Compare commits

..
251 Commits
Author SHA1 Message Date
zooqueenandhanzo-dev 1ba148db2f id: sync the repo that exists — hanzoai/id moved to hanzo-inc and went private
Hanzo CI/CD / cicd (push) Successful in 1m41s
CI/CD / cicd (push) Successful in 2m19s
This job fetched https://github.com/hanzoai/id.git and died with
`remote: Repository not found`, exit 128, on every run. The repo moved:
`gh api repos/hanzoai/id` resolves to hanzo-inc/id, and it is private.
GitHub answers 404 rather than 403 for a private repo a token cannot see,
so "not found" here means "no access", not "no such repo" — and a
member's SSH key reads it fine, which is how this stayed invisible from a
laptop while CI was blind.

The consequence was not cosmetic: the forge is what builds, so it fell
behind main and hanzo.id shipped stale source. It had already lost a
login fix ("honor a registration hint, so the signup funnel reaches
signup") before anyone noticed.

This points at the repo that exists. If the CI GH_PAT still cannot read a
private hanzo-inc repo, the job will now fail with a 404 naming the RIGHT
url, which is the honest next signal — the remaining half of the fix is a
token grant, not a code change.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 20:34:10 -07:00
zeekay 197be5fed7 Merge remote-tracking branch 'forge/main'
Hanzo CI/CD / cicd (push) Successful in 3m26s
CI/CD / cicd (push) Successful in 3m26s
Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 19:04:59 -07:00
zeekayandhanzo-dev f75a9e954d social: the bare arm signs in as the portal, never as the app
Clicking a social button with a downstream client_id on the query but no
redirect_uri sent IAM a hybrid the older model left behind: the app's
client_id paired with the portal's own /callback. No app registers that
pairing, so authorize answered "invalid redirect_uri" — and Callback,
which redeems as the portal client, could never have spent the code
anyway. createIam no longer takes a client override: an app that wants a
code arrives as a full authorize request and takes the forward arm.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 19:03:02 -07:00
zeekay 818e4c20ee Merge: offer Google first, with the provider order under test 2026-08-06 18:37:21 -07:00
zeekay 8e112b8a09 offer Google first, and put the order somewhere a test can hold it
Google is the account most people arrive already signed into, so it leads the
strip; the wallet trails as the specialist entry.

The order lived as an unexported constant inside the component that paints the
buttons, so nothing could assert on it and it had already drifted. It is
provider policy, so it moves to social.ts beside the other provider policy,
where it is a value a pure test can pin -- and one of the two new tests holds
the two policies in that module to each other: a key the strip renders must be
a key a provider_hint can resolve, or the console's one-click hand-off falls
back to the form for a provider plainly on screen.
2026-08-06 18:36:16 -07:00
zooqueenandhanzo-dev c473b3c851 read the client id the plain-link shape carries
hanzo.chat's homepage "Sign up" links to hanzo.id/signup/hanzo-chat -- the
client id in the PATH, from three live components (Visitor, MobileNav, the
error surface). App.tsx has always routed that shape (`path.startsWith
('/signup/')` and '/login/' are in its switch), so it is accepted by design.
The pages then read only `?client_id`, so the segment was matched and thrown
away.

What that produced: the page fell back to the host's default app, rendered a
form, and created the account under hanzo-console -- a DIFFERENT application
than the button that asked for it -- then left the new customer on hanzo.id
with no redirect_uri to return through. Signup "worked" and the funnel dropped
them.

One resolver, used by both pages. Query wins when both are present: it is the
OAuth request and the only one carrying a redirect_uri.

The segment must look like an IAM client id (<org>-<app>, the estate's one
naming rule) or it resolves to undefined and the host default applies, which is
exactly the previous behaviour. Tested: both shapes, precedence, and eight
rejects including a deeper path, wrong case, and a traversal attempt -- so
"accept anything" cannot pass. 175/175, typecheck clean.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 18:14:05 -07:00
zooqueenandhanzo-dev 2efb68dc30 sign out has to clear this browser, not just the IdP
Reported as "sign out is not working", and it half was. Portal navigated
straight to client.logout(), which builds the RP-initiated logout URL and does
nothing else. The server end is genuinely correct -- measured against prod, the
leftover token comes back 401 "the access token is invalid or revoked" -- so
this is not an access leak. What survived was the token STRING and its
siblings: every hanzo_iam_* key stayed in storage after signing out.

That is enough to read as broken, because the presence of that key is how other
surfaces decide you are signed in. hanzoai/playground's AuthGuard branches on
sessionStorage.getItem('hanzo_iam_access_token') before anything else. And a
token that outlives the session it names is worth deleting wherever it sits.

signOut() is now ONE call: clear, then hand back the IdP URL. A URL builder
that also mutates storage would be a trap, so logout() keeps its single job and
the complete operation gets its own name -- the call site asks for the outcome,
not the two halves.

Three details that are the bug in miniature:
  - Sweep the hanzo_iam_ PREFIX, never a list of literals. A list here is a
    second copy of the SDK's key set, and the copy is what goes stale when the
    SDK adds one. There are nine today.
  - Collect keys, THEN delete. Removing while iterating by index re-indexes the
    store and skips every other key.
  - Both storages: the token is in sessionStorage, but the PKCE verifier is in
    localStorage by design (it has to survive the redirect to the IdP).

Tested with a neighbour key that must SURVIVE, so "it cleared everything"
cannot pass; and the guard is known to bite -- neutering the prefix check fails
it by name, restoring passes. 172/172, typechecks clean.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 18:14:05 -07:00
zooqueenandhanzo-dev 25878f9d7a hanzo.id is ONE app, not whichever won the race
Reported from a phone: fill in email + password on hanzo.id/signup, press
Create account, get "the application does not allow to sign up new account".
Not reproducible on demand, which is the tell.

resolveOrg spreads the runtime catalog OVER the built-in table, and for
hanzo.id the two named DIFFERENT applications:

  built-in DEFAULT_TENANTS   hanzo-id        enableSignUp false
  catalog  /config.json      hanzo-console   enableSignUp true

So the answer to "can this person create an account" depended on whether a
second network fetch arrived before the form was submitted. When it did, signup
worked -- which is why it looks fine from a desk. When it did not, the page
still rendered a full form (the fallback app permits password login, just not
registration) and IAM refused the POST. On a phone a dropped request is
ordinary, so that is where it showed up.

Confirmed against prod by replaying the exact POST the SPA makes:

  clientId=hanzo-console application=hanzo-console -> past the gate
  clientId=hanzo-console application=hanzo-id      -> "the application does
                                                       not allow to sign up
                                                       new account"

The second line is the customer's screenshot, reproduced.

lux.id had the same divergence (lux-id vs lux-cloud); both are false there, so
it was latent rather than live. Fixed with hanzo's -- the invariant is one host
one app, not "the currently harmful ones".

The fix is NOT to make the fallback survivable. This table is a fallback for a
failed /config.json, not a second opinion, so the answer is that there is only
one answer per host. org.test.ts now asserts that resolving a host WITH and
WITHOUT the catalog yields the same clientId; reverting org.ts reproduces the
failure ("hanzo.id fallback", 1 failed | 15 passed) and restoring it passes,
so the guard is known to bite. 171/171, typecheck clean.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 18:14:05 -07:00
zooqueenandhanzo-dev da10820a18 password: an eye, so you can see what you typed
Every credential this portal takes was masked with no way to unmask. That is
the most common cause of a sign-in failure that looks like a wrong password,
and it is worst where typing is least reliable -- a phone keyboard, one glyph
at a time, with autocorrect and a shifted layout in the way. On SIGNUP it is
worse still: the account is created with whatever was actually typed, so a
typo behind the dots locks you out of an account you believe you just made.

ONE PasswordField, used by both forms. Login and signup had separate
hand-rolled <label>+<input> pairs; adding a toggle to each would have been two
answers to one question. Both now render the same primitive, so the field is
also the one place autoComplete is stated (current-password vs new-password --
getting that wrong is how a manager offers to "update" a saved credential
during a fresh signup).

Details that are the whole feature:
  - type="button". A bare <button> in a <form> defaults to submit, so the
    reveal would have submitted the form.
  - The button is OUTSIDE the <label>. A <button> inside one is activated
    twice -- as itself, and again by the label forwarding to its control -- so
    the toggle would fight itself. htmlFor keeps click-the-text-to-focus.
  - 44x44, the touch floor this stylesheet already holds everywhere else. An
    icon button only as big as its 20px glyph is a miss on a phone, which is
    the device this is for.
  - The input gives back 58px of right padding, so a long password never runs
    under the eye.
  - aria-pressed carries the state and the colour follows it, not :hover --
    on touch there is no hover, and "is my password on screen" must not
    depend on where a pointer is.
  - Icons live in ui/icons.tsx with the provider marks, inline and
    currentColor-driven. No icon dependency on the credential path.

Measured in a real browser on the built bundle, /login and /signup, at 390 /
768 / 1440 with the viewport actually resized between readings: button 44x44,
input 44 high, button inside the input box, no horizontal overflow, in all six.
Toggle across frames: password -> text -> password, aria-pressed false -> true
-> false. Suite 170/170, typecheck clean.

Default is masked. Nothing about the resting page changes.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 18:14:05 -07:00
zooqueenandhanzo-dev 569b1139c5 sync: the forge being ahead of GitHub is not a divergence
ff-main has been red every ten minutes. Eleven consecutive failures in the
hour I read it, and not one of them a divergence: the branch asked only
"is LOCAL an ancestor of REMOTE" and called every other answer DIVERGED,
so the forge simply being AHEAD of GitHub took the loud-failure path.

That is the normal state after every native push, and it is not brief --
the push-mirror is sync_on_commit with an 8h floor, so one commit can
produce up to 48 false alarms before GitHub catches up.

The cost is not the noise. A job that is always red cannot report the one
thing it exists to report, and this job's whole value is refusing to
force-push a real divergence. The loud failure is worth keeping only if it
is rare.

Four sibling repos (app, git, hanzo.ai, iam) already carry exactly this
two-direction check; id had the stale copy. Same logic, same order --
this is convergence onto the estate's answer, not a new one. HIPs,
mirrors and python-sdk are still stale and are being fixed with it.

Exercised against real git ancestry, all four states:

  state          old              new
  equal          in-sync(0)       in-sync(0)
  github-ahead   fast-forward(0)  fast-forward(0)
  native-ahead   DIVERGED(1)      native-ahead(0)   <- the fix
  diverged       DIVERGED(1)      DIVERGED(1)       <- alarm preserved

One row moves. The genuine-divergence alarm still fires, which is what
keeps this a fix rather than a mute button.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 18:14:05 -07:00
zooqueenandhanzo-dev 839c5d2c65 signup: ask for a password only where an account can be created
hanzo.id/signup rendered a full credential form for EVERY client_id, and
IAM refuses signup at SUBMIT — so on 48 of 51 hanzo applications a visitor
typed an email and a password and only then got "the application does not
allow to sign up new account". The provider buttons dead-ended the same
way: federation PROVISIONS a local user for a new identity, so a first-time
GitHub sign-up hit the same gate after a full round trip through GitHub.

enableSignUp was already on AppLogin and already fetched -- SignupForm reads
the same row inside onSubmit, too late to change what is on screen. Reading
it at the page turns the refusal from a surprise into a state.

It FAILS OPEN on purpose: an unreadable app config still renders the form and
the server still refuses. This is honesty about an answer we already have, not
a gate; internal/oidc/signup.go stays the only gate.

Measured against LIVE IAM, built bundle, all three directions:

  client_id               enableSignUp   <form>  input[type=password]
  hanzo-bot               true             1       1
  hanzo-chat              false            0       0
  does-not-exist-probe    unresolvable     1       1   (fail-open)

The middle row is the fix; the outer rows are what make it evidence rather
than a blanket removal of the form.

No new CSS: .hanzo-id-info already exists for exactly this.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 18:14:04 -07:00
zooqueenandhanzo-dev 93b64ae2d2 read the client id the plain-link shape carries
Hanzo CI/CD / cicd (push) Successful in 2m29s
CI/CD / cicd (push) Successful in 2m35s
hanzo.chat's homepage "Sign up" links to hanzo.id/signup/hanzo-chat -- the
client id in the PATH, from three live components (Visitor, MobileNav, the
error surface). App.tsx has always routed that shape (`path.startsWith
('/signup/')` and '/login/' are in its switch), so it is accepted by design.
The pages then read only `?client_id`, so the segment was matched and thrown
away.

What that produced: the page fell back to the host's default app, rendered a
form, and created the account under hanzo-console -- a DIFFERENT application
than the button that asked for it -- then left the new customer on hanzo.id
with no redirect_uri to return through. Signup "worked" and the funnel dropped
them.

One resolver, used by both pages. Query wins when both are present: it is the
OAuth request and the only one carrying a redirect_uri.

The segment must look like an IAM client id (<org>-<app>, the estate's one
naming rule) or it resolves to undefined and the host default applies, which is
exactly the previous behaviour. Tested: both shapes, precedence, and eight
rejects including a deeper path, wrong case, and a traversal attempt -- so
"accept anything" cannot pass. 175/175, typecheck clean.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 16:44:44 -07:00
zooqueenandhanzo-dev ffda55a2dc sign out has to clear this browser, not just the IdP
Hanzo CI/CD / cicd (push) Successful in 1m25s
CI/CD / cicd (push) Successful in 1m29s
Reported as "sign out is not working", and it half was. Portal navigated
straight to client.logout(), which builds the RP-initiated logout URL and does
nothing else. The server end is genuinely correct -- measured against prod, the
leftover token comes back 401 "the access token is invalid or revoked" -- so
this is not an access leak. What survived was the token STRING and its
siblings: every hanzo_iam_* key stayed in storage after signing out.

That is enough to read as broken, because the presence of that key is how other
surfaces decide you are signed in. hanzoai/playground's AuthGuard branches on
sessionStorage.getItem('hanzo_iam_access_token') before anything else. And a
token that outlives the session it names is worth deleting wherever it sits.

signOut() is now ONE call: clear, then hand back the IdP URL. A URL builder
that also mutates storage would be a trap, so logout() keeps its single job and
the complete operation gets its own name -- the call site asks for the outcome,
not the two halves.

Three details that are the bug in miniature:
  - Sweep the hanzo_iam_ PREFIX, never a list of literals. A list here is a
    second copy of the SDK's key set, and the copy is what goes stale when the
    SDK adds one. There are nine today.
  - Collect keys, THEN delete. Removing while iterating by index re-indexes the
    store and skips every other key.
  - Both storages: the token is in sessionStorage, but the PKCE verifier is in
    localStorage by design (it has to survive the redirect to the IdP).

Tested with a neighbour key that must SURVIVE, so "it cleared everything"
cannot pass; and the guard is known to bite -- neutering the prefix check fails
it by name, restoring passes. 172/172, typechecks clean.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 16:32:46 -07:00
zooqueenandhanzo-dev 21be2fecd7 base: spa 1.4.11 — asset 404s become no-store
Today's sign-in outage: during the 0.2.34 rollout a module fetch for the
new bundle hit an old replica, the bare 404 had no cache-control, and
Cloudflare's edge cached it for four hours. hanzo.id served a black page
with a healthy origin behind it. spa 1.4.11 marks asset-miss 404s
no-store (and reunites the spa lineage that shipped as 1.4.3-1.4.8 from
a branch no remote called main).

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 16:31:50 -07:00
zooqueenandhanzo-dev 573d3cdde2 hanzo.id is ONE app, not whichever won the race
Hanzo CI/CD / cicd (push) Successful in 1m37s
CI/CD / cicd (push) Successful in 1m39s
Reported from a phone: fill in email + password on hanzo.id/signup, press
Create account, get "the application does not allow to sign up new account".
Not reproducible on demand, which is the tell.

resolveOrg spreads the runtime catalog OVER the built-in table, and for
hanzo.id the two named DIFFERENT applications:

  built-in DEFAULT_TENANTS   hanzo-id        enableSignUp false
  catalog  /config.json      hanzo-console   enableSignUp true

So the answer to "can this person create an account" depended on whether a
second network fetch arrived before the form was submitted. When it did, signup
worked -- which is why it looks fine from a desk. When it did not, the page
still rendered a full form (the fallback app permits password login, just not
registration) and IAM refused the POST. On a phone a dropped request is
ordinary, so that is where it showed up.

Confirmed against prod by replaying the exact POST the SPA makes:

  clientId=hanzo-console application=hanzo-console -> past the gate
  clientId=hanzo-console application=hanzo-id      -> "the application does
                                                       not allow to sign up
                                                       new account"

The second line is the customer's screenshot, reproduced.

lux.id had the same divergence (lux-id vs lux-cloud); both are false there, so
it was latent rather than live. Fixed with hanzo's -- the invariant is one host
one app, not "the currently harmful ones".

The fix is NOT to make the fallback survivable. This table is a fallback for a
failed /config.json, not a second opinion, so the answer is that there is only
one answer per host. org.test.ts now asserts that resolving a host WITH and
WITHOUT the catalog yields the same clientId; reverting org.ts reproduces the
failure ("hanzo.id fallback", 1 failed | 15 passed) and restoring it passes,
so the guard is known to bite. 171/171, typecheck clean.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 16:20:30 -07:00
zooqueenandhanzo-dev ccca48b5c2 password: an eye, so you can see what you typed
Hanzo CI/CD / cicd (push) Successful in 1m6s
CI/CD / cicd (push) Successful in 1m7s
Every credential this portal takes was masked with no way to unmask. That is
the most common cause of a sign-in failure that looks like a wrong password,
and it is worst where typing is least reliable -- a phone keyboard, one glyph
at a time, with autocorrect and a shifted layout in the way. On SIGNUP it is
worse still: the account is created with whatever was actually typed, so a
typo behind the dots locks you out of an account you believe you just made.

ONE PasswordField, used by both forms. Login and signup had separate
hand-rolled <label>+<input> pairs; adding a toggle to each would have been two
answers to one question. Both now render the same primitive, so the field is
also the one place autoComplete is stated (current-password vs new-password --
getting that wrong is how a manager offers to "update" a saved credential
during a fresh signup).

Details that are the whole feature:
  - type="button". A bare <button> in a <form> defaults to submit, so the
    reveal would have submitted the form.
  - The button is OUTSIDE the <label>. A <button> inside one is activated
    twice -- as itself, and again by the label forwarding to its control -- so
    the toggle would fight itself. htmlFor keeps click-the-text-to-focus.
  - 44x44, the touch floor this stylesheet already holds everywhere else. An
    icon button only as big as its 20px glyph is a miss on a phone, which is
    the device this is for.
  - The input gives back 58px of right padding, so a long password never runs
    under the eye.
  - aria-pressed carries the state and the colour follows it, not :hover --
    on touch there is no hover, and "is my password on screen" must not
    depend on where a pointer is.
  - Icons live in ui/icons.tsx with the provider marks, inline and
    currentColor-driven. No icon dependency on the credential path.

Measured in a real browser on the built bundle, /login and /signup, at 390 /
768 / 1440 with the viewport actually resized between readings: button 44x44,
input 44 high, button inside the input box, no horizontal overflow, in all six.
Toggle across frames: password -> text -> password, aria-pressed false -> true
-> false. Suite 170/170, typecheck clean.

Default is masked. Nothing about the resting page changes.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 16:08:50 -07:00
zooqueenandhanzo-dev 5917e56a1e sync: the forge being ahead of GitHub is not a divergence
Hanzo CI/CD / cicd (push) Successful in 1m1s
CI/CD / cicd (push) Successful in 1m58s
ff-main has been red every ten minutes. Eleven consecutive failures in the
hour I read it, and not one of them a divergence: the branch asked only
"is LOCAL an ancestor of REMOTE" and called every other answer DIVERGED,
so the forge simply being AHEAD of GitHub took the loud-failure path.

That is the normal state after every native push, and it is not brief --
the push-mirror is sync_on_commit with an 8h floor, so one commit can
produce up to 48 false alarms before GitHub catches up.

The cost is not the noise. A job that is always red cannot report the one
thing it exists to report, and this job's whole value is refusing to
force-push a real divergence. The loud failure is worth keeping only if it
is rare.

Four sibling repos (app, git, hanzo.ai, iam) already carry exactly this
two-direction check; id had the stale copy. Same logic, same order --
this is convergence onto the estate's answer, not a new one. HIPs,
mirrors and python-sdk are still stale and are being fixed with it.

Exercised against real git ancestry, all four states:

  state          old              new
  equal          in-sync(0)       in-sync(0)
  github-ahead   fast-forward(0)  fast-forward(0)
  native-ahead   DIVERGED(1)      native-ahead(0)   <- the fix
  diverged       DIVERGED(1)      DIVERGED(1)       <- alarm preserved

One row moves. The genuine-divergence alarm still fires, which is what
keeps this a fix rather than a mute button.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 15:55:12 -07:00
zooqueenandhanzo-dev beb79dddca signup: ask for a password only where an account can be created
Hanzo CI/CD / cicd (push) Successful in 1m10s
CI/CD / cicd (push) Successful in 1m58s
hanzo.id/signup rendered a full credential form for EVERY client_id, and
IAM refuses signup at SUBMIT — so on 48 of 51 hanzo applications a visitor
typed an email and a password and only then got "the application does not
allow to sign up new account". The provider buttons dead-ended the same
way: federation PROVISIONS a local user for a new identity, so a first-time
GitHub sign-up hit the same gate after a full round trip through GitHub.

enableSignUp was already on AppLogin and already fetched -- SignupForm reads
the same row inside onSubmit, too late to change what is on screen. Reading
it at the page turns the refusal from a surprise into a state.

It FAILS OPEN on purpose: an unreadable app config still renders the form and
the server still refuses. This is honesty about an answer we already have, not
a gate; internal/oidc/signup.go stays the only gate.

Measured against LIVE IAM, built bundle, all three directions:

  client_id               enableSignUp   <form>  input[type=password]
  hanzo-bot               true             1       1
  hanzo-chat              false            0       0
  does-not-exist-probe    unresolvable     1       1   (fail-open)

The middle row is the fix; the outer rows are what make it evidence rather
than a blanket removal of the form.

No new CSS: .hanzo-id-info already exists for exactly this.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 15:52:05 -07:00
zooqueenandhanzo-dev e7864887b8 id: one name for the ingest key, and it is the one the estate uses
Hanzo CI/CD / cicd (push) Failing after 12m50s
CI/CD / cicd (push) Failing after 13m14s
This repo read EVENT_INGEST_KEY; hanzo.chat, hanzo.app, hanzo.ai,
platform and the shared hanzoai/ci `build_secrets` all read
PUBLISHABLE_KEY. Same key, same KMS path, two spellings — and this repo
was the last holdout, which is why the duplicate entry had to exist at
all.

That duplication is a live rotation hazard: two KMS entries holding one
value means rotating either one silently 403s whichever half still reads
the other. Collapsing to a single name is what lets the duplicate be
deleted.

Renamed end to end: the build-arg, the ENV, the Vite-inlined
VITE_PUBLISHABLE_KEY, the ambient type, hanzo.yml's build_secrets, and
every message that names the KMS address in a failure.

Also corrected a claim this file repeats and that turned out to be
false everywhere it appeared: an unkeyed beacon does NOT take an
"anonymous lane" that files rows under a $public tenant and answers 200.
Measured against the live door, with and without a browser Origin, for
pageviews and exceptions alike: a hard 401 ingest_key_required. Nothing
arrives. That myth is why "we still get errors, just not events" was
believed across four repos while keyless surfaces reported nothing at
all.

No behaviour change beyond the name: the fail-closed `case` gate and the
inlining assertion are untouched, and this repo's gate was already
sound — the case rejects an empty value before the grep, so it never had
the `grep -F ""` hole that let hanzoai/chat pass keyless builds.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 15:37:07 -07:00
zooqueenandhanzo-dev 313b27d537 fix(login): honor a registration hint, so the signup funnel reaches signup
Hanzo CI/CD / cicd (push) Canceled after 0s
CI/CD / cicd (push) Canceled after 38s
hanzo.app's "Get started" forwards `signup=true` on the authorize request and
this page ignored it. Measured end to end: the CTA lands on "Sign in to Hanzo ID"
with empty credentials, and appending `&signup=true` to the authorize URL by hand
changes nothing. Every net-new customer had to notice the small "Create account"
link to get past a form they could not fill in. The app was doing its part; the
IdP dropped the hint.

`screen_hint=signup` is honored too — the OIDC-standard spelling of the same
request, so a compliant client works without knowing our name for it.

It LOSES to silent SSO, and that ordering is the point. A browser already
carrying an issuer session belongs to someone who HAS an account, and sending
them to registration is worse than ignoring the hint. So this sits in `fallback`,
where `provider_hint` already sits, and only decides what happens when there is
no session to mint from.

The navigation carries the whole search string. Registration is its own page, and
client_id, redirect_uri, state and the PKCE challenge live in that query — a bare
`/signup` strands the new account with nowhere to return to. `replace` rather
than `assign` so Back reaches whatever sent the user here, not a login page that
would bounce forward again.

Both invariants are mutation-verified: short-circuit silent SSO and the ordering
test fails; drop the search string and the round-trip test fails.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-06 02:37:17 -07:00
zooqueenandhanzo-dev 6813e8a304 docs: the focus law, and the two-rule collision that hid under it
Hanzo CI/CD / cicd (push) Successful in 1m4s
CI/CD / cicd (push) Successful in 1m4s
The bullet claimed --ring was var(--neutral-500) at 4.43:1; it has been
var(--white-40) at 3.77:1 for the whole 0.4.x line. It also described base.css
as shipping ONE :focus-visible rule, which only became true at @hanzo/design
0.4.9 — before that a second, field-specific rule tied it at (0,1,0) and lost
to source order, so focused inputs here drew two indicators.

Records the part the next reader cannot derive: why a local focus override is
still the wrong fix, and why the treatment that rule was applying could not have
worked on this surface at all (its indicator rode on border-color, and
.hanzo-id-input states border unlayered).

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-05 06:46:47 -07:00
zooqueenandhanzo-dev 87ca9544a2 id: one focus ring, a CTA that answers the pointer, a lockup that stays 32px
Hanzo CI/CD / cicd (push) Successful in 1m31s
CI/CD / cicd (push) Successful in 1m33s
@hanzo/design 0.4.5 -> 0.4.9 fixes the focus indicator at the root. Two rules in
its base layer both computed to (0,1,0), so source order decided, and the generic
ring overrode the `outline:none` the field rule stated to prevent it: every
focused input on hanzo.id drew BOTH the 2px ring and a halo. Measured on the
shipped bundle, before 2 indicators and after 1. Nothing in this repo could fix
that — app.css deliberately keeps no focus rule of its own — and the same bump
carries it to lux.id, zoo.id and pars.id.

The rest is this surface's own:

The filled button had no hover. It has declared `transition: background` since it
was written and --primary-hover has always existed, so Sign in / Continue /
Create account animated a property nothing changed. `:not(.ghost)` keeps it from
tying with the ghost hover at (0,4,0), where order would otherwise pick the
winner — the same defect class as the ring above, so it is not written that way
here. Measured: #fafafa -> #e5e5e5 on hover, ghost unchanged.

The brand lockup is pinned to 32px. BrandHeader writes `height={32}`, but a
presentational attribute loses to design's `:where(img,video){height:auto}`, so
each portal was sized by whatever its brand package shipped. @hanzo/brand's SVG
has a viewBox and no width/height, so Hanzo landed on 32 and looked correct;
@luxfi/brand's declares 1024x1024, so lux.id has been rendering a 342px mark over
its sign-in form. Now 32x32 on all three, Hanzo unchanged.

min-height moves to 100dvh with 100vh kept underneath as the fallback: on a phone
`vh` is the URL-bar-retracted viewport, so a full-height page overflows the space
it actually has.

Styling and one dependency only — no flow, field, CSRF or handler is touched.
160 tests pass, tsc clean, all four brands verified.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-05 06:42:46 -07:00
hanzo-dev 05237b0470 the image installs what the tests ran
Hanzo CI/CD / cicd (push) Successful in 1m21s
CI/CD / cicd (push) Successful in 1m21s
hanzo.id has not published an image since the telemetry commit. The build
fails at `vite build` with

  Cannot find module '/build/apps/web/node_modules/vite/bin/vite.js'

which reads as a broken tree but is not: the same commit builds cleanly when
its dependencies come from pnpm-lock.yaml.

The image never used the lockfile. It was not copied, and the install ran
`--frozen-lockfile=false`, so the container re-resolved the entire graph from
the registry on every build while `pnpm test` on the runner installed from the
lockfile -- two different dependency graphs out of one commit, the tested one
and the shipped one. That agreed for as long as free resolution happened to
land on the same versions, which is not a property anyone was maintaining.
Adding @hanzo/event was enough to break the tie: the lockfile pins
vite@7.3.5_@types+node@25.9.3_..., free resolution chose
vite@7.3.6_@types+node@22.20.1, and the peer-suffixed store path the workspace
link pointed at no longer existed. The dependency did nothing wrong; it
disturbed a resolution that was never pinned.

So the lockfile ships and the install is frozen to it. The image now gets the
exact tree the tests passed against, and a stale lockfile fails at the install
naming the mismatch rather than quietly building a different application.

apps/account/package.json joins the pre-install copies: a frozen install
validates the lockfile against EVERY workspace member, and apps/account is in
the workspace even though this image does not build it.

Verified from a cold store on this exact commit: frozen install resolves the
pinned vite, `pnpm --filter @hanzo/id-web build` succeeds, and the ingest key
is present in apps/web/dist -- the assertion the next step makes.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-05 01:05:20 -07:00
zeekayandhanzo-dev ac70eb41d1 telemetry: report pageviews and errors, refused on auth-artifact routes
Hanzo CI/CD / cicd (push) Failing after 49s
CI/CD / cicd (push) Failing after 50s
hanzo.id is where every property's visitor lands and it reported nothing — the
live bundle carries no @hanzo/event, no /v1/event and no key — so the
arrival->session funnel had no denominator.

Mounts the ONE @hanzo/event client (product `id`, POST /v1/event) for pageviews
and errors, anonymously: no identify(), no interaction autocapture, no session
replay, no input capture, no email/name. A visitor's pre-signup pageviews still
join to whoever they become, because the client's per-browser anonymousId
survives sign-up and a post-auth surface identifies them there.

Telemetry is refused outright on /callback and /login/oauth/device. Passing a
clean pathname is NOT sufficient protection on those routes: @hanzo/event stamps
`url: window.location.href` onto every event it assembles, independently of the
`path` a caller passes, and its scrubber redacts secret SHAPES (JWT, sk-/pk-,
bearer, cloud keys, PAN) — an opaque OAuth authorization code is none of them.
Measured against the real client, a pageview from /callback?code=…&state=… put
the code and state on the wire while `path` read a tidy /callback. Neither route
is a funnel step; both are transient hops that redirect onward, and /, /login
and /onboarding still report. analytics.test.ts asserts BOTH that the gated
routes emit nothing and that the ungated client really does leak, so the gate
cannot decay into a decorative assertion.

The key is a publishable pk- read from KMS (deploy/EVENT_INGEST_KEY, env prod)
and passed as a build-arg. The build fails closed on an empty or
non-publishable value, then asserts the key actually reached dist: a green build
that ships an unattributed bundle is the failure mode here, not a red one.

No auth logic is touched — no OAuth/OIDC/PKCE, token storage, session restore,
or login template. 167 tests pass.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-04 23:48:20 -07:00
zeekayandhanzo-dev 18756a8ce2 ci: retire deploy.yml and the dead .github lane — one build path
Second half of the migration, and deliberately a separate commit: the canonical
lane was landed first and watched green while deploy.yml was still in place, so
this repo was never left with zero delivery.

Three files out, one line changed.

.hanzo/workflows/deploy.yml — superseded. The canonical lane builds the same
Dockerfile, reads the same package.json version, pushes ghcr.io/hanzoai/id, and
proves the manifest resolves before going green. It also runs the 160 tests and
the typecheck that this file never did.

.github/workflows/docker.yml — a second image lane. It had already been reduced
to an echo, but it kept the SHAPE of a build lane alive in a directory this forge
cannot even read (Gitea takes the first of WORKFLOW_DIRS present, and
.hanzo/workflows has existed here for months). Two lanes for one image is how the
two drifted in the first place.

.github/workflows/workflow-sanity.yml — imports
hanzoai/.github/.github/workflows/workflow-sanity.yml@main, which returns 404:
the reusable it calls does not exist. It could only ever have failed. It also
triggers on `paths: ['.github/workflows/**']`, a directory that is now empty, so
even a live version of it would be guarding nothing.

sync-from-github.yml keeps its cron and its fast-forward-only rule, and its
build dispatch is repointed deploy.yml -> cicd.yml. That reference is BY
FILENAME. Left alone it would have 404'd against the `|| echo` that makes it
non-fatal, so the symptom would not have been a red sync — it would have been
commits arriving on the forge and nothing building, silently, which is the exact
failure this repo already survived once (four commits, zero images, no red).

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-04 17:28:44 -07:00
zeekayandhanzo-dev 73346553ce ci: the canonical lane — root hanzo.yml + .hanzo/workflows/cicd.yml (0.2.27)
This repo had TWO image lanes and ZERO gates. 13 test files, 160 assertions, and
CI ran none of them — which is how pkgs/shared/src/org.test.ts sat RED on main
against code that is correct.

Two files. `hanzo.yml` says what this repo is (three gates, one image, no
deploy); `.hanzo/workflows/cicd.yml` is ~10 lines importing
hanzoai/ci/.hanzo/workflows/build.yml@v1. `.hanzo`, not `.github`: github.com has
zero self-hosted runners for this org, so a job asking for
hanzo-build-linux-amd64 there is never claimed — it waits out the 24h timeout
rather than failing. Gitea also reads only the FIRST of WORKFLOW_DIRS present in
the commit, so `.github/workflows` has been dark here all along.

deploy.yml is deliberately LEFT IN PLACE by this commit. It is the current green
lane and it does not come out until the new one has been watched green and shown
to publish the same image from the same source.

THE RED TEST, and why the fix is in the test. `org.test.ts` asserted that
`oauthCallbackOrigin` defaults to `iamIssuer`. That was the FIRST attempt at the
social-login fix and it was abandoned in the same change that shipped the real
one, because `hostSkeleton` derives the issuer from the REQUEST HOST — so on an
app host the issuer IS the brand host and the bug is unchanged. org.ts says
exactly this, in place, directly above the code. The assertion was simply left
behind.

It was also asking the question on hosts that carry no org at all: it called
resolveOrg('hanzo.app') and resolveOrg('hanzo.chat') with NO catalog, so both
fell to the deliberate unknown-host skeleton (empty orgId, fail closed, never
another brand's config). No design can make two ORG-LESS hosts agree on one
org's callback — that assertion could not have passed under either default.

So it now asks the question the way production does. The catalog (/config.json)
supplies the orgId, and the invariant that matters is ONE registered redirect_uri
PER ORG: hanzo.app, hanzo.chat, console.hanzo.ai and hanzo.id all resolve to
https://hanzo.id, none of them to their own host. Plus the assertion that pins
the abandoned attempt OUT: id.lux.network gets https://lux.id, which is NOT its
issuer — under the old default it would send its own origin and break again. A
strictly stronger guard than the one it replaces. 160/160 green.

org.ts also carried BOTH comment blocks — the superseded one ("The issuer is the
right default because...") still sitting above the one that corrects it. That is
one fact with two homes, and the stale half is the exact paragraph that would
talk the next reader into re-breaking social login. Deleted; the correct block is
untouched.

packageManager: pnpm@10.15.0 in package.json. corepack already reads it, so the
gate and the Dockerfile now resolve the SAME pnpm from ONE declaration instead of
the gate inheriting whatever corepack defaults to on the runner.

The gates: install (--frozen-lockfile, stricter than the image build's
--frozen-lockfile=false — lockfile drift should be caught in CI, not block a
build), typecheck (vite does NOT typecheck; `vite build` strips types, so a type
error ships a green image and this is the only thing that reads them across all 7
packages), unit (vitest, 160).

No `deploy:`, transcribing deploy.yml's own rule. id is governed by Hanzo CD:
the tag it runs is declared in universe (charts/app/values/hanzo/id.yaml, tag AND
digest) and reconciled back within ~60-90s, so a CI-side patch cannot stick — and
the reason to refuse it is not that it fails, it is that it LOOKS like it works.
It would also be actively wrong: the reusable's semver guard reads
infra/k8s/operator/crs/<svc>.yaml, a path this service does not use, so the guard
would not fire and a sha- tag would land on the live sign-in surface for every
property in the fleet.

0.2.27 / id-shared 0.1.4 — a release IS a version bump, and org.ts changed.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-04 17:08:39 -07:00
zeekayandhanzo-dev 23950b9ea4 onboarding: mint the bearer the writes need (0.2.26)
build / build (push) Successful in 39s
E2E caught the funnel dead-ending at consent: a password/form sign-in mints
a SESSION COOKIE and lands on /onboarding directly — the PKCE SDK holds no
token, update-user (rightly) refuses cookie-only writes, so every
saveOnboarding answered 401 and nothing ever persisted. The page now
bootstraps the token on mount: no SDK token → signinRedirect, which with a
live session is the silent-SSO authorize branch (no UI), /callback stores
the token and returns here. One bounce per session, guarded — a broken
mint degrades to the read-only state, never a redirect loop.

Also from the E2E: the plan step now SAYS when the catalog fetch failed
instead of silently rendering only pay-as-you-go ("there are no plans" is
false), and the popular plan wears a visible badge, not just a class.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-04 15:29:01 -07:00
zeekayandhanzo-dev 331675c945 onboarding: catalog prices are CENTS — pin the real contract (0.2.25)
build / build (push) Successful in 36s
The live catalog prices in cents (go=900 → $9/mo; priceAnnual=825 →
$8.25/mo billed annually ≈ $99/yr). 0.2.24's plan step would have rendered
"$900/mo". PlanInfo now carries priceCents/priceAnnualCents unscaled, the
UI formats ("$9/mo · $99/yr billed annually"), and listPlans keeps only the
personal+team product lines — dns-* and enterprise have their own surfaces.
Test pinned with production-shaped rows.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-04 14:43:48 -07:00
zeekayandhanzo-dev 5ab47aeb5d onboarding: consent + plan-or-payg steps, saved ONCE on the user (0.2.24)
build / build (push) Successful in 40s
The prepaid funnel's missing half. Two new steps close the flow:

- consent (required answer, either answer): the data-sharing agreement is
  asked exactly once and recorded on the USER (Properties, IAM row), so it
  never re-appears — on any browser.
- plan (LAST, required): pick a catalog plan or Pay as you go. Prices come
  from the pay catalog (GET <payUrl>/v1/billing/plans) — nothing hardcoded.
  The choice + completedAt persist BEFORE the redirect, so bouncing off the
  payment page never re-enters onboarding.

Completion gate: Onboarding.tsx reads onboarding.completedAt on mount and
skips straight to the portal for a user who already finished — the
DOES-NOT-REPEAT requirement, backed by the user row, not browser storage.

Handoff: plan slug → <payUrl>/cart?plan=<slug>; payg → <payUrl>/onboard
($5-minimum top-up). payUrl is a new optional OrgConfig field (catalog
override for white-label brands; defaults to https://pay.hanzo.ai).

Fixed underneath: every self-write now goes through ONE read-merge-write
(updateSelf). IAM's update-user is a FULL-ROW write that ignores the v1
columns= param, so the wallet step's minimal body was silently blanking
displayName/email on every link. Tests lock the round trip.

id-shared 0.1.3, id-onboarding 0.1.3, id-web 0.1.33, root 0.2.24.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-04 14:41:56 -07:00
zeekay fea1660363 merge origin/main
Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-04 11:22:27 -07:00
zeekayandhanzo-dev 5f01fb90f4 web: the sign-in control takes the system's edge and radius
@hanzo/design 0.4.2 put control rungs back on the alpha ladder, so
.hanzo-id-input draws --border-control (.15) instead of --border-strong,
which is now hover and emphasis — decoration a control must not reach
for. Its comment argued the opposite in every clause; rewritten. The 3:1
budget did not vanish, it moved to --ring, where a keyboard user actually
reads position. Lines 183 and 374 keep --border-strong: both are hover.

Radius follows the same recut: the field and the button under it were
6px while the system moved every control to 8px. Changing only the field
would have left two radii on two stacked controls in one 432px card, so
both move.

tokens.test.ts had to change to stay true. It asserted each tokens/*.css
was reachable through @import; 0.4.x flattens all ten groups into
styles.css so a bundler never resolves those subpaths, and the check went
red with nothing wrong. It now asks whether every token a group declares
is actually served, which is the cherry-picking this gate exists to catch
and survives however the package assembles itself. base.css needed its
own assertion — it ships element defaults as rules, and the one token it
declares is also in colors.css, so token coverage cannot see it go
missing; `@layer base` is its exact marker. Both paths proven by
reintroducing the defect and watching the right message fire.

Pins ^0.4.5: 0.4.2 and 0.4.3 shipped a stray */ that PostCSS rejects and
are now deprecated on the registry.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-04 11:20:24 -07:00
zeekayandhanzo-dev 0da7d5b33a fix(web): mobile floors + a hover that painted a wireframe (0.2.23)
Styling audit of iam.hanzo.ai. The headline finding is a NEGATIVE and it is the
useful part: this surface does not have the "classes without rules" defect that
broke hanzo.app, and structurally cannot — it uses neither Tamagui atomic nor
Tailwind, only hand-written semantic .hanzo-id-*. Measured in Chromium against
the served bundle, 7 routes x 3 viewports: gui atomic used 0 | rules 0, tailwind
used 0 | rules 0, and every rendered .hanzo-id-* class resolves.

An audit claim was WRONG and is corrected in LLM.md rather than "fixed": a build
with @hanzo/design absent does NOT silently emit a bare @import. Tried it —
Vite's postcss-import hard-fails with ENOENT and exit 1. The token layer cannot
vanish from this build. What was true is narrower: the tree was uninstalled, so
tokens.test.ts could not run. It passes once deps are present.

Four real defects, all verified by before/after measurement of the BUILT bundle
in a real browser (a green build is exactly what let the original defect ship):

- Ghost hover set border-color:var(--foreground) — #ededed, 17.9:1 — a
  near-white wireframe on the two most-hovered controls. The edge no longer
  moves; the surface carries the state, which the existing transition already
  animated. NOT simply dimmed: every rung brighter than --neutral-500 is worse
  on white (--neutral-400 is 8.3:1 on black but 2.52:1 on white) and would fail
  the WCAG 1.4.11 floor the resting border holds in BOTH themes. No token gets
  brighter on dark and darker on light, so the edge must not encode state.
- viewport-fit=cover was declared and never consumed — zero env(safe-area-inset)
  rules. Now max(24px, env(...)) per side; 0 insets still compute to 24px.
- The iOS-zoom comment described a protection the code did not implement:
  --text-base is 14px, so both credential fields zoomed on focus. 16px scoped to
  @media (pointer: coarse); the desktop ramp is untouched.
- Three tap targets under 44px, including a logo link whose target was SMALLER
  than the logo inside it (inline <a> = 32x18 around a 32px mark). Fixed with
  padding/negative-margin pairs — zero layout shift, screenshots pixel-identical
  at rest.

Measured BEFORE -> AFTER, built bundle served like prod:
  tap targets <44px           3 -> 0        ghost hover  #ededed -> #737373
  safe-area rules             0 -> 1        @media rules       1 -> 2
  input font-size (touch)  14px -> 16px     overflow           0 -> 0
  console errors              0 -> 0        atomic|tailwind  0|0 -> 0|0
  CSS transferred        16,733 -> 17,070 bytes (+337)   JS unchanged

The 5 page-variant classes with no rule (hanzo-id-login/-signup/-forgot/
-onboarding-page/-callback) are left alone: BEM-style hooks paired with
.hanzo-id-page, which carries the layout — the same pattern as .hanzo-id-device.

Pre-existing and NOT touched: pkgs/shared/src/org.test.ts fails on origin/main
(hanzo.app vs hanzo.chat), verified on a pristine checkout; it belongs to the
in-flight callback work.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-04 10:19:14 -07:00
hanzo-dev 787352b0eb fix(shared): restore catalogOf — c153004 deleted it with its own tests
The id image has failed to build since c153004 with

  "catalogOf" is not exported by "../../pkgs/shared/src/index.ts"

which is why 0.2.22 never existed and every fix behind it stayed unshipped.

c153004 removed catalogOf from org.ts AND removed its tests from org.test.ts in
the same commit, while apps/web/src/App.tsx kept importing it from the barrel.
Deleting a function together with the tests that pin it removes the very thing
that would have reported the deletion — the failure surfaced two steps later, in
a container build, as a rollup resolution error.

catalogOf owns which key the /config.json payload uses, and the key is the
SERVER'S name (iamTenantConfigJson). Reading any other key returns undefined,
App.tsx falls back to a window global the runtime never injects, and every
catalog-only host silently drops to the bundled defaults — a total catalog
outage that looks like nothing at all. Verified against the live payload:
https://hanzo.id/config.json serves {"iamTenantConfigJson": "...", "v": 1}
with 13 hosts.

Tests restored alongside, plus one the old set lacked: that catalogOf is
reachable from the package BARREL, which is how App.tsx imports it and the exact
edge that broke the build.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-03 18:11:09 -07:00
hanzo-dev 7c6f148435 docs(llm): handoff — social login root cause, what shipped, what is left
Everything measured against production. Records the one-line cause of the
social-login outage (oauthCallbackOrigin defaulted to the brand host, so every
property sent a different redirect_uri), the two IAM fixes reconciled into one
SSO seam, the four method errors that cost real time this session, and the
ordered list of what remains — including the platform build enqueue being down
on a missing org row and the ruling to delete DEFAULT_BUILD_ORG_ID.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-03 17:52:51 -07:00
hanzo-dev f2fc1e4539 release: id-shared 0.1.2 and the three packages that carry it
c153004 put the one-provider-callback fix in @hanzo/id-shared, but nothing
downstream can consume a fix that was never published. The internal deps are
`workspace:*`, which pnpm rewrites to the concrete version AT PUBLISH TIME — so
id-shared alone going out leaves id-auth, id-onboarding and id-idv still
resolving 0.1.1 for anyone installing from the registry.

  @hanzo/id-shared      0.1.1 -> 0.1.2   the fix itself
  @hanzo/id-auth        0.1.6 -> 0.1.7   consumes it (SocialButtons, client)
  @hanzo/id-onboarding  0.1.1 -> 0.1.2   consumes it
  @hanzo/id-idv         0.1.0 -> 0.1.1   consumes it
  @hanzo/id-web         0.1.30 -> 0.1.31 private app, kept in step
  @hanzo/id             0.2.21 -> 0.2.22 workspace root

Patch throughout: the change is a corrected DEFAULT, and any consumer that had
already set oauthCallbackOrigin explicitly is unaffected — the explicit value
still wins. Nothing here justifies a minor, let alone a major.

@hanzo/id-connect is deliberately NOT bumped: it does not depend on id-shared.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-03 17:33:42 -07:00
hanzo-dev c15300493c auth: one provider callback per org, and it is the hosted ID host
Social login was broken on every Hanzo property for days, and it was never the
credentials. Google's own error payload, decoded, reads `redirect_uri_mismatch`
— our client_id reached Google intact every time.

Google and GitHub each hold ONE OAuth client with a FIXED list of authorized
redirect URIs. `oauthCallbackOrigin` defaulted to `publicOrigin` — the brand's
OWN host — and no catalog entry overrode it. So hanzo.app sent
hanzo.app/callback, hanzo.chat sent hanzo.chat/callback, console sent
console.hanzo.ai/callback, and social login could work on at most ONE property,
whichever happened to be registered. Every new brand arrived broken by
construction, and the failure surfaced at the provider rather than in our logs,
which is why it read as a KMS/secrets problem.

The default is now the org's hosted ID host, read out of DEFAULT_TENANTS so the
`.id` hosts stay declared exactly once:

  hanzo.app / hanzo.chat / console.hanzo.ai / cloud.hanzo.ai -> hanzo.id/callback
  id.lux.network                                             -> lux.id/callback

A social provider therefore never learns about individual apps. hanzo.id
completes the exchange and forwards the browser back to the originating app.

FIRST ATTEMPT WAS WRONG, recorded so nobody repeats it: defaulting to
`iamIssuer` does NOT fix this. hostSkeleton derives the issuer from the REQUEST
HOST too, so it is per-brand for exactly the same reason. It has to be a
per-ORG constant. Caught by executing resolveOrg rather than reasoning about it.

An unknown host still falls through to the host-derived skeleton with orgId '',
so the fail-closed property that keeps a Zoo visitor off Hanzo's login is
untouched.

Two tests pin the default (there were none, which is how it drifted). They are
NOT run here: this checkout has no vitest installed and @hanzo/id-shared
declares no `test` script — `pnpm --filter @hanzo/id-shared test` exits 0 having
run nothing, which is its own false green. Behaviour verified by executing
resolveOrg directly under tsx against a production-shaped catalog.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-03 17:32:09 -07:00
zooqueenandhanzo-dev a39ebf4b4e docs: oauthCallbackOrigin is dead config, and its comment argues for the bug
Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-03 14:00:54 -07:00
zooqueenandhanzo-dev 36fb502b13 docs: the catalog key, and why a rename did not stop at the language boundary
Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-03 13:59:13 -07:00
zooqueenandhanzo-dev 75ea9b6bae config: read the catalog key the runtime actually serves
build / build (push) Successful in 46s
/config.json is emitted by hanzoai/spa, which derives its key mechanically
from the env var universe supplies: SPA_IAM_TENANT_CONFIG_JSON becomes
iamTenantConfigJson. Renaming TenantConfig -> OrgConfig in this codebase
also renamed the READ to iamOrgConfigJson, a key nothing emits.

So the catalog has been undefined on every host for as long as that
shipped, with no error anywhere — the fetch succeeds, the field is
absent, and resolveOrg falls back. Verified against the live endpoint:
https://hanzo.id/config.json returns iamTenantConfigJson, and the SPA
was reading past it.

Two consequences, both live:

hanzo.id resolved to the BUILT-IN hanzo-id (enableSignUp:false) instead
of the catalog's hanzo-console (true). A first-time federated user is
refused "the application does not allow to sign up new account" — so
this gates the GitHub fix in 0.2.20 for exactly the users it was for.

Every catalog-ONLY host — osage.id, zoolabs.id, id.zoo.network,
id.lux.network, iam.lux.network, id.pars.network, id.bootno.de,
iam.hanzo.ai — got the empty-clientId skeleton and could resolve no
application at all. The resolver's fail-closed behaviour is correct and
is why this leaked no brand; it is also why it was silent.

The key now lives in one place, catalogOf, next to the resolver it feeds,
with the reason it is not ours to rename. Guarded by two tests that fail
on the old spelling and pass on the new.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-03 13:56:11 -07:00
zooqueenandhanzo-dev 7183bb69c8 social: the browser was the relying party, and it cannot be one
build / build (push) Successful in 54s
GitHub sign-in succeeded at GitHub and dead-ended here — "you need to
login first", which is IAM's 401 from /v1/iam/onboard reached with no
session and no bearer. The SPA never obtained a token.

social.ts built the IdP URL in the browser and pointed it at our own
/callback, so GitHub returned a GitHub code that NOTHING can spend:
exchanging it needs the client secret, which a browser must never hold,
and IAM has no endpoint that takes a raw provider code. Callback.tsx
posted it to /v1/iam/login, which knows only password, device approval
and code minting.

IAM has implemented this correctly the whole time and was never called.
Naming a provider on the authorize endpoint IS the entry point: it
validates client_id, the exact redirect_uri and PKCE, then federates
server-side and returns an IAM code bound to the original challenge.
OAuthAuthorizeRequest.provider was already declared here; authorize()
simply never emitted it. That one parameter is the fix; the rest of this
diff deletes what existed to work around its absence.

The name is the record name — provider-github, never github:
federationProvider matches ProviderItem.Name exactly. A comment claimed
the opposite; corrected in place.

Two arms, the two the password path already branches on, differing only
in who owns the PKCE verifier: an app-initiated request is re-entered
verbatim so IAM mints against that app and returns the browser straight
to it; a bare portal sign-in goes through the SDK, which persists the
verifier in localStorage keyed by state so it survives the redirect.

Gone as dead: the IdP endpoint table, buildProviderAuthUrl,
startProviderLogin, isHoppableProvider, encodeState/decodeState,
client.providerLogin, ProviderExchangeRequest, and Callback's provider
branch. /callback now has one case, because a federated return is
indistinguishable from any other.

Verified in a browser against live IAM: both arms reach GitHub via
/v1/iam/oauth/authorize with redirect_uri=<brand>/v1/iam/oauth/callback
and the hanzo_fed cookie; the app arm creates no verifier slot, proving
the app's own request was forwarded; the return leg consumes the slot
and posts /v1/iam/oauth/token, which answers invalid_grant for a
deliberately fake code. The GitHub login itself is not exercisable here.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-03 13:49:22 -07:00
hanzo-dev 2fe32a731e auth: one provider list, not two — delete the one nothing could render
pkgs/auth/src/ui/ProviderButtons.tsx declared its own provider map:

  google: { label: 'Continue with Google', brand: 'google' },
  github: { label: 'Continue with GitHub', brand: 'github' },

Two providers, where SocialButtons renders four (ORDER = github, gitlab,
google, web3). It was imported by nobody, absent from pkgs/auth/src/ui/index.ts
and therefore never part of the package's public API, and unreferenced across
every repo in the estate. It could not render, so it could not be wrong in
production — it could only be COPIED, and the copy would be missing GitLab and
Wallet from the day it was made.

Login.tsx, Signup.tsx and DeviceApproval.tsx all use SocialButtons, which is
the correct design and stays: it resolves {application, providers} from IAM at
runtime and intersects with PROVIDER_META, so the buttons follow the
application's configuration rather than a hardcoded list.

social.ts is NOT a third list and is untouched: its provider map holds OAuth
authorize ENDPOINTS for buildProviderAuthUrl (11 call sites), which is where to
send the user, not what to show them. All six of its exports are live.

NOTE, and this commit does not fix it: the live asymmetry is IAM
per-application config, not this file. Measured against production in one
browser context —

  page                                    GitHub GitLab Google Wallet
  sign-in                                   y      y      y      n
  sign-up via client_id=hanzo-app           y      y      y      y
  sign-up direct at hanzo.id/signup         y      n      y      y

GitLab is enabled on the hanzo-app application and not on the portal default,
and Wallet renders only under the signup intent. Both are configuration, and
both need a decision about what the canonical provider set is per application.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-03 07:51:00 -07:00
zooqueenandhanzo-dev 1ac8c599b4 device approval: the header described a flow that does not exist
build / build (push) Successful in 36s
Three errors, each the kind that sent someone down a wrong path today.

`verification_uri_complete` does not add `?user_code=<code>`. IAM appends the
code as a PATH segment — `/login/oauth/device/<code>` — because that is the
route this page is registered on (device.go). readUserCode accepts the query
form too, so both work, but the minted URI is the path one.

Approval does not flip `UserSignIn=true`. No such field exists on the row.
approveDevice binds the approver onto `Token.User` as owner/name, and an EMPTY
User is what "not yet approved" means — that is the whole state machine, and a
reader looking for a boolean will not find one.

The CLI is `hanzo login`, not `dev login --device-auth`.

Also re-runs the 0.2.19 build: the previous one died in the oci.hanzo.ai login
step on a transient 502 (the registry answers 401 again, i.e. healthy), so no
image was ever pushed for this version and nothing is being re-tagged.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-01 20:52:51 -07:00
zooqueenandhanzo-dev 13413d41a6 0.2.19 — the device page names the right client, and the checkbox ships
build / build (push) Failing after 12s
0.2.18 was cut, built and published BEFORE the two fixes it is named after
landed. The build workflow refuses to rebuild an existing version — correctly,
since re-pushing a tag yields two digests for one name — so 5d411ff (drop the
confirmation checkbox) and everything after it never produced an image. The tag
in universe stayed 0.2.18 and the cluster kept serving the bytes it already had,
which is why the platform owner has asked four times for a checkbox that has
been gone from source the whole time.

A version is the only thing that publishes. Patch-bump, never a re-push.

Ships: the self-attestation checkbox removal (5d411ff), and the approval page
naming the client that actually minted the code (470aa48).

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-01 20:43:40 -07:00
zooqueenandhanzo-dev 470aa48408 device approval: name the client from the code, not the portal
The page said "You are about to authorize hanzo-console" for a sign-in started
by hanzo-cli. appLabel was org.appName — this PORTAL's own branding, a static
per-brand string — so every device code, whichever client minted it, was
approved under the same wrong name. The one job of the screen is to say WHICH
application you are authorizing, and a screen that names the wrong one does not
merely fail to help: it teaches people the name means nothing.

bd04657 removed the false name but concluded the right one could not be learned
without building a code-hunting oracle. That was true of an open lookup and is
not true of the one IAM now serves: POST /v1/iam/oauth/device/info is session
gated, org scoped, and answers unknown / expired / already-approved with a
single opaque refusal, so it reveals strictly less than the approval the same
caller could already attempt. The comment arguing the endpoint must not exist is
replaced by what is actually true of it.

The code rides the POST BODY. A user_code is the one secret in this flow and a
request line is copied into ingress and proxy access logs where a body is not —
this page already ships scrubUrl() to keep the code out of the address bar, and
a GET would have undone that on the server side.

Approve is gated on a resolved name and fails closed: no server-confirmed
application, no button. Nothing local is ever substituted, because rendering a
guess is the defect being fixed. THIS MAKES THE PORTAL DEPEND ON IAM b466bd63 —
ship IAM first or together, or every approval blocks.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-01 20:43:26 -07:00
zooqueenandhanzo-dev bd04657638 fix: the consent screen named the wrong client, and we have orgs not tenants
APPROVING THE WRONG NAME. The device page read "You are about to authorize
hanzo-console" for a sign-in started by hanzo-cli, because appLabel was
tenant.appName — this PORTAL's own branding, a static string, not the client that
asked. The backend was never confused: approveDevice binds the pending row's own
application, and says so ("the portal app the browser happens to be on is
irrelevant to WHAT is being approved").

The page cannot honestly name the client, and must not learn it: the user_code is
40 bits and the one secret in this flow, so an endpoint answering "which app is
this code for" would be the oracle device.go refuses to be — it returns one opaque
refusal for unknown / expired / already-approved precisely so live codes cannot be
hunted. So the page stops naming a party it cannot vouch for and asks about the
thing it can: the code the human transcribed off their own device. A consent screen
that names the wrong party is worse than one that names none — it teaches people
the name means nothing.

ORGS, NOT TENANTS. There is no tenant concept here; the identifier has always been
the org (the JWT `owner` claim, the IAM `<org>-<app>` namespace). TenantConfig →
OrgConfig, resolveTenant → resolveOrg, tenant → org, and the module renamed to
match. 28 files, no residual `tenant`, typecheck clean, 145 tests pass.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-01 20:32:41 -07:00
zooqueenandhanzo-dev 793b9b37bc device approval: stop naming the wrong app, and say what the page can vouch for
It read 'You are about to authorize hanzo-console' for a sign-in started by
hanzo-cli: appLabel was tenant.appName, this PORTAL's own branding, never the
client being approved. A consent screen that names the wrong party is worse than
one that names none — it teaches people the name means nothing.

It cannot honestly learn the right name either, and must not: the user_code is
40 bits and the only secret in this flow, so an endpoint that revealed which app
a code belonged to would be an oracle for hunting live codes — exactly what
device.go refuses to be, answering unknown, expired and already-approved with
one opaque refusal.

So the prompt asks about the thing this page CAN vouch for: the code in front of
the human, and whether they started the sign-in themselves. That is also where
the removed checkbox's intent belongs — in the sentence someone reads before
clicking, not a tickbox in front of it.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-01 20:30:38 -07:00
zooqueenandhanzo-dev 255c9622fb 0.2.18 — device approval drops the confirmation checkbox
A release is a version bump; 0.2.17 already shipped the signup fix and its image
is published, so the checkbox change needs its own.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-01 20:22:40 -07:00
zooqueenandhanzo-dev 5d411ff26a device approval: drop the confirmation checkbox
No device page anyone actually uses has one — Google, GitHub and AWS all show
the code and an Approve button. It sat in front of the only action on the screen
and disabled it until ticked.

It was there as an anti-phishing gate, on the theory that a victim arriving from
a crafted verification_uri_complete link must not approve with one click. A
tickbox is not evidence of that: someone being walked through a crafted link
ticks it exactly as readily as they click Approve. The property that does the
work is the one that stays — the code is displayed and remains EDITABLE, so
approving is an explicit act on a value the human can read and correct against
what their own device shows.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-08-01 18:47:36 -07:00
Antje Worringandhanzo-dev c78eddd735 signup must end in a session — new customers were stranded on the portal
build / build (push) Successful in 3m7s
A customer who clicked "Create account" from an app never came back. Two
defects composed, and every unauthenticated probe read healthy through both.

`Login` linked to a bare `/signup`, so the OIDC request the app sent —
client_id, redirect_uri, state, PKCE challenge — was dropped at a full page
load. `Signup` already parsed exactly those params; nothing ever supplied
them.

Then IAM's signup is CREATE-ONLY. It persists the user and answers with the
row: no session cookie, no authorization code. The `autoSignin: true` the
client posted has no field in `internal/oidc/signup.go`, so the Go decoder
dropped it — "signed up" and "signed in" were never the same event. That
response fell through `parseLoginResponse`'s no-redirect arm to
`{redirectUrl: '/onboarding'}`, so the new user was sent to the portal's own
onboarding, unauthenticated, while the app waited on a code no one minted.

The links now carry the request, and `signup()` finishes the job through the
one path that establishes a session: it logs in, which mints the code bound
to this request's challenge and returns the redirect back to the app. A
signup that leaves you logged out is not a signup.

`parseCreated` reads the create leg, deliberately apart from
`parseLoginResponse` — that one INVENTS a destination when no redirect_uri
was asked for, which is right for a session and wrong for a row.

Registration also resolves the app via `get-app-login` the way `LoginForm`
does, so the account is created under the app the user came FROM rather than
the portal's own tenant.

Contracts locked in client.test.ts; both regressions verified by reverting the
fix and watching them fail. Driven end to end in Chromium against the built
SPA with IAM stubbed at its real response shapes: the params survive the
click and the browser lands on the app's callback holding the code.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-30 20:05:05 -07:00
zeekayandhanzo-dev 1e919446b6 0.2.16 — read IAM's named envelope fields before legacy data2
The IAM envelope's untyped data2 slot splits into named fields — total
(count) / org (masked org) / mfa (challenge list). id touches two of the
meanings: the login MFA challenge allow-list (client.ts) now reads the
named mfa field first, and onboarding's list rows stay on the named data
slot; both keep the data2 fallback, deleted once IAM stops emitting it.
The MFA test table pins both spellings plus named-first precedence, and
listOrgs pins the legacy list slot.

@hanzo/id-auth 0.1.5, @hanzo/id-onboarding 0.1.1.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-30 10:59:28 -07:00
zooqueenandhanzo-dev 66c6ebee63 docs: name the identity server IAM in these comments
hanzoai/iam is Hanzo IAM — original, clean-room work; the vendor-derived server
was hanzoai/iam-v1, retired and in nothing we ship. The comments here named that
vendor for behavior that is IAM's own: get-account really does return the User
at the top level or under `data`, and add-organization really is entity CRUD
behind the authenticated Guard. Kept the facts, dropped the wrong name.

Comments and markdown only.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-28 17:13:43 -07:00
zandhanzo-dev ce9de2e9cb ci: log in to ghcr with GHCR_USER/GHCR_TOKEN, the pair that works
The build has failed every run on

  ::error::Error response from daemon: Get "https://ghcr.io/v2/": denied: denied

That is a LOGIN failure, not a push-permission one. It used GH_PAT with a
hardcoded username; every image build in this org that currently succeeds
(auto, base, sign, ui, papers) uses the GHCR_USER/GHCR_TOKEN org-secret pair
instead. The token identity is the only thing separating this job from those.

Checked before changing it, so this is not a guess about the registry side:
ghcr.io issues push,pull tokens for repository:hanzoai/id to the hanzo-dev
credential the cluster holds, and that account carries write:packages. So
neither the registry nor the target package rejects us — the credential this
step presents does.

One consequence worth recording: the hanzoai/id PACKAGE does not exist on ghcr
yet (the org has cloud, visor, vm), so the first successful push CREATES it.
That needs write:packages rather than read, which is exactly the kind of scope
gap a stale PAT produces.

Same registry, same tags, same build — only the credential changes.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-28 09:54:56 -07:00
zandhanzo-dev dddb1bb637 ci: the oci.hanzo.ai pull credential exists now — correct the note that says it does not
This comment described a blocker that has since been cleared, and it is the kind
of stale note that causes harm rather than merely being out of date: read
literally it argues AGAINST pointing an App CR at oci.hanzo.ai, because it says
the cluster cannot pull from there.

It claimed registry-credentials "currently only carries registry.hanzo.ai — so a
CR pointed at oci.hanzo.ai today would ImagePullBackOff". Measured against the
live cluster:

  registry-credentials auths -> registry.hanzo.ai (auth), oci.hanzo.ai (auth)
  App CR id                  -> oci.hanzo.ai/id:0.2.13
  pods                       -> 2/2 Running, 39h old, on that image

So the flip already happened and works; only the note lagged.

ghcr is still published to, and that is now stated as the actual reason —
a public mirror and a fallback — rather than as a workaround for a pull
credential that is no longer missing. Whether to drop it is a separate decision
this does not pre-empt.

No behaviour change: same two tags, same build.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-28 09:41:37 -07:00
zandhanzo-dev 06a360e71e merge github/hanzoai/id — reconcile five months of parallel history
build / build (push) Successful in 35s
The ff-main sync has been failing every 10 minutes. It was not broken: the forge
and GitHub genuinely diverged at 6fd3a63 on 2026-03-06 and never rejoined —
150 commits one side, 163 the other — and the job correctly refuses to force
either history away.

Direction was decided by what production actually runs, not by which remote is
labelled canonical:

  live (App CR)              oci.hanzo.ai/id:0.2.13
  forge/main version line    … 0.2.9, 0.2.10, 0.2.11   (stops)
  github/main version line   … 0.2.11, 0.2.13, 0.2.14, 0.2.15

0.2.13 exists ONLY on the GitHub side. Production was built from that lineage,
so where the two disagree GitHub wins.

The conflict list looked far worse than it was. 19 of the 21 conflicts are
add/add — files absent at the merge base that BOTH sides then created
independently (the whole auth UI: LoginForm, OTPForm, SignupForm, ForgotForm,
SocialButtons, OnboardingFlow, Portal, DeviceApproval). Only package.json and
pnpm-lock.yaml were edited on both sides from a common ancestor. The parallel UI
implementations differ mainly by class naming — `hanzo-id-login-form` vs the
systematic `hanzo-id-form` / `hanzo-id-field` / `hanzo-id-input` from GitHub's
design-token pass — so taking GitHub keeps the version that shipped.

One conflict settles itself on evidence rather than preference:

  forge:   ${{ github.server_url }}/api/v1/repos/…
  github:  ${{ github.server_url }}/v1/repos/…

This forge serves /v1 and 404s /api/v1 — measured today against the Service
directly, bypassing ingress. The forge side of its own sync workflow had the
path wrong; GitHub's is correct.

Nothing from either side is lost. 26 files auto-merged, and the forge-only work
is intact — verified by ancestry, including 3b06747 "a host with no entry must
fail closed, not become Hanzo" and its regression test, which is a brand-leak
security fix.

Verified before committing: `pnpm install --frozen-lockfile` succeeds (so the
merged lockfile and the merged package.json files agree), `pnpm -r build`
completes, and `pnpm test` is 139 passed / 139 across 13 files.

This does NOT deploy anything. The App CR still pins 0.2.13; reconciling git and
choosing to ship are separate decisions, and only the first is made here.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-28 05:37:03 -07:00
zooqueenandhanzo-dev 23c5b47eea 0.2.15 — take the whole token layer, mount the ONE account control, gate on resolution
0.2.14 adopted @hanzo/design and worked around it in three places. Those
findings are fixed IN @hanzo/design 0.3.0, so the workarounds go and this
surface takes the system's answer.

- ONE import: `@hanzo/design/styles.css`. app.css cherry-picked four of nine
  token groups, so --z-*, --shadow-*, --space-*, --font-* and the element
  defaults did not exist here. Nothing broke visibly, because an unresolved
  var() paints nothing and reports no error. @hanzo/iam's account menu alone
  reaches for --z-popover, --shadow-floating and --space-1..3.

- Geist is self-hosted in 0.3.0, so the reason for cherry-picking is gone:
  fonts.css no longer requests fonts.googleapis.com. Measured on the built
  bundle: rendered face is Geist, zero third-party font requests.

- The :focus-visible rule is deleted from this file. tokens/base.css ships it
  to every consumer and --ring is now var(--neutral-500): measured 2px solid
  rgb(115,115,115), 4.43:1 on --background (WCAG 2.4.11 wants 3:1). The local
  --primary override existed only because --ring was 1.66:1.

- Control boundaries are --border-strong, not --white-40. 0.3.0 moved it to
  var(--neutral-500) — 4.43:1 in BOTH themes. --white-40 cleared 3:1 on black
  and measures 1.00:1 on white.

- @hanzo/iam 0.13.1 -> 0.21.1 in apps/web, pkgs/auth, pkgs/onboarding, and the
  signed-in portal mounts <UserMenu> in place of a hand-rolled Billing/Sign out
  link row. Identity comes from resolveIdentity, so the portal cannot disagree
  with the console about who you are. No `brand` prop: omitting markSvg would
  put the Hanzo mark on lux.id and zoo.id.

- Portal.tsx's inline style object is gone — rgba(255,255,255,0.14),
  borderRadius 12, fontSize 13, fontWeight 600 and two bare opacities, six
  invented values for facts the token layer already states.

- apps/web/src/tokens.test.ts gates RESOLUTION, not declaration. The reference
  is built at runtime from a string, so it is invisible to the compiler and to
  grep. Verified by reintroducing both original defects: it names the 5 dropped
  groups and the 8 unresolved iam tokens.

Measured in Chromium, fresh context, 1440 and 390, on the built bundle served
as hanzo.id / lux.id / zoolabs.id: 0 unresolved of 76 tokens referenced on every
host; input edge and focus ring both 4.43:1; the account menu paints #0a0a0a /
12px / --shadow-floating / z-index 700. Brand switches with zero per-brand code
and no Hanzo mark reaches the Lux menu.

Open, against layers below this surface: the menu's own panel edge is --border
at 1.27:1 on --popover (0.3.0 raised --ring and --border-strong, not --border),
and BrandHeader loads the mark from cdn.jsdelivr.net/@latest — a third-party
request on the sign-in path, which is what this file refuses for fonts.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-27 17:49:37 -07:00
zeekayandhanzo-dev c2db6cc171 fix: complete the TypeScript 7 tsconfig migration
The first pass left three classes of config that BOTH tsc 5.9 and tsc 7
reject. Each was proven against both compilers before changing:

  TS5090  A `paths` target must be relative once `baseUrl` is gone. The
          first pass skipped the "./" prefix wherever baseUrl pointed at
          the config's own directory, reasoning it was semantically
          equivalent. It is not — without baseUrl a non-relative target
          is rejected outright, by 5.9 as well as 7.

  TS5110  `moduleResolution: node16` requires `module: node16`. The first
          pass mapped commonjs projects to node16 resolution alone, which
          BROKE those configs for the current toolchain. Both are now set.

  TS5102  `downlevelIteration` is also removed in TS7; it was missing from
          the dead-flag list.

Verified: repos that tsc 7 previously refused (base-studio, js-sdk, kv-js)
now report zero config errors on tsc 7 AND tsc 5.9.

Co-Authored-By: Hanzo Dev <dev@hanzo.ai>
2026-07-27 17:25:55 -07:00
zeekayandhanzo-dev 9c17d55ebb build: migrate tsconfig to TypeScript 7 (native compiler)
TypeScript 7 is the native Go compiler and removes `baseUrl` and
`moduleResolution: node|node10`. Both appear here, so `tsc` from TS7
refuses the config outright (TS5102 / TS5108) and cannot typecheck.

`paths` targets resolve relative to `baseUrl` when it is set and relative
to the tsconfig file otherwise. Every `baseUrl` folded here already
pointed at the config's own directory, so dropping it moves nothing and
the targets are left byte-identical. Where a baseUrl pointed elsewhere,
each affected target was rewritten as join(baseUrl, target).

`moduleResolution` was chosen from the declared `module`: commonjs ->
node16, esnext/preserve -> bundler. Configs whose `module` is unset or
exotic were left alone rather than guessed at.

The result is accepted by BOTH toolchains, so nothing has to upgrade
TypeScript in lockstep. Verified on hanzo/chat packages/api: tsc 5.9
779 -> 778 errors (no regression), and tsc 7.0.2 now runs the project
in 2s where it previously refused the config.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-27 14:18:19 -07:00
zeekayandhanzo-dev b4b87be598 account portal: move it here, where the identity UI lives
The portal serves account.hanzo.id / account.lux.id — branded account-management
pages, IAM access tokens over the OAuth code exchange, redirecting to each
brand's login. That is identity UI: same brands, same IAM, same login redirect as
apps/web. It had no business being its own repo.

It also had a worse problem. It lived in hanzoai/account, which is ALSO the Go
module github.com/hanzoai/account — the package that owns the billing-account
rule (Account, Payer, Subject) that cloud and commerce both import. One repo,
two unrelated codebases, one name: `main` was this Worker, and the Go module
survived only on a `go` branch plus its tags. A Go module whose default branch
has no go.mod cannot resolve for anything but an explicit tag, which is why it
was private-and-pinned rather than importable.

Moving the Worker here frees that repo to be the Go module and nothing else.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-27 13:05:47 -07:00
hanzo-dev c89793f14d ci: drop the Gitea mirror-sync nudge
Superseded: the Hanzo GitHub App pushes a webhook, so the forge tracks GitHub
without a per-repo workflow. This file called git.hanzo.ai/api/v1/.../mirror-sync
— a Gitea API for a system we no longer drive — and would sit inert in every repo.

One mechanism, in one place, instead of ~350 copies of a cron.
2026-07-27 10:34:44 -07:00
zooqueenandhanzo-dev 446a46a465 0.2.14 — reunite main with production, on top of the design-token pass
The running image (0.2.13, built from 9477777) was NOT on origin/main. The two
had diverged at 7a72225f with the live code unmerged: self-service org creation
founding through /v1/iam/onboard instead of the admin-only add-organization verb
(which is 401 without a bearer and 403 with one, so signup was dead), and the
App/Chat/Cloud launcher. A release cut from main would have silently regressed
both on hanzo.id, lux.id, pars.id and zoo.id at once.

This merges them. main is once again what runs, and 0.2.14 is the first tag that
carries BOTH the live fixes and the @hanzo/design token pass. The only conflict
was the version line.

136 unit tests and every workspace typecheck green; the login, device-approval
and spinner measurements re-taken in a real browser against the merged bundle.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-27 09:28:10 -07:00
zooqueenandhanzo-dev c9943fb883 id: style from @hanzo/design tokens, and stop painting controls by ancestor
A component's surface must never depend on WHERE it is mounted. app.css painted
controls with the descendant selectors `form input {…}` and `.hanzo-id-btn, form
button {…}`. DeviceApproval — the screen a human hits to authorise the CLI — has
2 inputs and 0 <form> ancestors, so its device-code field fell out of the
stylesheet and rendered as raw UA chrome: 31px tall, #3b3b3b, a 2px inset bevel,
square corners, sitting beside correctly-styled 44px siblings. That is the mirror
image of a component library shipping utility class names with no CSS behind
them — bare elements instead of bare class names — and it was live on hanzo.id,
lux.id and pars.id at once. Every control now carries its own class
(.hanzo-id-input / .hanzo-id-btn / .hanzo-id-field / .hanzo-id-form) and no
element-descendant selector for surface is left in the file.

.hanzo-id-spinner had NO rule at all: the loading state measured 0px and was
invisible on every portal. It is a real 28px ring now.

Tokens come from @hanzo/design — the same token layer hanzoai/pay renders from,
so the two halves of one flow (sign in -> pay) agree on the page black (#000000,
--background; the page was painted with the --card value), the type ramp (h1 21px
on both), the radii and the greys. Nothing in this file invents a colour, radius
or size. @hanzo/gui was declared as a dependency and imported exactly zero times;
it is removed rather than left as decoration.

ONE button: .hanzo-id-social-btn and the .primary modifier are gone. .hanzo-id-btn
IS primary and .ghost is the secondary surface, so 13 treatments become 1
primitive with 1 modifier. ONE focus indicator: the file had a single focus rule
(`form input:focus`), so every button, link and social entry fell back to
Chrome's blue outline: auto on a monochrome surface. `font: inherit` on every
control: buttons and inputs rendered in Arial while headings rendered in the
platform face — two typefaces in one 432px card, including on the Sign in CTA.

Control borders are --white-40 on purpose. On --background the semantic --border
(#1f1f1f) measures 1.27:1 and --border-strong (#404040) 2.03:1; neither clears
the 3:1 a control boundary needs (WCAG 1.4.11), and --white-40 measures 3.66:1
and is on the ladder. Same reasoning for the focus ring: --ring (#333333)
measures 1.66:1 and cannot carry one, so the ring is --primary.

Fonts are deliberately NOT imported from the design package: tokens/fonts.css
pulls Geist from fonts.googleapis.com and the sign-in path loads no third-party
font. --font-sans resolves to the platform stack, the identical value hanzoai/pay
sets.

Measured in a real browser against the built bundle, fresh context and empty
storage: page #000000, every control 44px, one 6px radius, h1 21px, a white
2px focus ring, nothing under the touch floor at 390px, no horizontal overflow.
134 unit tests and every workspace typecheck green.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-27 09:25:33 -07:00
hanzo-dev b774772a92 ci: nudge git.hanzo.ai to pull on push
git.hanzo.ai mirrors this repo by PULL on a ~10-minute interval, and arcd runs
CI/CD there — so every push waited out that interval before anything built.
This asks Gitea to pull HEAD immediately.

Latency only: the repo already mirrors via the App webhook, so a missing
HANZO_GIT_TOKEN or a failed curl is non-fatal and never fails the push.
Idempotent (mirror-sync just pulls HEAD) and concurrency-coalesced.
2026-07-27 08:20:36 -07:00
zeekayandhanzo-dev 9477777e6b chore(release): 0.2.13 — self-service org creation + App/Chat/Cloud launcher
0.2.12 is the live image (built from 3b067475); this is the next free release
above it.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-26 17:26:12 -07:00
zeekayandhanzo-dev e9cca37728 fix(launcher): Hanzo is App, Chat, Cloud — Console is Cloud's old name
The signed-in launcher listed six tiles and one product twice. "Console" is not a
separate product: it is what Cloud used to be called, console.hanzo.ai redirects
to cloud.hanzo.ai, and clicking either tile landed on the same page — so the list
taught people a product that does not exist and sent half the traffic through a
redirect. It is gone, and nothing here links to console.hanzo.ai any more.

Analytics, Platform and Storage went with it. Storage pointed at s3.hanzo.ai,
which answers a browser with a bare XML AccessDenied — it is an S3 API endpoint,
not a page — and a launcher that lands you on an error page is how people learn
the tiles are broken. The other two are surfaces inside Cloud, not products of
their own.

The tiles themselves were never the defect: they are plain <a href> anchors, they
hit-test clean, and they navigate — middle-click and copy-link always worked. What
made them feel dead was the destination. Every app bounced a signed-in person to a
credential form, so a click ended back on hanzo.id staring at a password box,
which looks exactly like nothing happened. That is IAM's missing silent-SSO branch
(hanzoai/iam v1.33.15), fixed there rather than papered over here.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-26 17:24:14 -07:00
zeekayandhanzo-dev e22fef1f2a fix(onboarding): found an org through the front door, not the admin verb
A new person typed an org name at hanzo.id/onboarding, clicked "Create
organization", and got HTTP 401. Self-service signup was dead.

createOrg posted to /v1/iam/add-organization — Casdoor entity CRUD behind IAM's
authenticated Guard, with owner:"admin" in the body. That is the wrong door
twice over. The Guard authenticates by BEARER only, and a bare portal sign-in
(type=login) establishes a session cookie and mints no token at all, so the
browser's request arrived anonymous: 401. Hand it a bearer and it is still 403 —
"admin" is a reserved platform org and a human may only write an org row whose
name is the org they are already in, so a person founding their FIRST org can
never pass that gate. Correct behaviour; wrong door. Widening it would mean
handing an app blanket authority over every tenant's orgs.

/v1/iam/onboard is the door built for this: it resolves the caller from their own
session or bearer and provisions the whole tenant under their authority as its
founder — org stamped with them as Founder, them moved in as its owner, one
metered API key — atomically and idempotently.

The display name now travels and the server's slug comes back: slug policy lives
in one place, so the client's preview is only a preview. The front door answers
{org}/{error}, not the casibase envelope, so its own words reach the user ("that
name is reserved", "you already have an organization") instead of a bare code.

Tests: the call goes to onboard and never to add-organization; both credentials
ride along; the front door's error text surfaces verbatim.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-26 17:24:03 -07:00
zeekay 7a72225f58 fix: point forge API calls at /v1 — /api/v1 is gone
The fork moved its API off /api to /v1, so every call built against
${{ github.server_url }}/api/v1/... now 404s. Verified live with a control:
/v1/version 200, /api/v1/version 404, a nonsense path 404.

This is the build-dispatch in sync-from-github, so a fast-forward from GitHub
was landing commits and then silently failing to trigger the build.
2026-07-26 15:51:46 -07:00
zeekay 0184f80346 ci: drop the last two GitHub workflows — id is native like the rest
Neither did any work. `docker.yml` was reduced to a `notice` job whose only step
echoes "native pipeline is .hanzo/workflows/deploy.yml; GitHub is a mirror" —
it burned a GitHub runner to print one sentence, and had run four times doing
exactly that. `workflow-sanity.yml` triggers only on
`pull_request: paths: ['.github/workflows/**']`, so with that directory empty it
can never fire on anything meaningful; it validated the thing being removed.

The real pipeline is unchanged and already here:
  .hanzo/workflows/deploy.yml           builds on hanzo-build-linux-amd64
  .hanzo/workflows/sync-from-github.yml fast-forward-only mirror pull

That matches console, cloud, universe and bootnode, all of which carry zero
GitHub workflows. git.hanzo.ai is the forge; GitHub is a mirror and needs no
CI of its own.
2026-07-26 14:28:05 -07:00
zeekay 3b067475d1 fix(tenant): a host with no entry must fail closed, not become Hanzo
resolveTenant's terminal branch returned DEFAULT_TENANTS[`${defaultOrg}.id`] —
Hanzo's tenant — for any host it could not resolve. Eight real identity hosts
have no built-in entry and exist only in the runtime catalog:

  zoolabs.id  www.zoolabs.id  id.zoo.network  id.lux.network
  iam.lux.network  id.pars.network  id.bootno.de  iam.hanzo.ai

and App.tsx deliberately tolerates a failed /config.json fetch. So whenever
that fetch failed, a Zoo, Lux, Pars or Bootnode visitor was handed orgId
`hanzo`, brandPackage `@hanzo/brand` and iamUrl `https://hanzo.id` — shown
"Sign in to Hanzo ID" under the Hanzo mark, and posting their credentials at
hanzo.id. Not a cosmetic brand leak: the wrong origin receives the password.

The function's own comment eleven lines above already promised this could not
happen ("Never another brand's config: a catalog-only host … must not inherit
Hanzo's issuer or brand package"). The guarantee held only while the catalog
loaded; the fallback broke it exactly when the catalog did not.

Unknown hosts now resolve to hostSkeleton(host) — the host's own origin, empty
orgId/clientId/brandPackage. The portal fails closed rather than silently
authenticating as another brand's IAM application. A login page that cannot
resolve its tenant must refuse, not guess. The `defaultOrg` option is deleted
with the branch that used it; the docblock no longer advertises a cross-brand
default.

Also removes the `zoo.id` built-in: crs/iam.yaml records the host as retired to
NXDOMAIN, and `dig zoo.id` returns nothing. It named a tenant that cannot be
reached.

One existing test asserted the defect as intended behaviour
(`orgId === 'hanzo'` for an unknown host) and is corrected. Added coverage
walks all seven brand hosts with NO catalog and asserts none of them comes back
as Hanzo. 134 tests pass across 12 files.
2026-07-26 12:33:30 -07:00
zeekay cdd9590e2a fix(tenant): a host with no entry must fail closed, not become Hanzo
resolveTenant's terminal branch returned DEFAULT_TENANTS[`${defaultOrg}.id`] —
Hanzo's tenant — for any host it could not resolve. Eight real identity hosts
have no built-in entry and exist only in the runtime catalog:

  zoolabs.id  www.zoolabs.id  id.zoo.network  id.lux.network
  iam.lux.network  id.pars.network  id.bootno.de  iam.hanzo.ai

and App.tsx deliberately tolerates a failed /config.json fetch. So whenever
that fetch failed, a Zoo, Lux, Pars or Bootnode visitor was handed orgId
`hanzo`, brandPackage `@hanzo/brand` and iamUrl `https://hanzo.id` — shown
"Sign in to Hanzo ID" under the Hanzo mark, and posting their credentials at
hanzo.id. Not a cosmetic brand leak: the wrong origin receives the password.

The function's own comment eleven lines above already promised this could not
happen ("Never another brand's config: a catalog-only host … must not inherit
Hanzo's issuer or brand package"). The guarantee held only while the catalog
loaded; the fallback broke it exactly when the catalog did not.

Unknown hosts now resolve to hostSkeleton(host) — the host's own origin, empty
orgId/clientId/brandPackage. The portal fails closed rather than silently
authenticating as another brand's IAM application. A login page that cannot
resolve its tenant must refuse, not guess. The `defaultOrg` option is deleted
with the branch that used it; the docblock no longer advertises a cross-brand
default.

Also removes the `zoo.id` built-in: crs/iam.yaml records the host as retired to
NXDOMAIN, and `dig zoo.id` returns nothing. It named a tenant that cannot be
reached.

One existing test asserted the defect as intended behaviour
(`orgId === 'hanzo'` for an unknown host) and is corrected. Added coverage
walks all seven brand hosts with NO catalog and asserts none of them comes back
as Hanzo. 134 tests pass across 12 files.
2026-07-26 12:33:30 -07:00
zeekay 48acbaaf92 test(tenant): a catalog host with NO brandUrl must not fall back to Hanzo
id.bootno.de had no catalog entry at all, so resolveTenant took the
`defaultOrg` path and a Bootnode customer was asked to "Sign in to Hanzo ID"
under the Hanzo mark — the white-label invariant broken at the highest-intent
moment in the product.

The existing osage.id case covers a catalog-only host that HAS a brandUrl.
id.bootno.de is the other shape: no bootnode brand package is published
(@bootnode/brand, @bootno/brand — all 404 on npm), so the entry carries no
brandUrl at all. That must resolve to an EMPTY brandPackage — the loader then
shows a neutral wordmark rather than another brand's mark — while idBrandLabel
still reads `orgId` first and renders "Sign in to Bootnode ID".

Also pins the intended default explicitly: a genuinely unknown host still
resolves to hanzo. That is the behaviour we want to keep; the bug was that a
KNOWN brand host was reaching it.

9 tests pass.
2026-07-26 11:02:37 -07:00
zeekay c8a14ce212 test(tenant): a catalog host with NO brandUrl must not fall back to Hanzo
id.bootno.de had no catalog entry at all, so resolveTenant took the
`defaultOrg` path and a Bootnode customer was asked to "Sign in to Hanzo ID"
under the Hanzo mark — the white-label invariant broken at the highest-intent
moment in the product.

The existing osage.id case covers a catalog-only host that HAS a brandUrl.
id.bootno.de is the other shape: no bootnode brand package is published
(@bootnode/brand, @bootno/brand — all 404 on npm), so the entry carries no
brandUrl at all. That must resolve to an EMPTY brandPackage — the loader then
shows a neutral wordmark rather than another brand's mark — while idBrandLabel
still reads `orgId` first and renders "Sign in to Bootnode ID".

Also pins the intended default explicitly: a genuinely unknown host still
resolves to hanzo. That is the behaviour we want to keep; the bug was that a
KNOWN brand host was reaching it.

9 tests pass.
2026-07-26 11:02:37 -07:00
zeekayandClaude Opus 5 6b12efdde5 ci: pull from GitHub fast-forward-only — the other half of the loop
The push-mirror carries native -> GitHub. This is the return leg, so the two
forges converge from either side.

They compose rather than fight: a native commit reaches GitHub via the mirror,
so this job then sees LOCAL == REMOTE and exits "in sync"; a GitHub commit
fast-forwards native here and the resulting mirror push is a no-op. No echo,
no loop.

Fast-forward ONLY -- a divergence fails loudly instead of force-pushing either
side and destroying whichever history lost the race. Same shape hanzoai/app has
run green 299 times.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-07-26 09:15:45 -07:00
zeekayandhanzo-dev 483b5e177c ci: pull from GitHub fast-forward-only — the other half of the loop
The push-mirror carries native -> GitHub. This is the return leg, so the two
forges converge from either side.

They compose rather than fight: a native commit reaches GitHub via the mirror,
so this job then sees LOCAL == REMOTE and exits "in sync"; a GitHub commit
fast-forwards native here and the resulting mirror push is a no-op. No echo,
no loop.

Fast-forward ONLY -- a divergence fails loudly instead of force-pushing either
side and destroying whichever history lost the race. Same shape hanzoai/app has
run green 299 times.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-26 09:15:45 -07:00
zeekayandClaude Opus 5 818a7d9079 build: do not build on docs-only commits
The previous commit touched only LLM.md and went RED: the version guard
correctly refused to rebuild an existing 0.2.11, but a docs edit cannot produce
a release, so it should never have started a build. A red run for a change that
could not possibly ship is how people learn to ignore red.

paths-ignore for markdown, workflow files and docs -- same shape hanzoai/app
uses. Real source changes still build; workflow_dispatch forces one.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-07-26 08:14:25 -07:00
zeekayandhanzo-dev d378b6d14b build: do not build on docs-only commits
The previous commit touched only LLM.md and went RED: the version guard
correctly refused to rebuild an existing 0.2.11, but a docs edit cannot produce
a release, so it should never have started a build. A red run for a change that
could not possibly ship is how people learn to ignore red.

paths-ignore for markdown, workflow files and docs -- same shape hanzoai/app
uses. Real source changes still build; workflow_dispatch forces one.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-26 08:14:25 -07:00
zeekayandClaude Opus 5 ad3e6daf48 docs: record that truth flows git.hanzo.ai -> GitHub
build / build (push) Failing after 23s
This repo is canonical on git.hanzo.ai and GitHub is now a push-mirror of it
(sync_on_commit). It used to be the reverse, and as a pull-mirror it could run
no CI at all -- four commits shipped zero images with nothing reporting a fault.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-07-26 08:11:55 -07:00
zeekayandhanzo-dev c6d53f678c docs: record that truth flows git.hanzo.ai -> GitHub
This repo is canonical on git.hanzo.ai and GitHub is now a push-mirror of it
(sync_on_commit). It used to be the reverse, and as a pull-mirror it could run
no CI at all -- four commits shipped zero images with nothing reporting a fault.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-26 08:11:55 -07:00
zeekayandClaude Opus 5 1c1cd9a803 build: publish to oci.hanzo.ai as well as ghcr, and 0.2.11
build / build (push) Successful in 56s
oci.hanzo.ai is our own registry and where images belong. It is the same
registry.hanzo.ai backend under a second hostname, gated by IAM token auth
(realm https://iam.hanzo.ai/v1/iam/registry/token). Verified hanzo-registry can
obtain a repository:id:pull,push token before wiring this.

Publishing to BOTH rather than cutting over, deliberately: Kubernetes matches
imagePullSecrets by HOSTNAME, and the KMS-synced registry-credentials secret
carries an auths entry only for registry.hanzo.ai. A CR pointed at
oci.hanzo.ai today would ImagePullBackOff every replica. Adding that entry
changes a shared credential every registry client holds, so it is not a change
to make blind. With both published, the CR flip is one line the moment it lands.

Creds are ORG-level Actions secrets (OCI_USER / OCI_TOKEN) sourced from the
KMS-synced credential already in-cluster -- KMS stays the source of truth.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-07-25 17:58:02 -07:00
zeekayandhanzo-dev 8a556d0094 build: publish to oci.hanzo.ai as well as ghcr, and 0.2.11
oci.hanzo.ai is our own registry and where images belong. It is the same
registry.hanzo.ai backend under a second hostname, gated by IAM token auth
(realm https://iam.hanzo.ai/v1/iam/registry/token). Verified hanzo-registry can
obtain a repository:id:pull,push token before wiring this.

Publishing to BOTH rather than cutting over, deliberately: Kubernetes matches
imagePullSecrets by HOSTNAME, and the KMS-synced registry-credentials secret
carries an auths entry only for registry.hanzo.ai. A CR pointed at
oci.hanzo.ai today would ImagePullBackOff every replica. Adding that entry
changes a shared credential every registry client holds, so it is not a change
to make blind. With both published, the CR flip is one line the moment it lands.

Creds are ORG-level Actions secrets (OCI_USER / OCI_TOKEN) sourced from the
KMS-synced credential already in-cluster -- KMS stays the source of truth.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-25 17:58:02 -07:00
zeekayandClaude Opus 5 55c7d907de build: tag images semver from package.json, never a sha
build / build (push) Successful in 47s
Deployments pin semver. package.json is the single source of the version, so a
release IS a version bump: edit the version, push, and the image that appears is
named after it -- no separate tagging step to forget and no sha in a CR.

Refuses to rebuild a version that already exists, so a tag other environments
may be pinning is never silently overwritten.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-07-25 17:42:55 -07:00
zeekayandhanzo-dev 41d2409117 build: tag images semver from package.json, never a sha
Deployments pin semver. package.json is the single source of the version, so a
release IS a version bump: edit the version, push, and the image that appears is
named after it -- no separate tagging step to forget and no sha in a CR.

Refuses to rebuild a version that already exists, so a tag other environments
may be pinning is never silently overwritten.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-25 17:42:55 -07:00
zeekayandClaude Opus 5 d05a53d3cb build: log in to ghcr before pushing
build / build (push) Successful in 52s
Every run died on 'https://ghcr.io/v2/: denied: denied'. The raw buildctl
invocation pushed with no registry login at all -- there was no auth step in the
job. Replaced with the same trio hanzoai/app and hanzoai/cloud use:
setup-buildx -> login-action (GH_PAT, write:packages, user hanzo-dev, an
ORG-level secret already inherited here) -> build-push-action.

Adds a manifest-inspect gate: build-push-action can exit 0 while the manifest is
not yet resolvable, and a CR bumped to a phantom tag becomes ImagePullBackOff.
Prove it pulls here instead.

Tag is sha-<full sha> so it is immutable and never floats.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-07-25 17:37:14 -07:00
zeekayandhanzo-dev 93b3524a42 build: log in to ghcr before pushing
Every run died on 'https://ghcr.io/v2/: denied: denied'. The raw buildctl
invocation pushed with no registry login at all -- there was no auth step in the
job. Replaced with the same trio hanzoai/app and hanzoai/cloud use:
setup-buildx -> login-action (GH_PAT, write:packages, user hanzo-dev, an
ORG-level secret already inherited here) -> build-push-action.

Adds a manifest-inspect gate: build-push-action can exit 0 while the manifest is
not yet resolvable, and a CR bumped to a phantom tag becomes ImagePullBackOff.
Prove it pulls here instead.

Tag is sha-<full sha> so it is immutable and never floats.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-25 17:37:14 -07:00
zeekayandClaude Opus 5 8aee66d029 build: target the runner pool that exists
build / build (push) Failing after 4s
runs-on was hanzo-linux-amd64. The fleet advertises hanzo-build-linux-amd64
(plus ubuntu-* aliases) and nothing else, so no runner could match and Gitea
created NO RUN AT ALL -- not a queued one, not a failed one. Four commits landed
on main between 2026-07-23 and today and produced zero images, with nothing
anywhere reporting a problem.

Same label hanzoai/app uses, where it has 299 green runs.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-07-25 17:35:29 -07:00
zeekayandhanzo-dev 2d4e759dfb build: target the runner pool that exists
runs-on was hanzo-linux-amd64. The fleet advertises hanzo-build-linux-amd64
(plus ubuntu-* aliases) and nothing else, so no runner could match and Gitea
created NO RUN AT ALL -- not a queued one, not a failed one. Four commits landed
on main between 2026-07-23 and today and produced zero images, with nothing
anywhere reporting a problem.

Same label hanzoai/app uses, where it has 299 green runs.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-25 17:35:29 -07:00
zeekayandClaude Opus 5 a7e917f348 build: separate build from deploy, and 0.2.10
build / build (push) Canceled after 0s
The pipeline ended with 'kubectl patch app id', but App/id is CD-governed --
id.yaml is listed in universe crs/kustomization.yaml and the live object carries
apps.hanzo.ai/tracking-id -- so that patch is reverted on the next sync (~90s)
to whatever git says. It could never stick, and it FAILED SILENTLY: the patch
applies, the pod rolls, and CD quietly restores the old tag a minute later.

Build and deploy are separate concerns. This emits an immutable image; git
declares desired state; CD applies it. Shipping a build is a tag bump in
universe crs/id.yaml -- the one way a tag reaches the cluster.

0.2.9 -> 0.2.10 carries the apex-login fix.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-07-25 17:32:52 -07:00
zeekayandhanzo-dev 6907c88383 build: separate build from deploy, and 0.2.10
The pipeline ended with 'kubectl patch app id', but App/id is CD-governed --
id.yaml is listed in universe crs/kustomization.yaml and the live object carries
apps.hanzo.ai/tracking-id -- so that patch is reverted on the next sync (~90s)
to whatever git says. It could never stick, and it FAILED SILENTLY: the patch
applies, the pod rolls, and CD quietly restores the old tag a minute later.

Build and deploy are separate concerns. This emits an immutable image; git
declares desired state; CD applies it. Shipping a build is a tag bump in
universe crs/id.yaml -- the one way a tag reaches the cluster.

0.2.9 -> 0.2.10 carries the apex-login fix.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-25 17:32:52 -07:00
zeekayandClaude Opus 5 bc890fd0d0 login: the apex form posts an organization — it was dead on every brand
deploy / deploy (push) Canceled after 0s
Typing a password at hanzo.id, lux.id, iam.hanzo.ai or pars.id did nothing. The
bare portal deliberately posted NO organization, and iam2 refuses that:

    HTTP 200  {"status":"error","msg":"organization, username and password are required"}

The 200 is why nobody caught it. The form renders that message as though the
USER's credentials were wrong, and every status-code monitor reads the response
as healthy. Four brand front doors, dead, green on the dashboard.

The omission was correct once. It let IAM resolve cross-org so a colliding
identity (z@hanzo.ai exists in both admin and hanzo) landed on admin/* with a
full multi-org session. iam2 removed that ON PURPOSE and calls the collision a
defect -- internal/registry/registry.go: "the F-2 bug where z@hanzo.ai collided
across admin and hanzo" -- because cross-org resolution coupled lockout counters
across rows and gave a brute-force oracle on the SuperAdmin. The server is not
regressed; this client was.

LoginForm now resolves the app's own org through get-app-login and posts it on
BOTH entry points, which is exactly what 0.2.2 already established for the
downstream-app path -- the apex path simply never got it. A global admin is no
longer reached by omission: they sign into an admin-org app (hanzo-admin-guard),
the explicit path 0.2.2/0.2.3 describe.

client.login() stays a pure passthrough -- it never invents an org. The two
tests pinning that are unchanged and still pass; only their now-dead rationale
is corrected, because a stale WHY is how this gets reverted. CLAUDE.md's 0.1.23
section is marked SUPERSEDED for the same reason: read alone, it argues for
putting the outage back.

130/130 green.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-07-25 17:17:01 -07:00
zeekayandhanzo-dev 876be989fd login: the apex form posts an organization — it was dead on every brand
Typing a password at hanzo.id, lux.id, iam.hanzo.ai or pars.id did nothing. The
bare portal deliberately posted NO organization, and iam2 refuses that:

    HTTP 200  {"status":"error","msg":"organization, username and password are required"}

The 200 is why nobody caught it. The form renders that message as though the
USER's credentials were wrong, and every status-code monitor reads the response
as healthy. Four brand front doors, dead, green on the dashboard.

The omission was correct once. It let IAM resolve cross-org so a colliding
identity (z@hanzo.ai exists in both admin and hanzo) landed on admin/* with a
full multi-org session. iam2 removed that ON PURPOSE and calls the collision a
defect -- internal/registry/registry.go: "the F-2 bug where z@hanzo.ai collided
across admin and hanzo" -- because cross-org resolution coupled lockout counters
across rows and gave a brute-force oracle on the SuperAdmin. The server is not
regressed; this client was.

LoginForm now resolves the app's own org through get-app-login and posts it on
BOTH entry points, which is exactly what 0.2.2 already established for the
downstream-app path -- the apex path simply never got it. A global admin is no
longer reached by omission: they sign into an admin-org app (hanzo-admin-guard),
the explicit path 0.2.2/0.2.3 describe.

client.login() stays a pure passthrough -- it never invents an org. The two
tests pinning that are unchanged and still pass; only their now-dead rationale
is corrected, because a stale WHY is how this gets reverted. CLAUDE.md's 0.1.23
section is marked SUPERSEDED for the same reason: read alone, it argues for
putting the outage back.

130/130 green.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-25 17:17:01 -07:00
z 8d2ad1891a chore(spa): pin the ONE canonical hanzoai/spa 1.4.8
hanzoai/spa is the SPA-friendly server (index.html/.html fallthrough + a CSP
that does not blank a client-routed app); hanzoai/static is the strict one.
Consumers had drifted to two untagged pins (implicit :latest — banned, and
non-reproducible) and three at 1.2.0, thirteen patches behind. One image,
one version. Published spelling is 1.4.8 (v1.4.8 does not exist).
2026-07-25 10:46:58 -07:00
z 1acb94ac65 chore(spa): pin the ONE canonical hanzoai/spa 1.4.8
hanzoai/spa is the SPA-friendly server (index.html/.html fallthrough + a CSP
that does not blank a client-routed app); hanzoai/static is the strict one.
Consumers had drifted to two untagged pins (implicit :latest — banned, and
non-reproducible) and three at 1.2.0, thirteen patches behind. One image,
one version. Published spelling is 1.4.8 (v1.4.8 does not exist).
2026-07-25 10:46:58 -07:00
hanzo-dev 2cd1979cb0 ci(docker): neutralize to sync-notice — native build is .hanzo/workflows/deploy.yml 2026-07-24 15:15:38 -07:00
hanzo-dev 08a19b8dcd ci(docker): neutralize to sync-notice — native build is .hanzo/workflows/deploy.yml 2026-07-24 15:15:38 -07:00
hanzo-dev 7bbfb77cd7 ci(deploy): native Hanzo pipeline — BuildKit → ghcr.io/hanzoai/id → operator reconcile 2026-07-24 15:15:17 -07:00
hanzo-dev ffe5d2b9b2 ci(deploy): native Hanzo pipeline — BuildKit → ghcr.io/hanzoai/id → operator reconcile 2026-07-24 15:15:17 -07:00
Hanzo AI 4226ef57a7 chore(release): 0.2.9 — one chain-agnostic Connect Wallet button 2026-07-23 01:38:36 -07:00
Hanzo AI e74c3ffdae chore(release): 0.2.9 — one chain-agnostic Connect Wallet button 2026-07-23 01:38:36 -07:00
Hanzo AI 403ff90b10 feat(id/auth): merge EVM + Solana wallet buttons into one chain-agnostic Connect Wallet
SocialButtons rendered one wallet button per ENABLED chain (Continue with Ethereum / EVM + Continue with Solana). Merge the ENTRY into a single monochrome 'Connect Wallet' button that keeps BOTH flows.

- web3.ts: add detectWalletChains() — a pure window sniff (EVM=window.ethereum, Solana=window.solana/solflare/backpack) derived from ENABLED_WALLET_CHAINS; injectable window, no DOM/connect/I-O.
- SocialButtons.tsx: one Connect Wallet button (data-wallet-connect). onConnectWallet -> detect: exactly one injected chain connects straight; zero or many reveals an inline chooser (.hanzo-id-wallet-chains, data-chain=evm|solana) so either chain stays reachable. Per-chain startWallet -> loginWithWalletChain unchanged. Sits beside the GitLab provider (0.2.8).
- app.css: indented monochrome chooser matching the GitHub/GitLab/Google buttons.
- 3 new detectWalletChains unit tests (130 vitest pass). Frontend-only; IAM/casdoor untouched.
2026-07-23 01:38:36 -07:00
Hanzo AI 09ae6190b6 feat(id/auth): merge EVM + Solana wallet buttons into one chain-agnostic Connect Wallet
SocialButtons rendered one wallet button per ENABLED chain (Continue with Ethereum / EVM + Continue with Solana). Merge the ENTRY into a single monochrome 'Connect Wallet' button that keeps BOTH flows.

- web3.ts: add detectWalletChains() — a pure window sniff (EVM=window.ethereum, Solana=window.solana/solflare/backpack) derived from ENABLED_WALLET_CHAINS; injectable window, no DOM/connect/I-O.
- SocialButtons.tsx: one Connect Wallet button (data-wallet-connect). onConnectWallet -> detect: exactly one injected chain connects straight; zero or many reveals an inline chooser (.hanzo-id-wallet-chains, data-chain=evm|solana) so either chain stays reachable. Per-chain startWallet -> loginWithWalletChain unchanged. Sits beside the GitLab provider (0.2.8).
- app.css: indented monochrome chooser matching the GitHub/GitLab/Google buttons.
- 3 new detectWalletChains unit tests (130 vitest pass). Frontend-only; IAM/casdoor untouched.
2026-07-23 01:38:36 -07:00
hanzo-dev 33cc55e837 merge(feat/brandcontract-from-hanzo-brand): consolidate onto main 2026-07-21 17:18:35 -07:00
hanzo-dev dfe7eff66d merge(feat/brandcontract-from-hanzo-brand): consolidate onto main 2026-07-21 17:18:35 -07:00
hanzo-dev cc7885c680 refactor(brand): re-export BrandContract from @hanzo/brand
id-shared's BrandContract already documented that it 'matches the consumer
contract in @hanzo/brand' — now it IS that contract, re-exported, not a second
copy that can drift. @hanzo/brand is the canonical home (its toBrandContract
projects the registry onto exactly this shape). Byte-identical fields, so the
8 auth-page consumers (Login/Signup/Callback/DeviceApproval/Portal/Onboarding/
Forgot/BrandHeader) see the same type — zero behavior change.

Verified: id-shared tc (tsc --noEmit) + 7 tenant tests green with @hanzo/brand
linked. App-level build gates on the id workspace install + @hanzo/brand
publish (deploy-gate).
2026-07-21 09:41:03 -07:00
hanzo-dev ec17532fb1 refactor(brand): re-export BrandContract from @hanzo/brand
id-shared's BrandContract already documented that it 'matches the consumer
contract in @hanzo/brand' — now it IS that contract, re-exported, not a second
copy that can drift. @hanzo/brand is the canonical home (its toBrandContract
projects the registry onto exactly this shape). Byte-identical fields, so the
8 auth-page consumers (Login/Signup/Callback/DeviceApproval/Portal/Onboarding/
Forgot/BrandHeader) see the same type — zero behavior change.

Verified: id-shared tc (tsc --noEmit) + 7 tenant tests green with @hanzo/brand
linked. App-level build gates on the id workspace install + @hanzo/brand
publish (deploy-gate).
2026-07-21 09:41:03 -07:00
zeekayandClaude Opus 4.8 f2596c4788 test(id): unify on vitest — one runner, 12/12 files, 127 tests green (0.2.8)
The connect crypto/connector suites (bitcoin, ton, xrp, caip122, verify,
connectors — 6 files, 71 tests) imported `vitest` but vitest was never
installed and connect had no `test` script: they had never run. The auth /
shared / onboarding suites ran green under three separate per-package
`node --test --experimental-strip-types` scripts — a divergent second runner.

Decomplect to ONE runner:
- add vitest 3.2.7 (built for the repo's vite 7) as the sole test dep + root
  `vitest.config.ts` (include pkgs/**/src/**/*.test.ts, node env)
- root `test` = `vitest run`; drop the three per-package node:test scripts
- swap `import { test } from 'node:test'` -> `'vitest'` in the 6 node:test
  files; node:assert/strict assertions kept verbatim (real assertions, no
  rewrites)
- add pkgs/connect/tsconfig.json (matching siblings) so `pnpm tc` is green —
  connect had no tsconfig, so its typecheck errored out entirely

Result: `pnpm test` 12 files / 127 tests pass; `pnpm tc` green across all 6
projects.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-19 19:53:40 -07:00
zeekayandhanzo-dev aa10455c43 test(id): unify on vitest — one runner, 12/12 files, 127 tests green (0.2.8)
The connect crypto/connector suites (bitcoin, ton, xrp, caip122, verify,
connectors — 6 files, 71 tests) imported `vitest` but vitest was never
installed and connect had no `test` script: they had never run. The auth /
shared / onboarding suites ran green under three separate per-package
`node --test --experimental-strip-types` scripts — a divergent second runner.

Decomplect to ONE runner:
- add vitest 3.2.7 (built for the repo's vite 7) as the sole test dep + root
  `vitest.config.ts` (include pkgs/**/src/**/*.test.ts, node env)
- root `test` = `vitest run`; drop the three per-package node:test scripts
- swap `import { test } from 'node:test'` -> `'vitest'` in the 6 node:test
  files; node:assert/strict assertions kept verbatim (real assertions, no
  rewrites)
- add pkgs/connect/tsconfig.json (matching siblings) so `pnpm tc` is green —
  connect had no tsconfig, so its typecheck errored out entirely

Result: `pnpm test` 12 files / 127 tests pass; `pnpm tc` green across all 6
projects.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-19 19:53:40 -07:00
zeekayandClaude 2b3f9330e7 fix(id): center auth card — pin .hanzo-id-page min-height:100vh so main can vertically center
Co-Authored-By: Claude <noreply@anthropic.com>
2026-07-19 19:23:47 -07:00
zeekayandhanzo-dev 45ffd21a85 fix(id): center auth card — pin .hanzo-id-page min-height:100vh so main can vertically center
Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-19 19:23:47 -07:00
zeekayandClaude 9bfbb8add2 fix(id): neutral "<Brand> ID" heading + title, WCAG-3:1 input border, centered auth card
- Login <h1> + document.title now use idBrandLabel(brand, orgId) → "Lux ID" /
  "Zoo ID" / "Hanzo ID" instead of the brand package's product name
  ("Lux Exchange" / "Zoo Exchange") leaking into the IAM portal heading + tab
  title. New helper in @hanzo/id-shared strips a trailing product word and
  appends " ID", preferring the tenant orgId.
- --border #262626 → #71717a: 1.25:1 → ~3.9:1 against the #111 input fill,
  clearing the WCAG 3:1 non-text-contrast threshold (inputs, social buttons,
  dividers, org rows).
- .hanzo-id-page main: justify-content center → the auth card is vertically
  centered instead of top-anchored (~15% → centered).

Co-Authored-By: Claude <noreply@anthropic.com>
2026-07-19 19:10:55 -07:00
zeekayandhanzo-dev 6fb61a4817 fix(id): neutral "<Brand> ID" heading + title, WCAG-3:1 input border, centered auth card
- Login <h1> + document.title now use idBrandLabel(brand, orgId) → "Lux ID" /
  "Zoo ID" / "Hanzo ID" instead of the brand package's product name
  ("Lux Exchange" / "Zoo Exchange") leaking into the IAM portal heading + tab
  title. New helper in @hanzo/id-shared strips a trailing product word and
  appends " ID", preferring the tenant orgId.
- --border #262626 → #71717a: 1.25:1 → ~3.9:1 against the #111 input fill,
  clearing the WCAG 3:1 non-text-contrast threshold (inputs, social buttons,
  dividers, org rows).
- .hanzo-id-page main: justify-content center → the auth card is vertically
  centered instead of top-anchored (~15% → centered).

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-19 19:10:55 -07:00
hanzo-dev da19c5f208 feat(id/social): render GitLab as a first-class sign-in provider (0.2.7)
hanzo.id showed only GitHub/Google/wallet because the portal gated the
dynamic get-app-login set on a hardcoded renderable-provider allowlist that
omitted GitLab, and the social hop had no GitLab authorize endpoint — so the
configured provider-gitlab (real clientId in IAM) never appeared.

- social.ts AUTH_INFO: GitLab -> https://gitlab.com/oauth/authorize (read_user).
- SocialButtons: gitlab in PROVIDER_META + ORDER (+ GitLabIcon). It still
  renders ONLY when get-app-login says the app has it configured + canSignIn —
  dynamic, one source of truth; the allowlist is just per-type presentation.
- Tests: GitLab hop URL + isHoppableProvider(GitLab) locked in social.test.ts.

Verified: 9/9 auth tests pass; built bundle contains GitLab. Same provider-gitlab
backs the cloud /v1/integrations/gitlab connector (one app, KMS hanzo/prod/gitlab-oauth).
2026-07-15 10:17:41 -07:00
hanzo-dev 7f609193ff feat(id/social): render GitLab as a first-class sign-in provider (0.2.7)
hanzo.id showed only GitHub/Google/wallet because the portal gated the
dynamic get-app-login set on a hardcoded renderable-provider allowlist that
omitted GitLab, and the social hop had no GitLab authorize endpoint — so the
configured provider-gitlab (real clientId in IAM) never appeared.

- social.ts AUTH_INFO: GitLab -> https://gitlab.com/oauth/authorize (read_user).
- SocialButtons: gitlab in PROVIDER_META + ORDER (+ GitLabIcon). It still
  renders ONLY when get-app-login says the app has it configured + canSignIn —
  dynamic, one source of truth; the allowlist is just per-type presentation.
- Tests: GitLab hop URL + isHoppableProvider(GitLab) locked in social.test.ts.

Verified: 9/9 auth tests pass; built bundle contains GitLab. Same provider-gitlab
backs the cloud /v1/integrations/gitlab connector (one app, KMS hanzo/prod/gitlab-oauth).
2026-07-15 10:17:41 -07:00
Hanzo AI a87b7500e5 docs(id): document provider_hint auto-federation (0.2.6) [skip ci] 2026-07-14 06:30:45 -07:00
Hanzo AI bf734cf3cf docs(id): document provider_hint auto-federation (0.2.6) [skip ci] 2026-07-14 06:30:45 -07:00
Hanzo AI 23d0c0e7a8 chore(release): 0.2.6 — provider_hint social login auto-federation 2026-07-14 06:18:41 -07:00
Hanzo AI 5ace3df75c chore(release): 0.2.6 — provider_hint social login auto-federation 2026-07-14 06:18:41 -07:00
Hanzo AI 436931bb53 feat(id/social): honor provider_hint so console social login lands straight in the provider
Clicking "Continue with GitHub/Google" on console.hanzo.ai now goes directly
into the provider OAuth flow instead of bouncing the user to the hanzo.id login
form. Three fixes, all reusing the EXISTING hop (no duplicated IdP config):

- Login.tsx: a `federate` phase honors `?provider_hint=provider-github` (the
  value the console already sends). After a silent-SSO miss it auto-launches the
  hinted provider via a headless SocialButtons instead of showing the form,
  falling back to the form only when the hint matches no configured provider.
- SocialButtons: `autoStart` runs the SAME startProviderLogin hop the button
  runs, matched by a new pure `matchProviderHint` (accepts provider-github /
  github, case-insensitive). getAppLogin now validates against the DOWNSTREAM
  app's own redirect_uri (read from the query), not the portal /callback — the
  latter is not registered for a cross-app clientId (hanzo-cloud), so IAM
  rejected the read (status:error) and no providers resolved.
- hop uses method=signup (IAM find-or-create-login) for interactive social
  login; method=signin is the account-LINK branch and errors "user doesn't
  exist" on a fresh sign-in with no session.

Contracts locked: matchProviderHint + getAppLogin redirect_uri. 39 tests green,
tsc + vite build clean.
2026-07-14 06:18:41 -07:00
Hanzo AI aee84762b9 feat(id/social): honor provider_hint so console social login lands straight in the provider
Clicking "Continue with GitHub/Google" on console.hanzo.ai now goes directly
into the provider OAuth flow instead of bouncing the user to the hanzo.id login
form. Three fixes, all reusing the EXISTING hop (no duplicated IdP config):

- Login.tsx: a `federate` phase honors `?provider_hint=provider-github` (the
  value the console already sends). After a silent-SSO miss it auto-launches the
  hinted provider via a headless SocialButtons instead of showing the form,
  falling back to the form only when the hint matches no configured provider.
- SocialButtons: `autoStart` runs the SAME startProviderLogin hop the button
  runs, matched by a new pure `matchProviderHint` (accepts provider-github /
  github, case-insensitive). getAppLogin now validates against the DOWNSTREAM
  app's own redirect_uri (read from the query), not the portal /callback — the
  latter is not registered for a cross-app clientId (hanzo-cloud), so IAM
  rejected the read (status:error) and no providers resolved.
- hop uses method=signup (IAM find-or-create-login) for interactive social
  login; method=signin is the account-LINK branch and errors "user doesn't
  exist" on a fresh sign-in with no session.

Contracts locked: matchProviderHint + getAppLogin redirect_uri. 39 tests green,
tsc + vite build clean.
2026-07-14 06:18:41 -07:00
hanzo-dev 2ceecf49a5 chore(release): 0.2.5 — forced TOTP MFA enforcement (RequiredMfa/NextMfa)
Docker / docker (push) Failing after 11s
Ships the portal-side forced-MFA fix: parseLoginResponse now branches on
IAM's string data signal (RequiredMfa/NextMfa) instead of the never-set
mfa_required boolean, so org-forced 2FA can no longer be silently bypassed
into /onboarding.

Claude-Session: https://claude.ai/code/session_01XsNmNwzHUSKXfN7gYruUsM
2026-07-08 07:02:33 -07:00
hanzo-dev e9d3ae9d07 chore(release): 0.2.5 — forced TOTP MFA enforcement (RequiredMfa/NextMfa)
Ships the portal-side forced-MFA fix: parseLoginResponse now branches on
IAM's string data signal (RequiredMfa/NextMfa) instead of the never-set
mfa_required boolean, so org-forced 2FA can no longer be silently bypassed
into /onboarding.
2026-07-08 07:02:33 -07:00
zeekayandhanzo-dev f5d2f13499 feat(auth): enforce forced TOTP MFA in the portal (RequiredMfa / NextMfa)
IAM signals MFA with a STRING in the login response `data`
("RequiredMfa" = org forces MFA, user not enrolled; "NextMfa" = user has
MFA, needs a challenge) — never the `mfa_required` boolean the portal was
checking. So every first login fell through to /onboarding, silently
bypassing 2FA.

parseLoginResponse now branches on `data` BEFORE the /onboarding return:
  - RequiredMfa -> { mfaRequired, mfaStage: 'enroll' }
  - NextMfa     -> { mfaRequired, mfaStage: 'challenge', mfaTypes }

New AuthClient methods drive the flow against the canonical /v1/iam surface:
  getAccount, mfaInitiate, mfaVerify, mfaEnable, mfaChallenge.

Wire contract (verified live vs iam.hanzo.ai): the /v1/iam/mfa/setup/*
calls put EVERY param on the query string with an EMPTY body — the only
shape that satisfies both IAM's authz self-match (objOwner/objName are
read from the query only when the body is empty; a urlencoded body fails
the JSON-unmarshal and yields "Unauthorized operation") and the MFA
controller. owner/name ride the query on every call, incl. verify.

UI: MfaEnrollForm renders forced TOTP enrollment — QR from the otpauth://
URI via @paulmillr/qr (secret never leaves the browser), manual-key
fallback, recovery code, and the existing OTPForm for code entry. No skip
control. Login.tsx routes onMfaRequired by stage; the existing OTPForm is
reused for the NextMfa challenge.

Tests: 8 new MFA wire-contract tests (pure, mocked fetch). 14/14 green.
(cherry picked from commit 3df982b9fc)
2026-07-08 07:01:57 -07:00
zeekayandhanzo-dev 09239ed16c feat(auth): enforce forced TOTP MFA in the portal (RequiredMfa / NextMfa)
IAM signals MFA with a STRING in the login response `data`
("RequiredMfa" = org forces MFA, user not enrolled; "NextMfa" = user has
MFA, needs a challenge) — never the `mfa_required` boolean the portal was
checking. So every first login fell through to /onboarding, silently
bypassing 2FA.

parseLoginResponse now branches on `data` BEFORE the /onboarding return:
  - RequiredMfa -> { mfaRequired, mfaStage: 'enroll' }
  - NextMfa     -> { mfaRequired, mfaStage: 'challenge', mfaTypes }

New AuthClient methods drive the flow against the canonical /v1/iam surface:
  getAccount, mfaInitiate, mfaVerify, mfaEnable, mfaChallenge.

Wire contract (verified live vs iam.hanzo.ai): the /v1/iam/mfa/setup/*
calls put EVERY param on the query string with an EMPTY body — the only
shape that satisfies both IAM's authz self-match (objOwner/objName are
read from the query only when the body is empty; a urlencoded body fails
the JSON-unmarshal and yields "Unauthorized operation") and the MFA
controller. owner/name ride the query on every call, incl. verify.

UI: MfaEnrollForm renders forced TOTP enrollment — QR from the otpauth://
URI via @paulmillr/qr (secret never leaves the browser), manual-key
fallback, recovery code, and the existing OTPForm for code entry. No skip
control. Login.tsx routes onMfaRequired by stage; the existing OTPForm is
reused for the NextMfa challenge.

Tests: 8 new MFA wire-contract tests (pure, mocked fetch). 14/14 green.
(cherry picked from commit 115408a9d9)
2026-07-08 07:01:57 -07:00
zeekayandClaude Opus 4.8 f8640af7bc fix(id/login): >=44px touch targets on sign-in + social buttons (cosmetic)
Sign-in button and the social/Web3 buttons rendered at 42px on mobile,
2px under the 44px minimum touch target. Bump sign-in vertical padding to
13px (44px) and add min-height:44px to the flex-centered social buttons.
CSS-only; no auth/OAuth logic touched. Branched off the deployed 0.2.3
(origin/main) so nothing but this cosmetic change ships. Version -> 0.2.4.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-05 07:29:01 -07:00
zeekayandhanzo-dev b914f74144 fix(id/login): >=44px touch targets on sign-in + social buttons (cosmetic)
Sign-in button and the social/Web3 buttons rendered at 42px on mobile,
2px under the 44px minimum touch target. Bump sign-in vertical padding to
13px (44px) and add min-height:44px to the flex-centered social buttons.
CSS-only; no auth/OAuth logic touched. Branched off the deployed 0.2.3
(origin/main) so nothing but this cosmetic change ships. Version -> 0.2.4.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-05 07:29:01 -07:00
zandClaude Opus 4.8 f4e7c0d145 fix(auth): org-scope silent SSO so admin operators reach god-mode
admin.hanzo.ai is gated by admin-guard (ForwardAuth: owner==admin only).
Login rides client_id=hanzo-admin-guard (org=admin) and the 0.2.2 form fix
correctly posts organization=admin -> IAM resolves admin/z (owner=admin).
But Login.tsx's silentLogin SSO fast-path minted a code from the ambient
iam_session_id session regardless of its org; operators carry a hanzo/*
session, so it minted owner=hanzo and the guard bounced them to console —
the org-scoped form was never shown. Silent SSO shadowed the fix.

silentLogin now resolves the app org (get-app-login) + session owner
(get-account) and mints ONLY when they match; on no session or an org
mismatch it returns {} so Login.tsx falls back to the interactive form,
which authenticates in the app's own org. Same-org SSO (hanzo->hanzo)
still mints silently — no UX change. Cross-org (hanzo->admin-guard) now
falls to the form -> org=admin -> owner=admin -> god-mode.

Guard still validates owner==admin server-side, so this is availability
(admins get IN), not a privilege boundary — a non-admin can never reach
god-mode. IAM (v1.31.14) unchanged; SSO-ATO exact-match fix untouched.

Tests: 4 silentLogin cases (same-org mint, cross-org no-mint form-fallback,
same admin-org mint, no-session no-mint) + fix a stale providerLogin
redirect-url expectation. tc + 29 tests green; web build green.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-03 14:30:17 -07:00
zandhanzo-dev 55d7d3f411 fix(auth): org-scope silent SSO so admin operators reach god-mode
admin.hanzo.ai is gated by admin-guard (ForwardAuth: owner==admin only).
Login rides client_id=hanzo-admin-guard (org=admin) and the 0.2.2 form fix
correctly posts organization=admin -> IAM resolves admin/z (owner=admin).
But Login.tsx's silentLogin SSO fast-path minted a code from the ambient
iam_session_id session regardless of its org; operators carry a hanzo/*
session, so it minted owner=hanzo and the guard bounced them to console —
the org-scoped form was never shown. Silent SSO shadowed the fix.

silentLogin now resolves the app org (get-app-login) + session owner
(get-account) and mints ONLY when they match; on no session or an org
mismatch it returns {} so Login.tsx falls back to the interactive form,
which authenticates in the app's own org. Same-org SSO (hanzo->hanzo)
still mints silently — no UX change. Cross-org (hanzo->admin-guard) now
falls to the form -> org=admin -> owner=admin -> god-mode.

Guard still validates owner==admin server-side, so this is availability
(admins get IN), not a privilege boundary — a non-admin can never reach
god-mode. IAM (v1.31.14) unchanged; SSO-ATO exact-match fix untouched.

Tests: 4 silentLogin cases (same-org mint, cross-org no-mint form-fallback,
same admin-org mint, no-session no-mint) + fix a stale providerLogin
redirect-url expectation. tc + 29 tests green; web build green.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-03 14:30:17 -07:00
zeekayandClaude Opus 4.8 0a97192506 chore(release): 0.2.2 — admin-org login resolution fix
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-03 12:49:47 -07:00
zeekayandhanzo-dev 9cf009a8ac chore(release): 0.2.2 — admin-org login resolution fix
Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-03 12:49:47 -07:00
zeekayandClaude Opus 4.8 d928ecdfe3 fix(auth): resolve login org from the app being signed into, not the brand
A downstream app that initiates login passes its own client_id
(props.clientIdOverride). LoginForm ignored it for org/app resolution and
always pinned application=tenant.appName + organization=tenant.loginOrg, so
EVERY login authenticated inside the brand portal's own org (hanzo). The
admin-guard (client_id=hanzo-admin-guard) lives in the `admin` org: its
operators resolved to hanzo/* (owner=hanzo) instead of admin/* (owner=admin),
so the admin.hanzo.ai forward-auth gate (predicate owner==admin) could never
be satisfied — god-mode was unreachable.

Fix: when a client_id override is present, resolve {application, organization}
from that app's get-app-login (the canonical clientId -> app/org map) and post
BOTH, so IAM scopes the credential check to the app's org. The bare brand-portal
sign-in (no override) stays org-agnostic (loginOrg unset) so a global admin
still resolves cross-org into the full multi-org session.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-07-03 12:42:40 -07:00
zeekayandhanzo-dev 47e477d599 fix(auth): resolve login org from the app being signed into, not the brand
A downstream app that initiates login passes its own client_id
(props.clientIdOverride). LoginForm ignored it for org/app resolution and
always pinned application=tenant.appName + organization=tenant.loginOrg, so
EVERY login authenticated inside the brand portal's own org (hanzo). The
admin-guard (client_id=hanzo-admin-guard) lives in the `admin` org: its
operators resolved to hanzo/* (owner=hanzo) instead of admin/* (owner=admin),
so the admin.hanzo.ai forward-auth gate (predicate owner==admin) could never
be satisfied — god-mode was unreachable.

Fix: when a client_id override is present, resolve {application, organization}
from that app's get-app-login (the canonical clientId -> app/org map) and post
BOTH, so IAM scopes the credential check to the app's org. The bare brand-portal
sign-in (no override) stays org-agnostic (loginOrg unset) so a global admin
still resolves cross-org into the full multi-org session.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-07-03 12:42:40 -07:00
87d2da9c74 Debrand: replace Casdoor name with Hanzo IAM in comments/docs/aliases (#19)
Co-authored-by: Hanzo AI <ai@hanzo.ai>
2026-06-30 14:57:54 -07:00
9517f73345 Debrand: replace Casdoor name with Hanzo IAM in comments/docs/aliases (#19)
Co-authored-by: Hanzo AI <ai@hanzo.ai>
2026-06-30 14:57:54 -07:00
z 218528569e docs(brand): add hero banner 2026-06-28 20:06:13 -07:00
z 4cc44563b8 docs(brand): add hero banner 2026-06-28 20:06:13 -07:00
z 6df3f160b8 chore(brand): dynamic hero banner 2026-06-28 20:06:11 -07:00
z 373e00516e chore(brand): dynamic hero banner 2026-06-28 20:06:11 -07:00
0c190b08e3 feat: RFC 8628 device-authorization approval page (#16)
* feat(web): RFC 8628 device-authorization approval page

Add the /login/oauth/device approval page — the final gap in device login
(dev login --device-auth) on hanzo.id/lux.id/zoolabs.id.

- pkgs/auth: AuthClient.approveDevice(userCode) POSTs /v1/iam/login
  {type:device, userCode, application, organization} over the issuer session
  cookie (no credentials in body), mapping the consent-required and error
  branches. userCode normalized to IAM's [0-9a-z] alphabet.
- LoginForm: optional onAuthenticated to keep the caller on-page after sign-in.
- apps/web: DeviceApproval page — sign-in (reused) then anti-phishing confirm
  of app + code, success/error states. White-label via tenant. URL scrubbed of
  tokens. Route wired in App.tsx before the /login catch.
- Tests: 5 approveDevice cases (posting/normalization/empty/error/consent).

* fix(device): require explicit human affirmation before approval (H2)

H2 (HIGH): the ?user_code= prefill is now DISPLAY-ONLY. A signed-in victim
arriving from a crafted verification_uri_complete link can no longer one-click
approve an attacker's device — the Approve button stays disabled until the user
ticks an explicit checkbox affirming the code matches the one their OWN device
shows. Anti-phishing copy retained.

Aligns the client with the iam M1 change: user_codes are now UPPERCASE over an
unambiguous alphabet, so normalizeUserCode uppercases (was lowercase) for an
exact-match send, the code field renders uppercase, and the placeholder/tests
use a realistic code. Adds the .hanzo-id-device-confirm style.

tc + build green; device client tests green.

---------

Co-authored-by: hanzo-dev <dev@hanzo.ai>
2026-06-28 08:12:34 -07:00
Hanzo AI cdb83c111d fix(social): URL-safe base64 state + method=signup — fixes 'code_verifier does not match code_challenge' on Google login
Two root causes of broken social login across apps (billing, console2, console):
1. state was standard btoa() (emits +,/,=). The provider reflects state on a URL
   query; URLSearchParams turns '+' into space → atob corrupts the encoded OIDC
   request INCLUDING code_challenge → app's token exchange fails invalid_grant
   'code_verifier does not match code_challenge'. Fix: URL-safe base64
   (encodeState/decodeState), byte-exact round-trip.
2. method defaulted to 'signin' (account-LINK branch, needs existing session →
   400 on a fresh 'Continue with Google'). Canonical Casdoor default is 'signup'
   (find-or-create-LOGIN). Fixed.
Completes the lane that was interrupted by the session limit.
2026-06-26 19:22:37 -07:00
Hanzo AI 33b46504b3 Standardize favicon to monochrome Hanzo H (transparent bg) 2026-06-26 14:33:05 -07:00
Hanzo AI f807b6a11c chore(iam): unify @hanzo/iam to ^0.13.1 (localStorage PKCE login fix) 2026-06-26 12:59:40 -07:00
Hanzo AI 7ce73662b2 docs(sso): document the silent single sign-on mechanism (0.1.26) 2026-06-25 18:02:51 -07:00
Hanzo AI 583e406fa6 feat(sso): silent single sign-on — auto-continue authorize from existing issuer session
Login.tsx now attempts a credential-less silentLogin when an app sends the user
to the authorize page (client_id + redirect_uri present). IAM's Login handler
mints an auth code from the existing iam_session_id cookie (its already-signed-in
branch), so the 2nd/3rd app logs in seamlessly with no form. Falls back to the
interactive form when there is no live session. Backend already supports this;
this wires the SPA leg. Contract locked in client.test.ts (no creds posted,
redirect built from minted code; { error } -> form fallback).
2026-06-25 17:55:40 -07:00
Hanzo AI 87ecfef2a2 social login: dedup provider= in state (RC#1), match exchange redirect_uri to hop (RC#2 hardening)
- buildProviderAuthUrl strips any pre-existing provider= before appending the
  real social provider, so the base64 state carries exactly ONE provider=.
  Callback.tsx reads URLSearchParams.get (FIRST match) — two providers made it
  post the upstream hint (hanzo-iam) instead of provider-google.
- providerLogin redirect_uri now derives from tenant.oauthCallbackOrigin (the
  same source the hop uses), never publicOrigin — IAM forwards it verbatim to
  the provider token endpoint; a mismatch is invalid_grant.
- tests: social.test.ts dedup lock + client.test.ts redirect_uri lock (20 pass).
- RC#2 backend proven live: junk-code probe returns invalid_grant not
  invalid_client → Google clientId/secret + iam.hanzo.ai/callback all valid.
- bump id 0.1.25, @hanzo/id-auth 0.1.2.
2026-06-25 16:04:26 -07:00
Hanzo AI e9ec7689a8 docs(LLM): provider-name resolution fix + org unified-login mechanism 2026-06-24 17:02:24 -07:00
Hanzo AI 43f7d74249 fix(auth): resolve social provider name from nested record, not outer link label
parseAppLogin read the outer Casdoor app-provider LINK name as the
provider identity. Some seeds label that link <org>-iam, so the social
hop POSTed provider=<org>-iam and the IAM backend rejected it ('The
provider: hanzo-iam does not exist'). The provider's real identity is the
nested provider record name (provider-github). Prefer the inner record
name; fall back to the outer label only when no nested record exists.

Regression: getAppLogin uses the nested name + falls back. 18/18 pass.
2026-06-24 16:50:57 -07:00
Hanzo 6c1f661f17 fix(docker): COPY pkgs/connect/package.json before pnpm install
pkgs/auth now depends on @hanzo/id-connect (workspace:*) for the
multi-chain wallet login. The build stage COPYs each workspace
package.json before `pnpm install` so pnpm can resolve the graph, but
pkgs/connect/package.json was missing -> install failed:
'@hanzo/id-connect' unresolved workspace dependency. Add the COPY line.
2026-06-23 21:55:53 -07:00
Hanzo AI 1ea5fe9cdb feat(web): native multi-chain wallet SIWX login (EVM+Solana), drop OAuth-redirect web3 fallback
Replace the web3 button's @hanzo/iam OAuth-redirect fallback in SocialButtons
with native Sign-In-With-X via the vendored @hanzo/id-connect connectors. The
connect->nonce->sign->verify orchestration is ONE function (loginWithWalletChain
in pkgs/auth/src/web3.ts): GET /v1/iam/web3/nonce -> connector.signLogin(challenge)
-> POST /v1/iam/web3/verify, returning the SAME LoginResponse/redirect the
password flow uses. Decomplected: connect+sign = browser (lazy-loaded wallet
libs, code-split), verify = server.

Enabled chains gated on the verifier-readiness matrix via one exported constant
ENABLED_WALLET_CHAINS=['evm','solana'] (TON/XRP/Bitcoin disabled: Go verifiers
are stubs, would fail closed). The wallet provider expands into one connect
button per enabled chain. NO WalletConnect, NO projectId, NO web3-onboard.

Tests: pkgs/auth/src/web3.test.ts (mock fetch + fake signer) asserts nonce fetch,
signLogin gets the challenge, proof POSTed, SSO code redirect, disabled chains
fail closed without network/signer, wallet-rejection -> {error}.
2026-06-23 21:47:38 -07:00
Hanzo AI 154d0b3b7d feat(id): wire @luxwallet/connect multi-wallet web3 (EVM/SOL/BTC/TON/XRP)
- vendor luxwallet/connect (MIT, not on npm) as workspace pkg
  @hanzo/id-connect (pkgs/connect) — exports map points at ./src/*.ts,
  Vite compiles it directly like the other id pkgs.
- replace the placeholder window.ethereum eth_requestAccounts connector
  in Onboarding.tsx with getConnector('evm').connect() (EIP-6963
  multi-injection via viem). connectWallet keeps its string|null contract
  so @hanzo/id-onboarding stays wallet-lib-agnostic.
- add wallet peerDeps to @hanzo/id-web: viem, @tonconnect/sdk,
  sats-connect, @crossmarkio/sdk (connectors.ts eagerly imports all 5
  chains). No WalletConnect projectId needed — connect uses pure injected
  discovery, not the WC bridge protocol.

IAM provider-web3 already exists and is attached to the hanzo-id app.
vite build green (607 modules); secp256k1 chunk bundled.
2026-06-23 21:39:53 -07:00
Antje Worring 8745f62bfd fix(auth): thread OIDC nonce through the password-login path
Confidential OIDC clients that validate strictly (LibreChat openid-client with
OPENID_REUSE_TOKENS) send a nonce on authorize and require the id_token to echo
it. The portal's password-login path read code_challenge from the authorize URL
but DROPPED nonce, so the IAM minted a code (and id_token) with no nonce ->
openid-client 'unexpected JWT claim value encountered' -> hanzo.chat callback
HTTP 500. Thread nonce: Login.tsx reads ?nonce, LoginForm forwards it, and
client.login() puts it on the /v1/iam/login query (next to code_challenge).
Additive + forward-only (never defaulted); social login already rode authorize()
which carried nonce. Verified: id SPA builds green.
2026-06-23 18:17:30 -07:00
Antje Worring d89222ad27 ci(docker): compute version without node (ARC runner has no node)
The ARC runner image ships git+docker but not node, so the Compute-tags step's
`node -p "require('./package.json').version"` died with exit 127 (node: command
not found), skipping the build. Parse the version from package.json with
portable sed instead. Verified locally -> 0.1.23.
2026-06-23 15:42:29 -07:00
Antje Worring db7acceca8 ci(docker): target ARC scale set by name + mode=min cache
runs-on was [self-hosted, linux, amd64], which GitHub matched to the OFFLINE
classic evo-* org runners instead of the live ephemeral ARC scale set, so every
Docker run since the self-contained-build switch queued forever (never built).
Use runs-on: hanzo-build-linux-amd64 (the exact scaleSetName), matching the
hanzoai/iam convention, so jobs route to scale set 31.

Also switch buildx gha cache to mode=min: the runner builds via DinD on a 32G
node and mode=max exported every layer, ballooning the DinD store past the
kubelet eviction threshold (NodeHasDiskPressure -> build killed 'no space').
2026-06-23 15:40:29 -07:00
Antje Worring fe13917088 fix(auth): org-agnostic password login — resolve real owner-org, not pinned brand
LoginForm no longer pins organization=<brand> on POST /v1/iam/login. It passes
the new (normally-unset) TenantConfig.loginOrg, and client.login() OMITS the
organization field when empty so IAM runs cross-org resolution and the session
encodes the user's REAL owner-org (GetOrganizationByUser), never the hint.

Fixes: global admins (z@/a@hanzo.ai, woo@lux.network — owner=admin) were
truncated to a 1-org 'hanzo' session via the UI because the pinned org made
GetUserByFields hit the colliding hanzo/<name> row first, so the cross-org
fallback to admin/<name> never ran. IAM IsGlobalAdmin()==(Owner=='admin') and
get-organizations returns all orgs only for a global admin. Verified live on
hanzo.id: z@hanzo.ai -> owner=admin, 45 orgs; brand-only user -> 1 org.

Boundaries preserved: signup still sends a concrete org; per-app SSO
(client_id+redirect_uri, type=code) still mints a code bound to the resolved
user; a brand can force single-org login via loginOrg in the catalog ConfigMap.
Contract locked in pkgs/auth/src/client.test.ts.

Bump 0.1.22 -> 0.1.23.
2026-06-23 15:24:32 -07:00
fbb5da8c70 fix(onboarding): never list other tenants' orgs; create-or-skip only (#15)
The org step fetched service.listOrgs() and rendered a pick-list of every
existing organization to a brand-new user — leaking the tenant directory to
anyone who signs up, and not the intended UX. A new user should only create
their own organization or skip; joining an existing org is invitation-based
and handled outside onboarding.

Drop the listOrgs() fetch and the pick-list entirely; the org step now always
shows the create form with a 'Skip for now' action.

Co-authored-by: Darkhorse7stars <z@hanzo.ai>
2026-06-23 12:56:31 -07:00
zeekay 23cac42399 Merge branch 'feat/sms-consent-disclosure' 2026-06-23 11:32:36 -07:00
zeekay 0e10986cc6 feat(auth): A2P SMS consent disclosure on the SMS verification surface
Twilio A2P 10DLC requires the SMS consent disclosure wherever a user enters
or uses a phone number to receive messages. Add one shared, verbatim consent
component (matches hanzo.ai/sms-opt-in SMS_CONSENT_TEXT and the IAM phone-login
UI) and render it on the portal's SMS surface, gated to the SMS channel only.

- New pkgs/auth/src/ui/SmsConsent.tsx: single source of SMS_CONSENT_TEXT plus
  <SmsConsentNotice/> (disclosure + Terms/Privacy links). Exported from the ui
  barrel for reuse by any future phone-collection surface.
- OTPForm: render SmsConsentNotice beneath the code input only when
  channel === 'sms' (not for totp/email).
- app.css: muted styling for .hanzo-id-sms-consent, matching footer-links.

Note: this portal does not render its own phone-number field — phone-number
COLLECTION happens in the IAM-hosted UI (covered separately), where signup uses
a required, submit-gating consent checkbox. The portal's only SMS-facing step
today is the verification-code entry, so the disclosure-only notice applies
here. Typecheck + tests green (pnpm -r tc, @hanzo/id-auth test).
2026-06-23 09:38:27 -07:00
Hanzo 0cf50f1f1b ci(docker): self-contained build on self-hosted runners
The shared reusable workflow (hanzoai/.github docker-build.yml@main) fails graph
validation for EVERY caller right now (org-wide startup_failure, 0 jobs created),
so id never built via CI (0.1.x images were pushed manually). Build the amd64
image directly on our self-hosted runners instead — no shared-infra dependency.
Tags: sha-<short> + package.json version.
2026-06-22 23:30:19 -07:00
Hanzo ddd5f4c409 fix(social): real Google endpoint + registered redirect_uri (iam.hanzo.ai/callback)
The provider hop sent redirect_uri=${origin}/callback (= hanzo.id/callback),
which the shared Google/GitHub OAuth client does NOT accept — verified live the
Google client accepts ONLY https://iam.hanzo.ai/callback (every other URI →
redirect_uri_mismatch). It also used the invalid authorize endpoint
accounts.google.com/signin/oauth.

- social.ts: Google endpoint → https://accounts.google.com/o/oauth2/v2/auth
  (canonical/stable); add callbackOrigin param to buildProviderAuthUrl/
  startProviderLogin (defaults to origin — local/single-host unchanged).
- TenantConfig.oauthCallbackOrigin (catalog-driven; defaults to publicOrigin);
  SocialButtons passes it so the hop returns to the provider's REGISTERED
  /callback. iam.hanzo.ai serves the same @hanzo/id SPA: the headless Callback
  completes the exchange and forwards to the originating app.
- tests: lock the registered-callback override + canonical Google endpoint.

bump id 0.1.1→0.1.22, id-auth/id-shared 0.1.0→0.1.1
2026-06-22 23:26:35 -07:00
Antje Worring e303575006 fix(portal): root IS the login form (logged-out) / apps launcher (logged-in)
- '/' rendered a static 'Welcome' marketing hero; for an identity portal the
  root should BE the login form when signed out, and the org's apps launcher
  when signed in. Portal now reads /v1/iam/get-account (same-origin, first-party
  cookie) and renders <Login> (anon) or the apps grid (authed).
- Onboarding completion redirected to '/' → the hero, which read as 'looped back
  to the beginning' after the last (wallet) step's Skip. It now lands on
  '/?signed_in=1' → the authed apps launcher. (The step machine was already
  correct: nextStep('wallet')==='done'.)
- Port marketing.ts (appsFor/billingFor) from the restore-design line.
2026-06-22 22:05:14 -07:00
zeekay 106818b9a1 fix(auth): use feat's clean LoginForm/SignupForm (getAppLogin via SocialButtons; password+PKCE) — auto-merge had mixed main's appLogin/sendLoginCode API with feat's client 2026-06-21 19:43:15 -07:00
zeekay 5d29cf4a55 fix(docker): COPY pkgs/onboarding/package.json — merged workspace needs it for pnpm install 2026-06-21 19:37:00 -07:00
zeekay b879084af0 merge(id): integrate feat/full-auth-onboarding — PKCE password-login fix + composable auth UI + social hop into main
# Conflicts:
#	Dockerfile
#	apps/web/src/App.tsx
#	apps/web/src/app.css
#	pkgs/auth/src/client.ts
#	pkgs/auth/src/index.ts
#	pkgs/auth/src/ui/index.ts
#	pkgs/shared/src/brand.ts
#	pkgs/shared/src/tenant.ts
2026-06-21 19:35:05 -07:00
zeekay b425ce9979 feat(social): wire the provider hop + return exchange (gated until creds seeded)
Prep social login end-to-end so only OAuth-app registration + KMS creds remain.

- The hop (social.ts/startProviderLogin): redirects straight to GitHub/Google with a base64 state that round-trips the original authorize request — the Casdoor getAuthUrl contract. Replaces the @hanzo/iam signinRedirect that looped back to the login page. Pure URL builder is unit-tested (social.test.ts, 5 cases).

- SocialButtons now keeps the full provider records (type/clientId/scopes surfaced from get-app-login) and routes OAuth providers through the hop, wallet through the SDK.

- Callback gained a GATED provider-return branch: a provider state (base64 with provider+application) is exchanged via client.providerLogin at the IAM backend, then follows the continue-URL back into the normal OIDC path. The existing OIDC/password path is the untouched else-branch — and the social branch is unreachable until a configured provider exists, so zero impact today.

- Cred-sync already exists (init_data.json ${IAM_GITHUB_CLIENT_ID} <- iam-kms-sync KMSSecret, project hanzo-iam/prod). Runbook in LLM.md. End-to-end needs live verification once real creds are seeded.
2026-06-20 10:46:26 -07:00
zeekay be4ea25d2f fix(login): deliver tenant catalog via /config.json; correct pars/osage/lux client wiring
Root cause osage.id still showed Hanzo: the SPA read the catalog from window.__ID_CATALOG__, which the runtime never injects — so every catalog-only host silently fell back to the bundled Hanzo default. Read the catalog from /config.json (what the static server actually templates from SPA_IAM_TENANT_CONFIG_JSON), with the global as a fallback.

Also corrected verified-broken client wiring: built-in pars-id does not exist (use pars-console, which carries the /callback redirect); added an osage built-in so a catalog-load failure can never leak the Hanzo brand. lux uses lux-id (lux-cloud lacks the portal redirect) — fixed in the ConfigMap. Tests: 7/7.
2026-06-20 02:51:08 -07:00
zeekay 29e85db41d fix(tenant): catalog-only hosts resolve their OWN brand; drop orphan divider
osage.id (and zoolabs.id) rendered as Hanzo — a white-label leak. Root cause: the runtime catalog entries carry brandUrl (+orgId/clientId), but a host with NO built-in DEFAULT_TENANTS entry used DEFAULT_TENANTS['hanzo.id'] as its merge base, so brandPackage/iamUrl silently inherited Hanzo. Now a catalog-only host derives iamUrl/iamIssuer/publicOrigin from the host itself and maps brandUrl -> brandPackage; it can never inherit another brand. Unknown brand asset -> neutral wordmark, not the wrong brand. Regression test added (pkgs/shared now has a test runner).

Drop the orphan 'or' divider: it lived between the social block and the password form, but with unconfigured social hidden it dangled above the form on every brand. Moved the divider INTO SocialButtons so it renders only alongside actual buttons; removed the standalone one from Login + Signup.
2026-06-20 02:38:31 -07:00
zeekay d6af481d72 fix(login): only show configured social providers; never show a broken brand logo
Two things made login feel broken across every brand on the portal (hanzo.id/lux.id/pars.id/osage.id):

1) Social SSO dead-ended. The GitHub/Google/Apple/Web3 providers are seeded with PLACEHOLDER credentials (clientId=GITHUB_CLIENT_ID_PLACEHOLDER, etc.), so the OAuth redirect can never complete — clicking 'Continue with GitHub' just looped back to the login form. Render ONLY providers IAM actually holds a real clientId for (AppProvider.configured, computed from the nested provider record). With placeholders that hides all social buttons, leaving the working methods (email/password + email/SMS code) — and the buttons reappear automatically once real OAuth creds are seeded, no code change. When get-app-login is unreadable, render none rather than risk a dead-end.

2) Broken logo. brand.json points logo/favicon at jsdelivr @hanzo/brand assets, but the published brand packages ship no assets/ dir (404). BrandHeader now falls back to the brand NAME as a text wordmark on image error/absence, so a missing logo never renders a broken-image icon.

Enabling real social login later needs: (a) register OAuth apps (GitHub/Google/Apple) with callback https://<brand>/v1/iam/callback, (b) put their client_id/secret in KMS → IAM provider records, (c) the SPA provider-redirect hop. (a)+(b) are external provisioning; until then social is correctly hidden.
2026-06-20 02:21:35 -07:00
zeekay a22f3c9648 fix(auth): forward PKCE code_challenge on password login
login() built /v1/iam/login with clientId/responseType/redirectUri/state
but dropped the PKCE code_challenge/code_challenge_method that authorize()
already forwards. A downstream public SPA client (e.g. hanzo-platform)
sending a user to password sign-in got an auth code minted with an empty
stored challenge; the downstream /auth/callback token exchange then failed
PKCE validation and fell back to a client_secret check the public client
can't satisfy (token.CodeChallenge: empty -> invalid_client).

IAM's Login handler (controllers/account.go) reads code_challenge from the
query first, body fallback, so forward it on the query mirroring authorize().
Plumb it from the OAuth params already on the /login URL through Login page
-> LoginForm -> client.login(). Social login was unaffected (rides authorize).
2026-06-20 01:36:59 -07:00
zeekay 6c7e083100 fix(docker): bake SPA-appropriate CSP into the image (HANZO_STATIC_CSP)
hanzoai/static defaults CSP to `default-src 'none'` which blocks the
SPA's own bundle (no script-src) -> blank page. Bake a widened-but-locked
CSP that permits: self scripts + CF beacon, same-origin + jsDelivr brand
fetches, https/data images (brand logos), inline styles (@hanzo/gui).
Baking it (vs a deploy-time env) makes the image render correctly
standalone and survives universe reconcile.
2026-06-20 00:17:30 -07:00
zeekay f6b7b9327c fix(brand): emit + fetch brand.json at flat encoding-safe /brand/<scope>.json
Two coupled bugs broke runtime brand loading on the hanzoai/static
(scratch) image:

1. vite.config.ts is ESM ("type":"module") so the brandJsonPlugin's
   require.resolve('@scope/brand/brand.json') threw ReferenceError
   (no global require), was swallowed by the silent catch, and NO
   brand.json was ever emitted into dist/. Fixed with
   createRequire(import.meta.url).

2. The brand fetch path /brand/@hanzo/brand/brand.json carries a literal
   @ and an encoded %2F that the production static server cannot map to
   the on-disk file, so it falls through to the SPA catch-all and returns
   index.html — the SPA then parses HTML as JSON. Emit and fetch at a
   FLAT slug path instead: @hanzo/brand -> /brand/hanzo.json. The plugin
   and loadBrand derive the slug identically (npm scope).

With defensive loadBrand (prev commit) a miss degrades to a neutral
brand; this restores real per-brand branding.
2026-06-20 00:07:06 -07:00
zeekay dc3f888355 fix(brand): never blank the login form on a transient brand.json failure
The browser brand loader threw on any non-ok /brand/<pkg>/brand.json fetch, so a single transient 502 on the cosmetic brand asset (intermittent behind Cloudflare) left the IAM login SPA blank — no form, just 'brand.json fetch failed: 502'. Retry the flaky asset up to 3x with backoff, then fall back to a neutral per-tenant brand so the form always renders. brand.json stays the source of truth on the happy path.
2026-06-19 23:54:11 -07:00
zeekay 6e33166f13 fix(docker): serve SPA from /public with --spa for hanzoai/static contract
hanzoai/static:0.4.1 is FROM scratch + ENTRYPOINT ["/static"]; the
binary defaults to -root /public -port 3000 and writes
/public/config.json from SPA_* env at boot. The previous final stage
copied the bundle to /spa and set dead ENV ROOT/PORT (the binary reads
flags, not env), so /public never existed and the runtime-config write
crashed (open /public/config.json.tmp: no such file or directory).

Copy the dist to /public (the default root + config.json target) and
pass --spa so client-routed paths (/auth/*, /callback) fall back to
index.html. Port 3000 matches the id Service and probes.
2026-06-19 23:48:00 -07:00
zeekay 7d91d22904 fix(docker): COPY pkgs/onboarding/package.json into install stage
apps/web depends on @hanzo/id-onboarding (workspace:*) since the
onboarding flow landed, but the Dockerfile's per-package manifest
COPY list (the layer feeding the dependency-resolution
`pnpm install` before the full `COPY pkgs pkgs`) was never
updated. In-cluster builds therefore failed at install with
ERR_PNPM_WORKSPACE_PKG_NOT_FOUND for @hanzo/id-onboarding.

Local builds masked this because node_modules was already
populated from a prior full install.
2026-06-19 23:40:16 -07:00
zeekay 670f69f604 test(onboarding): unit tests for step machine + IAM wire contracts; docs
9 tests via the Node built-in runner (--experimental-strip-types, no
test-framework dependency): step machine (org→project→wallet→done),
listOrgs mapping + error-resilience, createOrg request shape + error
surfacing, linkWallet address validation + get-account→update-user
(web3onboard, column-scoped) + fail-closed. tc excludes *.test.ts so the
build gate stays type-only.

LLM.md: corrected issuer to per-brand *.id host (not iam.hanzo.ai),
documented the full method set + get-app-login source-of-truth, the
onboarding pkg + its IAM routes + admin-gating reality, and the /v1/iam
OIDC paths.
2026-06-19 15:34:37 -07:00
zeekay 647bb88353 fix(auth): social/Web3 sign-in honors downstream redirect_uri
Social + Web3 always return to the portal's own /callback (the SDK's fixed
redirectUri), so a downstream app's redirect_uri would be lost. SocialButtons
now stashes it in post_login_redirect before signinRedirect; Callback reads it
back and forwards the tokens there, else lands on /onboarding. Login + Signup
pass redirect_uri through. Password path already forwards directly via the
auth-code response — both methods now reach the same downstream target.
2026-06-19 15:31:21 -07:00
zeekay 2365f83855 fix(onboarding): correct IAM contracts — get-account+web3onboard wallet, authz-aware org/project
linkWallet now resolves the signed-in user via /v1/iam/get-account (owner/name)
and writes the lowercase `web3onboard` column scoped via ?columns= so the rest
of the user row is untouched — IAM's update-user is keyed by owner/name, not a
self alias.

Org/project creation is admin-gated in IAM authz (add-organization needs the
admin role; add-project default-denies non-admins). The common path — pick the
org you signed up into, listed via the *-allowed get-organizations — works for
everyone; create-new surfaces a plain permission message and the step stays
skippable so onboarding never hard-blocks. ProjectStep tolerates an org-less
flow (continue-only).
2026-06-19 15:29:42 -07:00
zeekay 3262d9e47e feat(onboarding): post-login org → project → wallet flow + serverUrl=hanzo.id
@hanzo/id-onboarding: domain (serializable step machine + types) / service
(IAM-backed writes via /v1/iam/{get-organizations,add-organization,add-project,
update-user}, cookie+bearer) / UI (self-contained 3-step OnboardingFlow, no
router lib). Web app mounts it at /onboarding; both password (cookie) and
social/Web3 (SDK token) sign-in land there for a bare portal login.

Callback now completes via the @hanzo/iam SDK handleCallback (matches
SocialButtons' signinRedirect) and forwards tokens to a downstream app only
when one initiated the flow.

tenant.ts: iamUrl is the per-brand OIDC issuer host (hanzo.id / lux.id /
zoo.id / pars.id), never iam.hanzo.ai — HIP-0111 host-relative discovery.
clientId is the brand -id app (hanzo-id), matching init_data.json.
2026-06-19 15:25:04 -07:00
zeekay 2132c13e0c feat(auth): full method set — email/password + GitHub + Google + Web3
SocialButtons reads live enabled providers from /v1/iam/get-app-login and
drives the @hanzo/iam PKCE redirect (provider param) per method. Login/Signup
render the social row + divider above the email form. AppLogin/AppProvider
types model the get-app-login view; createIam() centralizes the one PKCE
client both SocialButtons and Callback share.
2026-06-19 15:19:50 -07:00
818ae385e5 fix(brand): resilient brand loader — never blank the login page (#14)
loadBrand did res.json() on the SPA-fallback HTML (the spa server answers
unknown paths with index.html + HTTP 200 for client routing), so a missing
brand.json threw 'Unexpected token <' and crashed App boot -> blank login
page. brand.json is not bundled and the path also wrongly encoded the '/'
in '@hanzo/brand' (%2F), so it never resolved.

Fix:
- Prefer the tenant's brandUrl (the working jsDelivr brand.json from
  config.json), then the app-local path; reject non-JSON (SPA-fallback HTML)
  responses by content-type; on total failure return a minimal fallback brand
  derived from the package scope. Branding is cosmetic and must never block login.
- Add TenantConfig.brandUrl (already supplied by config.json) and pass it from App.
- Drop encodeURIComponent on the whole package (it mangled the scope slash).

Co-authored-by: Antje Worring <worringantje@gmail.com>
2026-06-18 17:56:31 -07:00
5b9555ada6 fix(docker): pin hanzoai/spa:1.2.0 (1.3.0 image not published) (#13)
The v1.3.0 git tag exists but no 1.3.0 image was published to ghcr
(build pipeline blocked). 1.2.0 is the latest published spa image and the
version the canonical RECIPE.md uses.

Co-authored-by: Antje Worring <worringantje@gmail.com>
2026-06-18 16:18:13 -07:00
704b6989a4 fix(docker): serve SPA via hanzoai/spa, not hanzoai/static (#12)
hanzoai/static defaults to Content-Security-Policy: default-src 'none'
(no script-src) — built for static assets, not a SPA that loads its own
JS bundle. That CSP blocks index-*.js, so React never mounts and the
login page renders blank. hanzoai/spa is the purpose-built base: history-
API fallthrough for client-side routes + a SPA-safe CSP. Defaults
PORT=3000 / ROOT=/public, matching the id deploy probe.

Immediate prod was unblocked by setting HANZO_STATIC_CSP on the live
deployment; this is the durable, one-way fix so the override hack isn't
needed and every rebuild serves correctly.

Co-authored-by: Antje Worring <worringantje@gmail.com>
2026-06-18 16:14:27 -07:00
b0b0c1addf fix(docker): static needs -root/-spa flags, not ROOT/PORT env (#11)
hanzoai/static:0.4.1 reads the serve root from the -root flag (default
/public) and port from -port (default 3000) — the ENV ROOT=/spa PORT=8080
were no-ops, so the built image served /public (crash) on :8080 while the
deploy probes :3000. Set ENTRYPOINT [/static -root /spa -spa] (default
port 3000). Matches the hand-built id:0.1.5 now live.

Co-authored-by: Antje Worring <worringantje@gmail.com>
2026-06-18 15:33:36 -07:00
172b4ddf6d fix(id): appLogin uses /v1/iam/get-app-login (main dropped /api/*) (#10)
Merged IAM main serves the canonical /v1/iam/* only; /api/get-app-login
now returns the SPA shell. Point the providers/methods fetch at the
canonical path so social buttons + sign-in methods load.

Co-authored-by: Antje Worring <worringantje@gmail.com>
2026-06-18 14:08:32 -07:00
c0f11dcf28 feat(id): render social logins + email + SMS code on login/signup (#9)
The Vite portal rendered only email/password. Add — driven by the live
IAM app config (new AuthClient.appLogin -> get-app-login), no hardcoded
lists:
- social provider buttons (GitHub/Google/Apple/Web3) via authorize?provider=
- passwordless email / SMS code sign-in (send-verification-code + OTPForm)
- new ProviderButtons component; LoginForm + SignupForm render providers
Email/password sign-in unchanged. tsc --noEmit clean; vite build clean.

Functional prerequisites (display works now; these make it WORK):
- real OAuth client IDs for github/google/apple (IAM currently has
  GITHUB_CLIENT_ID_PLACEHOLDER etc.)
- an SMS gateway provider (Twilio/etc.) in IAM for SMS codes
- redirectUris backfill so the /callback OAuth return is allowed (iam#51)

Co-authored-by: Antje Worring <worringantje@gmail.com>
2026-06-18 13:42:41 -07:00
f04eb29f28 chore(deps): clear Dependabot alerts — drop legacy-nextjs, patch esbuild/uuid (#8)
Resolves all 35 open Dependabot alerts on hanzoai/id:

- Delete legacy-nextjs/ (frozen Next.js predecessor, not wired into any
  build/Dockerfile/workspace glob; docs slated it for deletion after
  v0.1.0 — v0.1.1 has shipped). Removes 31 alerts (next, undici, ws,
  picomatch, postcss, js-yaml, cookie, defu, ...).

- pnpm overrides for the 3 active-workspace transitives:
    esbuild ^0.28.1  (was 0.27.7) — GHSA-gv7w-rqvm-qjhr (high), GHSA-g7r4-m6w7-qqqr (low)
    uuid    ^11.1.1  (was 7.0.3/10/x via xcode@3.0.1) — GHSA-w5hq-g745-h8pq (medium)

Verified: pnpm install + typecheck (4/4) + build green (vite 7.3.5,
esbuild 0.28.1, 52 modules); 'pnpm audit' → no known vulnerabilities.

Co-authored-by: Antje Worring <worringantje@gmail.com>
2026-06-18 12:23:57 -07:00
70b1330df2 fix(tenant): use real IAM clientId (<org>-id, not <org>-id-portal) (#7)
The IAM apps are registered with clientId '<org>-id' (hanzo-id, lux-id,
zoo-id, pars-id) — there is no '-portal' variant. clientId 'hanzo-id-portal'
returns 'Invalid client_id' from IAM, breaking the login flow. Align all
four tenant clientIds with their appName and the IAM seed convention.

Co-authored-by: Antje Worring <worringantje@gmail.com>
2026-06-18 11:34:50 -07:00
583e3454dc fix(id): pin @hanzo/iam to published ^0.9.4 (0.10.0 unpublished, breaks pnpm install) (#6)
Co-authored-by: hanzo-dev <dev@hanzo.ai>
Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-06-17 04:03:28 -07:00
7a9ac97526 chore(decomplect): deploy + per-host catalog live in universe, not the app repo (#5)
The id repo is brand-neutral (zero brand data bundled), but apps/web/k8s/
carried the per-host brand catalog + Deployment/Ingress — duplicated in
hanzoai/universe infra/k8s/id/ (the single deploy source-of-truth). Removed
the overlay so app (brand-neutral image) and deploy (per-host catalog +
manifests, in universe) are separated. One way, one place.

Co-authored-by: hanzo-dev <dev@hanzo.ai>
Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-06-16 18:05:31 -07:00
cff508e2f0 fix(auth): IAM wire contract (body params) + brand-app catalog (v0.1.1) (#4)
* fix(auth): IAM wire contract + brand-app catalog

- client.ts: send type/application/organization in the request BODY (IAM reads
  auth fields from body, OAuth params from query) so the branded login form
  authenticates; build the code-redirect from the response.
- canonical /v1/iam/oauth/{authorize,token,logout} paths.
- tenant-catalog: point each host at its brand's real IAM app
  (hanzo-console/lux-cloud/zoo-console/pars-console).

Co-authored-by: Hanzo Dev <dev@hanzo.ai>

* chore(release): 0.1.1

---------

Co-authored-by: hanzo-dev <dev@hanzo.ai>
Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-06-16 18:00:41 -07:00
hanzo-dev ef6263d666 chore: update 2026-06-10 14:12:32 -07:00
hanzo-dev 98c78c7ad1 merge: id-vite-monorepo (-X theirs) 2026-06-01 18:09:45 -07:00
hanzo-dev 4c89defe95 merge: ci/canonical-docker-build-1776995843 (-X theirs) 2026-06-01 18:09:44 -07:00
hanzo-dev fb06c01189 refactor: brand-neutral identity portal (zero brand-specific code)
Image carries no brand identity. Every per-tenant value comes from the
runtime catalog (K8s ConfigMap → /config.json → window.__ID_CATALOG__)
or is derived from the hostname.

Source changes:
- TenantConfig.brandPackage → brandUrl (absolute URL to brand.json).
  Brands self-host (npm + jsDelivr is convention; any URL works).
- DEFAULT_TENANTS = {} (deleted). Hostname-derivation is the new fallback:
    foo.id / www.foo.id / id.foo.net / iam.foo.net → orgId=foo,
    clientId=foo-id-portal, appName=foo-id,
    brandUrl=https://cdn.jsdelivr.net/npm/@foo/brand@latest/brand.json
  Works out-of-the-box when the npm scope matches the org. The catalog
  handles mismatches.
- brand.ts loadBrand(brandUrl) fetches the URL directly. localizeAssets()
  removed — brand.json's URLs are absolute, used as-is.
- vite.config.ts: removed brandJsonPlugin + BRAND_PACKAGES list. Vite
  bundles no brand assets. dist drops from ~250kB to 205kB (64kB gzip).
- apps/web/package.json: dropped @hanzo/brand, @luxfi/brand, @zooai/brand,
  @parsdao/brand deps. Zero brand packages in the image.
- main.tsx: fetches /config.json before mount; sets
  window.__ID_CATALOG__ from cfg.iamTenantConfigJson (templated by
  hanzoai/spa runtime from SPA_IAM_TENANT_CONFIG_JSON env var).
- App.tsx: loadBrand(t.brandUrl).
- LLM.md: documents the brand-neutral architecture and how to add a
  brand (publish brand.json + add to deploy's catalog ConfigMap; image
  never changes).

Deploy changes:
- apps/web/k8s/tenant-catalog.yaml: NEW ConfigMap carrying the Hanzo
  deployment's full host→tenant map for 11 hosts (hanzo / lux / zoo /
  pars / osage and their id.*.network / iam.*.network / www.* variants).
  Brand-specific knowledge lives ONLY here.
- apps/web/k8s/deployment.yaml: envFrom: configMapRef: id-tenant-catalog.
- apps/web/k8s/kustomization.yaml: includes tenant-catalog.yaml first.

Adding a brand from here on is:
  1. publish @<scope>/brand to npm with brand.json
  2. add the host(s) to the deployment's ConfigMap
  3. add the host to ingress.yaml (cert-manager auto-provisions TLS)
  4. DNS → cluster ingress IP

No source change. No image rebuild.
2026-05-29 11:37:05 -07:00
hanzo-dev 5dfbbad5c1 feat(k8s): add www.zoolabs.id + osage.id + www.osage.id to id ingress
cert-manager.io/cluster-issuer=letsencrypt-prod-cf provisions per-host
TLS via DNS-01 (the Traefik file-route certResolver path is non-functional
in this cluster — CF env vars on the ingress deployment are empty, ACME
in /data/acme.json is empty). cert-manager owns TLS for identity hosts.

After apply + cert issuance (~90s): all 7 identity hosts return their
own OIDC issuer with valid LE certs:
  lux.id, hanzo.id, pars.id, zoolabs.id, www.zoolabs.id, osage.id, www.osage.id
2026-05-29 11:07:17 -07:00
hanzo-dev d43e700f41 feat(tenant): add osage.id + www.osage.id + www.zoolabs.id tenants
DEFAULT_TENANTS now resolves all 4 active identity hosts that didn't
have built-in entries:
- www.zoolabs.id → orgId=zoo, brand=@zooai/brand
- osage.id      → orgId=osage, brand=@osage/brand
- www.osage.id  → orgId=osage, brand=@osage/brand

zoolabs.id was already present. orgId, iamIssuer, clientId, appName,
publicOrigin, brandPackage all follow the existing per-host shape.

@osage/brand is unpublished (~/work/osage/brand v0.1.0 needs build +
publish). Until then, /brand/@osage/brand/brand.json 404s and the
SPA's loadBrand falls back to the @hanzo/brand default — adequate
until osage.id DNS flips off Cloudflare Pages.
2026-05-28 18:27:27 -07:00
hanzo-dev be087c496a chore: update 2026-05-25 15:14:35 -07:00
hanzo-dev b7628eee35 feat: add zoolabs.id (canonical Zoo identity host; was zoo.id)
Per user clarification, the Zoo identity domain is zoolabs.id (zoo.id is
not owned). Add tenant entry + Ingress host + TLS secret. DNS A record
already repointed in CF to 129.212.164.5.
2026-05-24 14:10:33 -07:00
hanzo-dev 7753b0ce78 fix(tenant): hanzo.id clientId was mangled by previous sed (id-portal-portal → hanzo-id-portal) 2026-05-24 10:32:18 -07:00
hanzo-dev 8d7031cf3d style: rename CSS classes hanzo-id-* → id-portal-*
Brand-neutral CSS class names match the public-facing identity of the
portal (id-portal). Cosmetic only; no rendered text changes. Per
playwright agent's source-view note.
2026-05-24 10:31:03 -07:00
hanzo-dev 30c7d0a8c4 feat: extend id portal to .network identity hosts
Add iam.lux.network + id.lux.network (Lux brand) + id.zoo.network
(Zoo brand) to the id Ingress + tenant catalog. Same image, same code
path — only DEFAULT_TENANTS + Ingress hosts/tls grow.

Cutover plan (post-merge):
  1. kubectl apply -k infra/k8s/id  (cert-manager issues 3 new TLS certs via DNS-01)
  2. Repoint CF DNS A records → 129.212.164.5
  3. Delete the 3 hanzo-id Worker routes
  4. Archive hanzoai/hanzo.id-worker repo (no routes remaining)
2026-05-24 10:12:23 -07:00
hanzo-dev 9b54655329 feat(brand): bundle brand logo SVGs as static assets
The published @hanzo/brand / @luxfi/brand / @zooai/brand / @parsdao/brand
npm tarballs don't ship the assets/ directory (only dist/ + brand.json
per the pkg's 'files' field). The Vite plugin can only emit what's
present on disk; result: 404 on /brand/<pkg>/assets/logo/logo.svg.

Workaround until the brand pkgs republish with assets included: commit
the SVGs into apps/web/public/brand/<pkg>/assets/logo/ so they ship in
the SPA bundle directly. Same path the localizeAssets rewrite produces,
so loadBrand() output is unchanged.

To update: re-copy from ~/work/<org>/brand/assets/logo/ and bump
@hanzo/id patch.
2026-05-24 09:39:59 -07:00
hanzo-dev 17b98ae11c fix(brand): regex now matches @scoped/name pkg URLs
The old regex [^@/]+ excluded the leading @ in @scope/name pkg paths
on jsdelivr, so the rewrite silently passed through to the upstream
CDN URL and the browser rendered a broken-image glyph. Add @? prefix.
2026-05-24 09:32:47 -07:00
hanzo-dev feca00917a fix(brand): serve logo + assets from /brand/<pkg>/, switch to DNS-01 issuer
playwright agent flagged: brand.json works but logo SVG 404s on
jsdelivr because the brand pkgs' npm tarballs don't ship assets/logo/.

Fix at this side instead of waiting on a brand-pkg republish:
  1. vite plugin now serves ALL files under <pkg>/assets/ (not just
     brand.json), at /brand/<pkg>/assets/...
  2. loadBrand() in pkgs/shared rewrites the brand.json logoUrl and
     faviconUrl from /npm/<pkg>@latest/<rest> → /brand/<pkg>/<rest>.
     Only the brand's own pkg URLs rewrite; third-party CDN refs pass
     through unchanged.

Also flip the cert-manager cluster-issuer for the id Ingress from
letsencrypt-prod (HTTP-01) to letsencrypt-prod-cf (DNS-01 via
Cloudflare). The Ingress unconditionally 308-redirects all HTTP→HTTPS
including /.well-known/acme-challenge/*, so HTTP-01 never solves.
DNS-01 via CF API works regardless of HTTP path routing.
2026-05-24 09:17:36 -07:00
hanzo-dev 43371d76f4 fix(vite): brand-json plugin actually emits files at build time
The previous plugin used require.resolve in ESM context which failed
silently — generateBundle had no warning and no assets emitted. Result:
production image had no /brand/<pkg>/brand.json files, hanzoai/spa
fell back to index.html for those paths, and the browser tried to JSON-parse
HTML → 'Unexpected token <' on every host.

Fix: use createRequire(import.meta.url) + paths fallback walk through
node_modules. emitFile now writes dist/brand/<pkg>/brand.json.

Verified by: pnpm build outputs dist/brand/@hanzo/brand/brand.json etc.
playwright report flagged this as the root cause of the white-label
blank page across hanzo.id / lux.id / pars.id.
2026-05-24 08:43:02 -07:00
hanzo-dev 6678b9b9fd fix(docker): switch to hanzoai/spa (zero-config SPA server)
hanzoai/static is for traditional static-file serving (no SPA fallback,
needs --spa flag for client-side routing, /healthz only). hanzoai/spa is
purpose-built for SPAs: SPA mode always on, runtime config via SPA_* env
vars, /health endpoint. K8s probes flip /healthz → /health to match.
2026-05-24 07:46:08 -07:00
hanzo-dev 6748c62e22 fix(docker): static image serves /public on :3000 (not /spa:8080)
hanzoai/static:0.4.1 is hardcoded to read /public and listen on :3000.
PORT/ROOT env vars from the Dockerfile are ignored. Move dist there.

K8s deployment containerPort + probes flipped from 8080 to 3000 to match.
Service targetPort already :3000. Service port 80 stays the same.
2026-05-24 07:12:28 -07:00
hanzo-dev 7e359f279c deps: pin @hanzo/iam to ^0.9.4 (latest published); drop @hanzo/gui (not published yet)
The 0.10.0 / 7.2.4 versions in source are unpublished WIP. Use the latest
public-npm versions so the Docker build can resolve. @hanzo/gui re-adds
once 7.2.x publishes to npm (per hanzoai/gui PR #2 dist/ emit fix).
2026-05-24 06:57:56 -07:00
hanzo-dev 981cce26e3 ci: amd64-only (arm64 ARC pool paused on DOKS) 2026-05-24 06:50:23 -07:00
hanzo-dev 86c961bbaf ci: drop pre-build-command — Dockerfile self-contains pnpm build 2026-05-24 06:38:53 -07:00
hanzo-dev fce00d8549 ci: drop CF Pages deploy (was Next.js-only; Vite SPA ships via Docker) 2026-05-24 06:25:55 -07:00
hanzo-dev 919d261d0a ci: add id-token permission + allow pnpm to generate lockfile
Tag-push of v0.1.0 hit startup_failure because the caller workflow lacked
id-token: write. The hanzoai/.github reusable docker-build.yml requires it
at the caller's top-level (see universe LLM.md, 2026-05-05 sprint notes).

Also drop --frozen-lockfile since this is a fresh rewrite without a
checked-in lockfile yet — let pnpm generate one.
2026-05-24 06:09:28 -07:00
hanzo-devandGitHub 8b85781327 rewrite: Vite + @hanzo/gui monorepo (drops CF Worker + Next.js) (#2)
Replaces the Cloudflare Worker (hanzo.id-worker) and the Next.js portal
with a pnpm monorepo following the a-monorepo/id pattern.

Layout:
  apps/web/        Vite + React 19 SPA, embeds @hanzo/gui shell
    k8s/           Deployment(2) + Service + Ingress (4 hosts, 4 TLS)
  pkgs/shared/     @hanzo/id-shared — TenantConfig, resolveTenant,
                                       loadBrand (browser + node)
  pkgs/auth/       @hanzo/id-auth   — AuthClient (wraps @hanzo/iam REST)
                                       + LoginForm/SignupForm/ForgotForm/OTPForm
  pkgs/idv/        @hanzo/id-idv    — pluggable IDV (stub, persona,
                                       onfido, veriff) behind one
                                       IDVProvider interface
  legacy-nextjs/   Frozen — Next.js predecessor. Delete after v0.1.0 ships.
  Dockerfile       Two-stage: pnpm build → hanzoai/static:0.4.1 serves /spa
  README.md
  LLM.md           Architecture, dev, deploy, cutover plan

Tenant resolution: hostname → TenantConfig (orgId, iamUrl, clientId,
appName, publicOrigin, brandPackage). Built-in defaults for
hanzo.id/lux.id/zoo.id/pars.id; runtime override via
IAM_TENANT_CONFIG_JSON env (served as /config.json at pod startup).

Brand resolution: each per-org brand pkg (@hanzo/brand, @luxfi/brand,
@zooai/brand, @parsdao/brand) ships brand.json. The Vite plugin
brandJsonPlugin emits /brand/<pkg>/brand.json verbatim; the browser
fetches the right one based on the resolved tenant. No bundle bloat.

IDV: stub (default for dev), persona, onfido, veriff. Each adapter
implements `IDVProvider` from pkgs/idv/src/provider.ts. Swap providers
with one registration call at boot — portal code unchanged.

Cutover (separate ops PR — not in this commit):
  1. Tag + push image ghcr.io/hanzoai/id:0.1.0
  2. Apply k8s manifests, cert-manager issues TLS
  3. Remove CF Worker routes for hanzo.id/lux.id/zoo.id/pars.id
  4. CF A records → 129.212.164.5 (hanzo ingress LB)
  5. Archive hanzo.id-worker repo
2026-05-24 06:00:20 -07:00
hanzo-dev 3d2001dab7 rewrite: Vite + @hanzo/gui monorepo (drops CF Worker + Next.js)
Replaces the Cloudflare Worker (hanzo.id-worker) and the Next.js portal
with a pnpm monorepo following the a-monorepo/id pattern.

Layout:
  apps/web/        Vite + React 19 SPA, embeds @hanzo/gui shell
    k8s/           Deployment(2) + Service + Ingress (4 hosts, 4 TLS)
  pkgs/shared/     @hanzo/id-shared — TenantConfig, resolveTenant,
                                       loadBrand (browser + node)
  pkgs/auth/       @hanzo/id-auth   — AuthClient (wraps @hanzo/iam REST)
                                       + LoginForm/SignupForm/ForgotForm/OTPForm
  pkgs/idv/        @hanzo/id-idv    — pluggable IDV (stub, persona,
                                       onfido, veriff) behind one
                                       IDVProvider interface
  legacy-nextjs/   Frozen — Next.js predecessor. Delete after v0.1.0 ships.
  Dockerfile       Two-stage: pnpm build → hanzoai/static:0.4.1 serves /spa
  README.md
  LLM.md           Architecture, dev, deploy, cutover plan

Tenant resolution: hostname → TenantConfig (orgId, iamUrl, clientId,
appName, publicOrigin, brandPackage). Built-in defaults for
hanzo.id/lux.id/zoo.id/pars.id; runtime override via
IAM_TENANT_CONFIG_JSON env (served as /config.json at pod startup).

Brand resolution: each per-org brand pkg (@hanzo/brand, @luxfi/brand,
@zooai/brand, @parsdao/brand) ships brand.json. The Vite plugin
brandJsonPlugin emits /brand/<pkg>/brand.json verbatim; the browser
fetches the right one based on the resolved tenant. No bundle bloat.

IDV: stub (default for dev), persona, onfido, veriff. Each adapter
implements `IDVProvider` from pkgs/idv/src/provider.ts. Swap providers
with one registration call at boot — portal code unchanged.

Cutover (separate ops PR — not in this commit):
  1. Tag + push image ghcr.io/hanzoai/id:0.1.0
  2. Apply k8s manifests, cert-manager issues TLS
  3. Remove CF Worker routes for hanzo.id/lux.id/zoo.id/pars.id
  4. CF A records → 129.212.164.5 (hanzo ingress LB)
  5. Archive hanzo.id-worker repo
2026-05-24 01:50:39 -07:00
hanzo-dev 3bdccc0095 fix(middleware): proxy /v1/iam/* canonical IAM surface
Bug: hanzo.id returned 405 on POST /v1/iam/login because:

1. The matcher had no /v1/iam/:path* rule — the middleware never fired,
   the static SPA had no POST handler at that path, CF Pages returned
   405 directly.
2. IAM_PATH_PREFIXES carried '/api/' but not '/v1/iam/' — even if the
   matcher did fire, shouldProxyToIAM() returned false.
3. PATH_REWRITES mapped RFC OAuth paths (/oauth/token, /oauth/userinfo,
   …) onto legacy /api/* targets — wrong direction; IAM serves /v1/iam/*
   natively.

Fix: drop the entire /api/* hop. RFC aliases now collapse onto canonical
/v1/iam/* targets (one-way mapping, no legacy detour). The matcher lists
/v1/iam/:path*. SPA components, lib/oauth.ts, and the server-side logout
handler all call /v1/iam/* directly. The discovery rewriter strips the
canonical /v1/iam/* form back to RFC-public /oauth/* for OIDC clients.

Net: one canonical surface (/v1/iam/*), three RFC-spec public aliases
(/oauth/*, /login/oauth/*, /.well-known/*). No /api/* anywhere in the
caller path.
2026-05-15 14:35:21 -07:00
hanzo-devandGitHub a8a6ad353d ci: migrate to canonical hanzoai/.github/docker-build.yml reusable (#1) 2026-04-23 18:58:37 -07:00
hanzo-dev dddac31dc5 ci: migrate to canonical hanzoai/.github/docker-build.yml reusable 2026-04-23 18:57:29 -07:00
hanzo-dev 4578efbfc5 feat: add id.lux.cloud as Lux tenant (was defaulting to Hanzo) 2026-04-20 21:37:05 -07:00
hanzo-dev d4eaeabedb refactor: flatten id-{dev,test}.hanzo.ai so *.hanzo.ai Universal SSL covers them 2026-04-20 21:31:45 -07:00
hanzo-dev 64de2b7441 feat: zoolabs.id as canonical Zoo tenant (replaces zoo.id)
zoolabs.id was just acquired to replace zoo.id which we no longer own.
Same Zoo branding + content + socialProviders as id.zoo.network.
2026-04-20 19:45:15 -07:00
hanzo-dev b6748bf8f5 feat(branding): TENANT_BRANDING_JSON env var for runtime white-labeling
Allows any deployment of hanzo-login image to add/override tenants without
white-label deployment to inject their own domain branding + auth providers.

Example:
  env:
    - name: TENANT_BRANDING_JSON
      value: |
        {
            "orgId": "liquidity",
            "orgName": "",
            "content": { "title": "Trade digital securities" },
            "auth": { "socialProviders": ["google", "apple"] }
          }
        }
2026-04-20 19:35:09 -07:00
hanzo-dev bc2106db08 remove: zoo.id tenant — domain no longer owned; use id.zoo.network instead
Zoo Labs ecosystem logins go through id.zoo.network / id.zoo-dev.network / id.zoo-test.network.
zoo.id was removed from TENANTS map, staticBranding, and the hanzo-login ingress.
2026-04-20 19:34:02 -07:00
hanzo-dev 0e2f42256f feat(branding): add id.{hanzo,hanzo-dev,hanzo-test,zoo-dev,zoo-test}.network tenants
Match the {brand}-{env}.network naming scheme. Each env gets its own
subtitle + title so login flow reflects which network the user is on.
2026-04-20 19:26:19 -07:00
hanzo-dev 4d0f06ee34 ci: use ghcr.io/hanzoai/hanzo-login (new package) 2026-04-20 19:15:03 -07:00
hanzo-dev 3f2ceaf839 ci: push to ghcr.io/hanzoai/login (avoid conflict with existing ghcr.io/hanzoai/id) 2026-04-20 19:12:15 -07:00
hanzo-dev c525074995 feat(branding): add dev/test tenants for lux-dev/test.network + dev/test.hanzo.ai 2026-04-20 19:11:37 -07:00
hanzo-dev ffdf3eab1f ci: add docker workflow for ghcr.io/hanzoai/id image 2026-04-20 19:07:51 -07:00
hanzo-dev 1d3b81bd92 feat(login): full marketing panel for lux/zoo + apple social provider
- Add Apple OAuth button (rendered when branding.auth.socialProviders contains 'apple')
- Lux tenant: custom tagline 'Lux-powered infrastructure', title 'Start deploying in seconds'
- Zoo tenant: custom tagline 'Open AI research network', title 'Build the future of DeAI'
- Pars tenant: gets its own content structure (ready for extension)
- All tenants explicitly declare auth feature flags (password/code/webauthn/faceid/social)

Tenants that want apple instead of github: set socialProviders: ['metamask','google','apple']
via IAM /api/branding response — hanzo/id never hardcodes /liquidity.
2026-04-20 19:06:31 -07:00
hanzo-dev 8462c5a4c8 chore: symlink AGENTS.md and CLAUDE.md to LLM.md
Canonical project context lives in LLM.md. Symlinks ensure
agentic coding tools (agent, Cursor, etc.) find context
automatically regardless of which filename they look for.
2026-04-01 14:37:38 -07:00
Darkhorse7starsandhanzo-dev cc64e11712 fix(signup): add username field to IAM signup request
Casdoor's /api/signup expects a `username` field. Without it, the username
defaults to empty string which fails the "at least 2 characters" validation.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-04-01 12:06:37 -05:00
Darkhorse7starsandhanzo-dev d1810354f6 fix: use hanzo-id as IAM application name
IAM enforces that app names start with the org prefix (hanzo-).
The old name 'app-hanzo' fails validation. Changed to 'hanzo-id'
which matches the IAM application entry and passes the prefix check.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-04-01 11:51:09 -05:00
hanzo-dev 8f0ee95754 fix: pass PKCE code_challenge in POST body (Casdoor reads from body, not query) 2026-03-25 21:47:20 -07:00
hanzo-dev 9655226fa6 feat: referral code bridging — capture ?ref= on signup/login, claim on callback 2026-03-22 19:09:56 -07:00
hanzo-dev 9b3b5c8238 security: gate localhost OAuth origins behind NODE_ENV !== production 2026-03-21 01:19:23 -07:00
hanzo-dev fbc3b1bf42 rebrand: 'Hanzo AI' → 'Hanzo' — Hanzo ID is universal, not AI-specific 2026-03-20 17:41:29 -07:00
hanzo-dev 77f8d6e17d fix: add permanentAvatar to UserInfo type — unbreaks CF Pages build
The TypeScript error on permanentAvatar caused the build to fail
silently (swallowed by || true), deploying zero functions and 404'ing
the entire site including /login and /account.

Also adds post-build verification to catch future silent build failures.
2026-03-11 18:38:22 -07:00
hanzo-dev fff4b8f38f security: fix critical OAuth redirect and cross-tenant vulnerabilities
- Add redirect URI allowlist validation to social-callback handler
  (was unvalidated open redirect — CRITICAL)
- Enforce org resolution from client map over untrusted state/cookie
  sources to prevent cross-tenant bypass (CRITICAL)
- Reject malformed/oversized state parameters
- Remove sensitive debug info from error redirects
- Add PKCE code_verifier forwarding in bridge handler
- Expand ALLOWED_ORIGINS to include billing, hanzo.id, lux/zoo/pars.id
2026-03-11 18:29:48 -07:00
hanzo-dev 22f1eb3dd1 fix(ci): use apiToken instead of CLOUDFLARE_API_KEY for wrangler
The wrangler-action needs apiToken input for API Token auth.
CLOUDFLARE_API_KEY env var is for Global API Key auth which
requires CLOUDFLARE_EMAIL — not what KMS provides.
2026-03-11 18:01:44 -07:00
hanzo-dev 171ba8cb18 fix: resolve user profile display on account page
- Use same-origin /oauth/userinfo proxy to avoid CORS
- Prefer id_token claims for displayName, email, avatar
- Store id_token in localStorage from both PKCE and passthrough flows
- Fetch /oauth/userinfo after password login for full profile
2026-03-11 17:40:13 -07:00
hanzo-dev 1cd915a4e9 fix(ci): use v3 KMS API with workspaceId and Universal Auth
- environment=production → environment=prod (matches KMS config)
- v1/secrets/raw → v3/secrets/raw with workspaceId
- Added KMS_CLIENT_ID/SECRET Universal Auth with HANZO_API_KEY fallback
2026-03-11 16:00:58 -07:00
hanzo-dev 7e8c9c4084 ci: migrate CF Pages deploy to HANZO_API_KEY + KMS 2026-03-11 14:50:51 -07:00
hanzo-dev 94d53d7f17 docs: add LLM.md project guide 2026-03-11 10:31:35 -07:00
hanzo-dev 5d44b6a246 fix(middleware): fall through to login UI when IAM doesn't auto-authorize
When proxying /oauth/authorize with a session cookie, IAM may return
200 (its built-in login page) instead of a 3xx redirect. This happens
when the session isn't valid for the requested application. In this
case, fall through to our own login UI instead of proxying IAM's page.
2026-03-10 20:18:46 -07:00
hanzo-dev 0d506cc23b fix(oauth): revert to direct code grant flow for PKCE login
The two-step approach (authenticate then redirect to /oauth/authorize)
doesn't work because IAM doesn't auto-authorize from session cookie.
Revert to the proven direct approach: POST to /api/login with
code_challenge as query param, get authorization code directly,
redirect to client callback with code+state.
2026-03-10 19:21:49 -07:00
hanzo-dev fafbf9fe9d fix(oauth): two-step login for proper PKCE support
Step 1: POST /api/login with type=login to authenticate (set session)
Step 2: Redirect to /oauth/authorize which middleware now proxies to IAM
when session cookie exists, producing a properly PKCE-bound auth code.

Previously, LoginForm tried to get the code directly from /api/login,
which didn't bind the code_challenge, causing PKCE verification failures
at the client's token exchange.
2026-03-10 19:11:09 -07:00
hanzo-dev cdb960ffa9 feat(i18n): expand language dropdown to 71 languages
Cover all major world languages including South/Southeast Asian,
African, Central Asian, Baltic, Nordic, and Caucasian languages.
2026-03-10 18:58:26 -07:00
hanzo-dev 6f1868e3cf fix(login): use middleware proxy instead of cross-origin IAM fetch
Root cause: LoginForm.tsx constructed absolute URLs to iam.hanzo.ai for
the /api/login POST, which fails due to CORS in browser. Now uses
relative URLs (/api/login) so requests go through the Next.js middleware
proxy.

Also:
- Add CLIENT_APP_MAP fallback for appName resolution (prevents sending
  raw clientId like "hanzo-console-client-id" when /api/get-app-login
  fails silently)
- Add LanguageDropdown component (wired globe icon to actual lang picker)
2026-03-10 17:36:01 -07:00
hanzo-dev b31552bb49 fix(branding): monochrome Hanzo theme + H logo mark
Replace red (#ef4444) accent with monochrome zinc-200 (#e4e4e7) across
all Hanzo domain branding configs. Update logo SVG from text-only
wordmark to geometric H mark + wordmark. Fix hardcoded red box-shadow
in input focus state.
2026-03-10 17:23:17 -07:00
Darkhorse7starsandhanzo-dev b29d16334c fix: extract app name from IAM response regardless of status
IAM's get-app-login API returns the app data (including name) even
when the redirect URI validation fails (status: "error"). Previously
we only extracted appName when status was "ok", causing the login to
fall back to using the raw clientId as the application name — which
IAM rejects with "does not exist".

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-03-10 17:18:31 -05:00
Darkhorse7starsandhanzo-dev 8a9485b7ba fix(ci): use Global API Key auth for Cloudflare Pages deploy
Switch from apiToken to CLOUDFLARE_API_KEY + CLOUDFLARE_EMAIL env vars
for wrangler-action, since we're using a Global API Key instead of a
scoped API Token.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-03-10 16:05:07 -05:00
Darkhorse7starsandhanzo-dev 5fc7cbe72d fix(ci): use npm packageManager for wrangler-action
The wrangler-action auto-detects pnpm from pnpm-lock.yaml and fails
when trying to install wrangler via pnpm. Force npm to avoid this.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-03-10 09:18:07 -05:00
hanzo-dev ed2445da6d feat: add multi-tenant login with org logos and IAM integration 2026-03-09 23:27:37 -07:00
hanzo-dev fa14c7e0ca fix: update Lux brand color from orange to zinc-200
Change lux.id primary color from #f97316 (orange) to #e4e4e7 (zinc-200)
with dark text (#09090b) for a clean, professional blockchain brand look.
2026-03-09 23:27:37 -07:00
Darkhorse7starsandhanzo-dev 66a21facf8 ci: add GitHub Actions workflow for Cloudflare Pages auto-deploy
Triggers on push to main and manual dispatch. Requires
CLOUDFLARE_API_TOKEN and CLOUDFLARE_ACCOUNT_ID secrets.

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-03-10 00:11:25 -05:00
Darkhorse7starsandhanzo-dev f6fa601c9d fix: resolve IAM app name from clientId and use camelCase query params
The login and signup forms were sending the OAuth client_id as the
`application` field to IAM's /api/login, but IAM expects the
application NAME (e.g. "app-hanzobot" not "hanzobot-client-id").
Additionally, IAM reads query params using camelCase keys (clientId,
responseType, redirectUri) but the forms were sending snake_case.

Changes:
- LoginForm/SignUpForm: resolve app name via get-app-login on mount
- Use camelCase query param keys matching IAM convention
- Include clientId/redirectUri in POST body as fallback
- Forward PKCE code_challenge params through login endpoint

Co-authored-by: Hanzo Dev <dev@hanzo.ai>
2026-03-09 23:59:21 -05:00
73 changed files with 7672 additions and 7689 deletions
+9
View File
@@ -0,0 +1,9 @@
<svg xmlns="http://www.w3.org/2000/svg" width="1280" height="640" viewBox="0 0 1280 640" role="img" aria-label="id">
<rect width="1280" height="640" fill="#0A0A0A"/>
<svg x="96" y="215" width="210" height="210" viewBox="0 0 67 67"><path d="M22.21 67V44.6369H0V67H22.21Z" fill="#fff"/><path d="M66.7038 22.3184H22.2534L0.0878906 44.6367H44.4634L66.7038 22.3184Z" fill="#fff"/><path d="M22.21 0H0V22.3184H22.21V0Z" fill="#fff"/><path d="M66.7198 0H44.5098V22.3184H66.7198V0Z" fill="#fff"/><path d="M66.7198 67V44.6369H44.5098V67H66.7198Z" fill="#fff"/></svg>
<text x="378" y="276" font-family="Inter,system-ui,-apple-system,sans-serif" font-size="78" font-weight="800" letter-spacing="-2" fill="#ffffff">id</text>
<text x="378" y="322" font-family="Inter,system-ui,sans-serif" font-size="30" fill="#ffffff" opacity=".66">Hosted login pages for Hanzo IAM - configurable per organization</text>
<rect x="378" y="338" width="806" height="3" rx="1.5" fill="#ffffff" opacity=".9"/>
<text x="378" y="390" font-family="Inter,system-ui,sans-serif" font-size="24" font-weight="600" fill="#ffffff" opacity=".5">github.com/hanzoai</text>
<text x="1184" y="390" text-anchor="end" font-family="Inter,system-ui,sans-serif" font-size="24" font-weight="600" fill="#ffffff" opacity=".5">hanzo.ai</text>
</svg>

After

Width:  |  Height:  |  Size: 1.3 KiB

-59
View File
@@ -1,59 +0,0 @@
name: Docker
# Self-contained build on Hanzo self-hosted runners. The shared reusable
# workflow (hanzoai/.github docker-build.yml@main) is currently failing graph
# validation for every caller (org-wide startup_failure), so this repo builds
# its own image directly. Cluster nodes are linux/amd64 → build that arch.
on:
workflow_dispatch:
push:
branches: [main, dev, test]
tags: ['v*']
permissions:
contents: read
packages: write
jobs:
docker:
# Target the ARC scale set by NAME (the org convention, cf. hanzoai/iam).
# A label array like [self-hosted, linux, amd64] matches the OFFLINE classic
# `evo-*` runners instead of the live ephemeral scale set, so every run
# queued forever. ARC scale-set jobs route on the exact scaleSetName.
runs-on: hanzo-build-linux-amd64
steps:
- uses: actions/checkout@v4
- name: Compute tags
id: tags
run: |
# The ARC runner image ships git + docker but NO node, so read the
# version from package.json with portable shell (sed), not `node -p`.
SHA="sha-$(git rev-parse --short HEAD)"
VER="$(sed -n 's/.*"version"[[:space:]]*:[[:space:]]*"\([^"]*\)".*/\1/p' package.json | head -1)"
if [ -z "$VER" ]; then echo "could not parse version from package.json" >&2; exit 1; fi
echo "sha=$SHA" >> "$GITHUB_OUTPUT"
echo "ver=$VER" >> "$GITHUB_OUTPUT"
echo "Tags: $SHA, $VER"
- uses: docker/setup-buildx-action@v3
- name: Login to GHCR
uses: docker/login-action@v3
with:
registry: ghcr.io
username: ${{ github.actor }}
password: ${{ secrets.GITHUB_TOKEN }}
- name: Build and push (linux/amd64)
uses: docker/build-push-action@v6
with:
context: .
platforms: linux/amd64
push: true
tags: |
ghcr.io/hanzoai/id:${{ steps.tags.outputs.sha }}
ghcr.io/hanzoai/id:${{ steps.tags.outputs.ver }}
cache-from: type=gha,scope=hanzoai-id
# mode=min (not max): the ARC runner builds via DinD on a 32G node;
# exporting every intermediate layer (mode=max) ballooned the DinD
# store past the kubelet eviction threshold and killed builds with
# "no space left". min caches only the final image layers.
cache-to: type=gha,scope=hanzoai-id,mode=min
-7
View File
@@ -1,7 +0,0 @@
name: Workflow Sanity
on:
pull_request:
paths: ['.github/workflows/**']
jobs:
sanity:
uses: hanzoai/.github/.github/workflows/workflow-sanity.yml@main
+50
View File
@@ -0,0 +1,50 @@
name: CI/CD
# The caller, and deliberately nothing else: triggers plus the import. All real
# config is the repo-root hanzo.yml, which platform.hanzo.ai reads too.
#
# `.hanzo/workflows`, NOT `.github/workflows`, and the difference is whether
# anything runs at all:
#
# * github.com has ZERO self-hosted runners for this org
# (/orgs/hanzoai/actions/runners -> total_count 0), so a job asking for
# `hanzo-build-linux-amd64` there is never claimed. An unclaimable job does
# not fail — it waits out the 24h timeout while the next push queues behind
# it. Silence, not an error.
# * The `git-runner` StatefulSet registers against the FORGE only
# (GIT_INSTANCE_URL=http://hanzo-git.hanzo.svc), and that is the pool which
# actually advertises that label.
# * Gitea collects workflows from the FIRST of WORKFLOW_DIRS present in the
# commit (modules/actions/workflows.go, listWorkflowsInDirs breaks on the
# first hit). `.hanzo/workflows` already existed here, so on this forge the
# whole of `.github/workflows` was already dark — including the Docker lane
# that used to live there.
#
# The `uses:` path points at .hanzo/workflows for the same reason: reusables
# resolve through services/actions.ResolveUses, which enforces the WORKFLOW_DIRS
# allowlist on the referenced path too. hanzoai/ci publishes build.yml at both
# paths from the same tag, byte-identical apart from the path each names for
# itself, so this is the same pipeline at the same @v1.
#
# NO `paths-ignore`. deploy.yml carried one so that a docs commit would not trip
# its "this version already exists" refusal — a guard this lane does not need,
# because it derives the next patch instead of going red. Dropping the filter
# means the GATES run on every commit to main, including the commits that only
# touch a workflow. A change to CI that breaks CI should be caught by CI.
on:
push:
branches: [main]
# A `v*` tag is a RELEASE and publishes an image named after it, verbatim.
tags: ['v*']
pull_request:
workflow_dispatch:
concurrency:
group: cicd-${{ github.ref }}
cancel-in-progress: true
jobs:
cicd:
uses: hanzoai/ci/.hanzo/workflows/build.yml@v1
secrets: inherit
+97
View File
@@ -0,0 +1,97 @@
name: Sync from GitHub
# git.hanzo.ai is CANONICAL and builds natively; development also lands on
# github.com/hanzo-inc/id. That repo MOVED — it was hanzoai/id, and it is now
# PRIVATE, which is why this job had been failing: GitHub answers 404 (not 403)
# for a private repo a token cannot see, so the old URL produced
# `remote: Repository not found` and the forge silently fell behind main.
# A member's SSH key reads it fine, which is exactly how that stayed invisible
# from a laptop. Together with the push-mirror going the other way
# (native -> GitHub, sync_on_commit) this is the full bidirectional loop.
#
# The two compose rather than fight: a native commit reaches GitHub via the
# push-mirror, so this job then sees LOCAL == REMOTE and exits "in sync". A
# GitHub commit fast-forwards native here, and the resulting push-mirror is a
# no-op because GitHub already has it. No echo, no loop.
#
# ONE deterministic direction per job: an in-cluster PULL. The runner reaches
# both ends (GitHub outbound, this forge via the instance URL actions/checkout
# already uses), so the sync has no ingress dependency.
#
# Fast-forward ONLY. A divergence fails LOUDLY here rather than force-pushing
# either side and destroying whichever history lost the race.
on:
schedule:
- cron: '*/10 * * * *'
workflow_dispatch: {}
concurrency:
group: sync-from-github
cancel-in-progress: false
jobs:
ff-main:
runs-on: [hanzo-build-linux-amd64]
steps:
- name: Checkout main (full history for the ancestry check)
uses: actions/checkout@v4
with:
fetch-depth: 0
persist-credentials: true
- name: Fast-forward main from github.com/hanzo-inc/id
env:
GH_PAT: ${{ secrets.GH_PAT }}
run: |
set -euo pipefail
git fetch --quiet "https://x-access-token:${GH_PAT}@github.com/hanzo-inc/id.git" main
LOCAL="$(git rev-parse HEAD)"
REMOTE="$(git rev-parse FETCH_HEAD)"
if [ "$LOCAL" = "$REMOTE" ]; then
echo "in sync at $LOCAL"
exit 0
fi
if git merge-base --is-ancestor "$LOCAL" "$REMOTE"; then
echo "fast-forwarding $LOCAL -> $REMOTE"
git push origin "$REMOTE:refs/heads/main"
# A push made with the workflow token does NOT trigger other workflows
# (loop prevention), so synced commits would never build. Dispatch it
# explicitly — a real fast-forward means real commits arrived.
#
# This names the workflow BY FILENAME, so it is a hard reference to a
# file in this repo and moves when that file does. It pointed at
# deploy.yml, which no longer exists; the `|| echo` below makes that a
# non-fatal 404, so the symptom would not have been a red sync — it
# would have been commits arriving on the forge and NOTHING building,
# silently, which is the exact failure this repo already survived once.
curl -fsS --max-time 20 -X POST \
-H "Authorization: token ${{ secrets.GITHUB_TOKEN }}" \
-H "Content-Type: application/json" \
"${{ github.server_url }}/v1/repos/${{ github.repository }}/actions/workflows/cicd.yml/dispatches" \
-d '{"ref":"main"}' || echo "build dispatch failed (non-fatal)"
elif git merge-base --is-ancestor "$REMOTE" "$LOCAL"; then
# Native is AHEAD of GitHub. This is the NORMAL state for the whole
# window between a native push and the push-mirror carrying it over,
# and that window is long: the mirror is sync_on_commit with an 8h
# floor, so it can legitimately last hours.
#
# It used to land in the `else` below, because that branch asked only
# "is LOCAL an ancestor of REMOTE" and called every other answer a
# divergence. So every native commit produced up to 48 consecutive
# DIVERGED failures, ten minutes apart, describing a repository that
# was in exactly the state this design intends. 11 of them were red on
# the hour I read them, none of them a real divergence.
#
# The cost is not the noise, it is what the noise hides: a job that is
# always red cannot report the one thing it exists to report. The loud
# failure below is worth keeping ONLY if it is rare.
#
# There is nothing to pull here and nothing to fix. Do NOT push native
# to GitHub from this job — that is the push-mirror's single direction,
# and duplicating it is how two writers start racing one destination.
echo "native is ahead at $LOCAL; GitHub at $REMOTE will follow via the push-mirror"
exit 0
else
# Genuine divergence: neither ref contains the other, so a commit
# exists on each side that the other has never seen. Nothing can be
# fast-forwarded and either direction would destroy history.
echo "DIVERGED: native $LOCAL and GitHub $REMOTE share no ancestry line." >&2
echo "Resolve by hand; this job will not force-push either side." >&2
exit 1
fi
+72 -4
View File
@@ -5,18 +5,86 @@ WORKDIR /build
ENV PNPM_HOME=/pnpm PATH=$PNPM_HOME:$PATH
RUN corepack enable && corepack prepare pnpm@10.15.0 --activate
COPY pnpm-workspace.yaml package.json tsconfig.base.json ./
# THE LOCKFILE SHIPS, and the install is frozen to it.
#
# This used to omit pnpm-lock.yaml and run `--frozen-lockfile=false`, so the
# image resolved the whole tree FRESH on every build while `pnpm test` on the
# runner resolved it from the lockfile. Two different dependency graphs from one
# commit: the tested one, and the shipped one. It went green for as long as free
# resolution happened to agree, and stopped the moment it did not — adding one
# dependency (@hanzo/event) moved vite from the lockfile's
# 7.3.5_@types+node@25.9.3_… to 7.3.6_@types+node@22.20.1 and the build died on
# `Cannot find module '/build/apps/web/node_modules/vite/bin/vite.js'`. Nothing
# was wrong with the source: the same commit builds cleanly when installed from
# the lockfile.
#
# A resolver free to drift ships a bundle no one has run. Frozen, the image gets
# the exact tree the tests passed against, and a lockfile that has gone stale
# fails HERE — loudly, naming the mismatch — instead of silently building
# something else.
#
# EVERY workspace member's package.json must be present before a frozen install:
# pnpm validates the lockfile against all of them and refuses if one is missing.
# apps/account is not built into this image, but it IS in the workspace, so its
# manifest is required for the check to pass.
COPY pnpm-workspace.yaml pnpm-lock.yaml package.json tsconfig.base.json ./
COPY apps/web/package.json apps/web/
COPY apps/account/package.json apps/account/
COPY pkgs/shared/package.json pkgs/shared/
COPY pkgs/auth/package.json pkgs/auth/
COPY pkgs/connect/package.json pkgs/connect/
COPY pkgs/idv/package.json pkgs/idv/
COPY pkgs/onboarding/package.json pkgs/onboarding/
RUN pnpm install --frozen-lockfile=false
RUN pnpm install --frozen-lockfile
COPY apps apps
COPY pkgs pkgs
RUN pnpm --filter @hanzo/id-web build
# Publishable event-ingest key (pk-live-…), inlined by Vite into the bundle.
#
# PUBLISHABLE_KEY is the name in KMS (org `hanzo`, path `deploy`, env `prod`)
# and on the --build-arg; the VITE_ prefix is what makes Vite inline it, and it
# is a property of THIS build, so it is applied here and the secret store keeps
# the ONE plain name.
#
# Publishable and write-only by design — it authorizes a write into one org and
# can read nothing — so shipping it in a bundle is the documented use. It is
# still a credential: it comes from KMS via CI. Never commit a value here.
#
# Deliberately NO default. An absent key is not a degraded mode: the door refuses
# the beacon outright. Measured against the live endpoint with and without a
# browser Origin, for pageviews AND exceptions alike: a hard 401
# `ingest_key_required`. Nothing arrives.
#
# This comment used to say the unkeyed beacon took an "anonymous lane" that filed
# rows under a `$public` tenant and answered 200. That lane is not implemented in
# the deployed cloud, and the belief was expensive: it is why "we still get
# errors, just not events" was accepted across four repos while the true answer
# was that a keyless surface reports nothing at all. hanzo.id ran exactly that
# way, which is the failure this gate exists to make loud.
ARG PUBLISHABLE_KEY
ENV VITE_PUBLISHABLE_KEY=$PUBLISHABLE_KEY
# Fail closed, and gate HERE because this is the one path every builder passes
# through — a guard in a workflow protects that lane only.
RUN case "$PUBLISHABLE_KEY" in \
pk-*) : ;; \
'') echo "PUBLISHABLE_KEY is empty - pass --build-arg PUBLISHABLE_KEY=<pk-...> (KMS deploy/PUBLISHABLE_KEY, env prod)" >&2; exit 1 ;; \
*) echo "PUBLISHABLE_KEY is not a publishable key (expected a pk- prefix)" >&2; exit 1 ;; \
esac
# Do NOT re-declare ARG VITE_PUBLISHABLE_KEY below this line. A later ARG of the
# same name shadows the ENV set above with an empty default, so the key resolves,
# passes the gate, and is then blanked before Vite inlines it — every step green,
# the bundle unattributed. That is exactly how hanzo.chat 1.0.58 shipped.
#
# `&&`, not `;`: with `;` the RUN exits with the status of the LAST command and a
# failed build would be masked. Assert on the bytes that actually ship — a key
# present in the environment and absent from the bundle is indistinguishable from
# success everywhere except the warehouse, where the traffic simply stops being
# attributable.
RUN pnpm --filter @hanzo/id-web build && \
{ grep -rqF "$VITE_PUBLISHABLE_KEY" apps/web/dist || \
{ echo "ERROR: the ingest key is not in apps/web/dist - hanzo.id would ship unattributed" >&2; exit 1; }; }
# SPA server stage — hanzoai/spa is the correct base for a Vite SPA:
# history-API fallthrough for client-side routes AND a SPA-safe CSP.
@@ -24,6 +92,6 @@ RUN pnpm --filter @hanzo/id-web build
# (built for static assets, not an app that loads its own bundle), which
# blocks the SPA's own scripts and leaves a blank page. hanzoai/spa serves
# index.html for all routes with a sane CSP. Defaults: PORT=3000, ROOT=/public.
FROM ghcr.io/hanzoai/spa:1.2.0
FROM ghcr.io/hanzoai/spa:1.4.11
COPY --from=build /build/apps/web/dist /public
EXPOSE 3000
+805 -30
View File
@@ -1,14 +1,516 @@
# LLM.md — Hanzo ID
## Mobile floors and a hover that stopped painting a wireframe (0.2.23)
A styling audit of iam.hanzo.ai. **The headline finding is a negative, and it is
the useful part: this surface does NOT have the "classes without rules" defect**
that broke hanzo.app, and structurally cannot. It uses neither Tamagui atomic nor
Tailwind — it is hand-written semantic `.hanzo-id-*` CSS. Measured in Chromium
against the served bundle, 7 routes x 3 viewports:
gui atomic (_bg- _dsp- _pos- _fs- _col- t_dark …) used 0 | rules 0
tailwind (flex items-center rounded-lg text-sm …) used 0 | rules 0
semantic .hanzo-id-* every rendered class resolves
The only classes with no rule are five page-variant markers
(`hanzo-id-login/-signup/-forgot/-onboarding-page/-callback`), each always paired
with `.hanzo-id-page`, which carries the layout. They are BEM-style hooks with
nothing to paint yet, not a defect — `.hanzo-id-device` and `.hanzo-id-portal`
are the same pattern with rules attached. Left alone deliberately.
**An audit claim that was wrong, corrected here so nobody re-fixes it.** It was
reported that a build with `@hanzo/design` absent "emits a bare unresolved
@import and every token vanishes with no error" — i.e. the hanzo.app failure
class, one `pnpm install` away. It does not. Removing the package and building
was tried: Vite's postcss-import **hard-fails**,
`[vite:css] [postcss] ENOENT: no such file or directory, open
'@hanzo/design/styles.css'`, exit 1. The token layer cannot silently disappear
from this build. What was actually true is narrower: the dev tree had not been
installed, so `tokens.test.ts` could not run. It passes once deps are present.
Four real defects fixed, all in `apps/web/src/app.css`, all verified by
before/after measurement of the BUILT bundle in a real browser (not by a green
build — a green build is what let the original defect ship):
- **Ghost-button hover painted a near-white wireframe.** It set
`border-color: var(--foreground)`#ededed, **17.9:1** on `--background` — on
the two most-hovered controls on the page (Continue with GitHub / Google).
Measured rest→hover: `rgb(115,115,115)``rgb(237,237,237)`. Now
`rgb(115,115,115)``rgb(115,115,115)`: the edge no longer moves and the
SURFACE carries the state (`--white-05``--white-10`), which is what the
`transition` on `.hanzo-id-btn` was already animating.
**Why not simply a dimmer border:** every rung brighter than `--neutral-500` is
worse on white. `--neutral-400` measures 8.3:1 on black but **2.52:1 on white**
and would fail the same WCAG 1.4.11 floor the resting border is documented to
hold in BOTH themes. No token gets brighter on dark and darker on light, so the
edge must not encode state at all. The resting `--border-strong` (4.43:1, both
themes) is untouched — that deviation from a low-alpha hairline is deliberate
and documented, and a control boundary still needs 3:1.
- **`viewport-fit=cover` was declared and never consumed.** Zero
`env(safe-area-inset-*)` rules in the whole stylesheet; `.hanzo-id-page` padded
a flat 24px, so content ran under the notch and home indicator. Now
`max(24px, env(...))` per side — 0 insets compute to exactly 24px, so nothing
moves on hardware without a notch (verified: page padding still `24px`).
- **The iOS-zoom comment described a protection the code did not implement.**
`font-size: var(--text-base) /* <16px makes iOS Safari zoom on focus */`
`--text-base` is 0.875rem = **14px**, so both credential fields zoomed on focus
and iOS never zooms back out. The scale has no 16px rung and should not grow
one (16 is Safari's threshold, not a design value), so it is a literal scoped to
`@media (pointer: coarse)`; the desktop ramp is untouched. Verified under iPhone
emulation: input 14px→16px, `iosWouldZoom` true→false.
- **Three tap targets under the 44px floor**, including a logo link whose target
was SMALLER than the logo inside it: an inline `<a>` takes its box from its own
line box, not from a replaced child, so it measured **32x18** around a 32px
mark. Fixed with `inline-flex` + a padding/negative-margin pair; the footer
links ("Forgot password?", "Create account", "Sign in", "Back to sign in") got
the same treatment — vertical padding on an inline box is hit-tested but does
not enter line-box height, so all of it is **zero layout shift**. Screenshots
before/after are pixel-identical at rest; only the targets and the hover moved.
Measured, BEFORE → AFTER, built bundle served like prod, 7 routes x 390/768/1280:
tap targets < 44px 3 → 0 (iPhone-emulated, pointer:coarse)
ghost hover border #ededed#737373 (17.9:1 → 4.43:1)
safe-area rules in CSSOM 0 → 1
@media rules 1 → 2 (added pointer:coarse)
input font-size (touch) 14px → 16px
horizontal overflow 0 → 0 (unchanged, all routes)
console errors / pageerrors 0 → 0
atomic + tailwind used|rules 0|0 → 0|0 (no regression)
CSS transferred 16,733 → 17,070 bytes (+337, +2.0%)
JS transferred 884,469 → 884,469 (unchanged)
**Still open, deliberately not taken here** (each is a finding against a layer
below this repo, and one is already recorded above under 0.2.15):
- `BrandHeader` loads the mark from `cdn.jsdelivr.net/npm/@<brand>/brand@latest/
assets/logo/logo.svg` — a third-party request pinned to a floating tag on the
credential-entry path. The `logoUrl` comes from `@hanzo/brand`'s own
`brand.json`, so it cannot be fixed here; the asset already exists locally at
`apps/web/public/brand/@hanzo/brand/assets/logo/logo.svg`. Note the catalog's
`brandUrl` is NOT fetched — `brandPackageFromUrl` only parses it for the npm
scope, and the brand JSON itself is same-origin `/brand/<slug>.json`.
- 939 KB of decoded JS to paint a two-field form; the eager bundle still carries
WalletConnect/viem that the email+GitHub+Google path never uses. Code splitting
works (ccip/login/secp256k1 chunks are correctly not fetched on `/`).
- The 70 KB Geist woff2 is discovered only after CSS parse, with no
`<link rel=preload>`; the filename is content-hashed, so a correct preload
needs `transformIndexHtml` to read the real name out of the bundle.
- `pkgs/shared/src/org.test.ts` "oauthCallbackOrigin is the org's hosted ID host"
FAILS on origin/main (`hanzo.app` vs `hanzo.chat`), pre-existing and unrelated
to this change — verified by running it on a pristine checkout. It belongs to
the in-flight callback work; not touched.
## The whole token layer, the ONE account control, and a gate on resolution (0.2.15)
0.2.14 adopted @hanzo/design and, in three places, worked around it. Those
findings have been fixed IN @hanzo/design 0.3.0, so the workarounds are gone and
this surface takes the system's answer.
- **ONE import: `@hanzo/design/styles.css`.** `app.css` cherry-picked four of the
nine token groups, so `--z-*`, `--shadow-*`, `--space-*`, `--font-*` and the
element defaults did not exist here at all. Nothing broke visibly, because an
unresolved `var()` paints nothing and reports no error — that silence is the
whole defect. @hanzo/iam's account menu alone reaches for `--z-popover`,
`--shadow-floating` and `--space-1..3`.
- **Geist is SELF-HOSTED in @hanzo/design 0.3.0**, so the reason for
cherry-picking is gone: `tokens/fonts.css` no longer requests
fonts.googleapis.com and the sign-in path can take the typeface with the
colours. Measured on the built bundle: rendered face is Geist, served from
`/assets/Geist-Variable-*.woff2`, zero third-party font requests.
- **The focus rule is DELETED from this file**, and it stays deleted.
`tokens/base.css` ships `:focus-visible{outline:2px solid var(--ring)}` to
every consumer; `--ring` is `var(--white-40)` in the 0.4.x line, measured
`2px solid rgba(255,255,255,.4)` = **3.77:1** on `--background` (WCAG 2.4.11
wants 3:1). Do not restore a local override — the last one existed only
because `--ring` was 1.66:1, which is fixed upstream.
Until **@hanzo/design 0.4.9** that layer carried a SECOND rule, a
field-specific `:where(input,select,textarea):focus-visible` that suppressed
the outline and drew a brightened edge + halo instead. Both rules computed to
(0,1,0) — `:where()` zeroes what it wraps — so the cascade fell through to
source order, the generic ring was written later, and it overrode the
`outline:none` the field rule stated to prevent it: every focused input here
drew BOTH indicators. Worse, the treatment it was trying to apply could never
have worked on this surface — its indicator rode on `border-color`, and
`.hanzo-id-input` states `border` unlayered, which beats a layer whatever its
specificity, so focused fields sat at the resting `.15` (1.47:1) while both
stylesheets read as correct. 0.4.9 deleted the field rule; one ring now covers
every focusable thing, and an `outline` is immune to the `border` this file
declares. `check-tokens.mjs` fails the build if a second focus rule, an
`outline:none` or a focus `box-shadow` ever returns.
- **Control boundaries are `--border-control`** (0.4.2 onward; this bullet read
`--border-strong` when 0.3.0 briefly cut the control rungs from the neutral
ladder). 0.4.2 put them back on alpha — `--border-control` .15, `--border-focus`
.22, `--border-selected` .30 — because an edge that clears 3:1 on a near-black
page is a mid-grey box, and a form of them reads as a wireframe rather than a
surface. The contrast budget moved to `--ring`, the focus indicator: the one
boundary a keyboard user actually navigates by, and the only one still pinned
at 3:1. `--border-strong` survives, but its duty is hover and emphasis — it is
decoration, and a control must not reach for it.
- **The signed-in portal mounts `<UserMenu>` from @hanzo/iam** (bumped
0.13.1 → 0.21.1 in `apps/web`, `pkgs/auth`, `pkgs/onboarding`) in place of a
hand-rolled "Billing / Sign out" link row. Identity comes from
`resolveIdentity`, the same resolution every Hanzo surface shows, so the portal
cannot disagree with the console about who you are. No `brand` prop is passed:
omitting `markSvg` would put the HANZO mark on lux.id and zoo.id.
- **Portal.tsx's inline `style={{…}}` is gone.** It carried
`rgba(255,255,255,0.14)`, `borderRadius 12`, `fontSize 13`, `fontWeight 600`
and two bare opacities — six invented values for facts the token layer already
states. `.hanzo-id-apps` / `.hanzo-id-applink` now, class-keyed like the rest.
- **`apps/web/src/tokens.test.ts` is the gate, and it tests RESOLUTION.**
"It is declared" was never evidence, and neither was "it type-checks": the
reference is built at runtime from a string, so it is invisible to the compiler
AND to grep. The test walks `app.css`'s `@import` graph into the installed
@hanzo/design, then asserts (1) every token group is served, (2) every
`var(--x)` under `src/` resolves, (3) every token @hanzo/iam's bundle paints
the menu with resolves. Verified by reintroducing both original defects: it
names the 5 dropped groups and the 8 unresolved iam tokens. `vitest.config.ts`
now includes `apps/**` so it actually runs.
Measured in Chromium, fresh context, empty storage, 1440 AND 390, against the
built bundle served as hanzo.id / lux.id / zoolabs.id: **0 unresolved tokens of
76 referenced** on every host; input edge and focus ring both 4.43:1; the account
menu paints `#0a0a0a` fill / 12px radius / `--shadow-floating` / `z-index 700`
(it was transparent before the iam fix). Brand switches with zero per-brand code:
hanzo.id → "Hanzo ID" + Hanzo apps, lux.id → "Lux ID" + Lux apps, and no Hanzo
mark reaches the Lux menu.
STILL OPEN, and they are findings against layers below this surface, not against
this repo: the menu's own panel edge is `--border` at 1.27:1 on `--popover` —
0.4.2 recut the control and focus rungs but left `--border` on the .10 rung, so a
floating panel still has no perceivable edge. And `BrandHeader` loads the brand
mark from `cdn.jsdelivr.net/npm/@<brand>/brand@latest/...` — a third-party
request pinned to `@latest` on the sign-in path, which is exactly what this file
refuses for fonts.
## One token layer, and no control painted by its ancestor (0.2.14)
The portal is now styled from **@hanzo/design tokens** — the same token layer
hanzoai/pay renders from, so the two halves of one flow (sign in → pay) agree on
the page black, the type ramp, the radii and the greys. `apps/web/src/app.css`
invents no colour, radius or size; `@hanzo/gui` was declared as a dependency but
never imported once, and is removed rather than left as decoration.
- **Surface is keyed to a CLASS the component carries, never to where it is
mounted.** The stylesheet used to paint controls with the descendant selectors
`form input {…}` and `.hanzo-id-btn, form button {…}`. `DeviceApproval.tsx` —
the screen a human hits to authorise the CLI — has 2 inputs and **0** `<form>`
ancestors, so its device-code field fell out of the stylesheet and rendered as
raw UA chrome: 31px tall, `#3b3b3b`, `2px inset` bevel, square corners, beside
correctly-styled 44px siblings. Every control now carries `.hanzo-id-input`,
`.hanzo-id-btn`, `.hanzo-id-field` or `.hanzo-id-form`, and there is not one
element-descendant selector for surface left in the file. Same class of defect
as a component library shipping utility class names with no CSS behind them —
the mirror image, bare elements instead of bare class names.
- **`.hanzo-id-spinner` had no rule at all.** The loading state measured 0px and
was invisible on hanzo.id, lux.id and pars.id at once. It is a real 28px ring
now, verified rendering and animating.
- **ONE focus indicator.** The file had exactly one focus rule (`form
input:focus`), so every button, link and social entry fell back to Chrome's
`outline: auto` — a blue ring on a monochrome surface. `:focus-visible` is now
global, 2px `--primary`. Note `--ring` (#333333) measures **1.66:1** on
`--background` and cannot carry a focus indicator; that is a finding against
@hanzo/design, not a licence to invent a value.
- **ONE button.** `.hanzo-id-social-btn` and the `.primary` modifier are gone:
`.hanzo-id-btn` IS primary, `.ghost` is the secondary surface (social sign-in,
Skip/Back). 13 treatments → 1 primitive with 1 modifier.
- **`font: inherit` on every control.** Buttons and inputs rendered in Arial
while headings rendered in the platform face — two typefaces in one 432px card,
including on the "Sign in" CTA.
- **Control borders are `--white-40`, deliberately.** On `--background` the
semantic `--border` (#1f1f1f) measures 1.27:1 and `--border-strong` (#404040)
2.03:1 — neither clears the 3:1 a control boundary needs (WCAG 1.4.11).
`--white-40` measures 3.66:1 and is on the ladder.
- **Fonts are NOT imported from the design package.** `tokens/fonts.css` pulls
Geist from fonts.googleapis.com; the sign-in path loads no third-party font.
`--font-sans` resolves to the platform stack, the identical value hanzoai/pay
sets. Self-hosting Geist would let both import `fonts.css` unchanged.
Measured in a real browser against the built bundle (fresh context, empty
storage): page `#000000`, controls 44px, one 6px radius, h1 21px, white focus
ring, no control under the touch floor at 390px, no horizontal overflow.
**This release also reunites `main` with production.** The running image (0.2.13,
built from `9477777`) was NOT on `origin/main`: the two had diverged at
`7a72225f`, with the self-service-signup fix (`/v1/iam/onboard` instead of the
admin-only `add-organization` verb) and the App/Chat/Cloud launcher live but
unmerged. A release cut from `main` would have silently regressed both. 0.2.14 is
the merge, so `main` is once again what runs.
## One chain-agnostic "Connect Wallet" button — merge EVM + Solana entries (0.2.9)
The login page rendered ONE wallet button PER enabled chain (`SocialButtons`
mapped `ENABLED_WALLET_CHAINS` → "Continue with Ethereum / EVM" +
"Continue with Solana"), so a two-chain build showed two near-identical buttons
above the divider. Merged into a SINGLE chain-agnostic "Connect Wallet" entry —
the ENTRY is merged, both underlying flows are kept intact. (Rides on 0.2.8:
GitLab provider + vitest-unified test runner.)
- **Detection is a pure function, in the ONE web3 module.** `detectWalletChains()`
(`pkgs/auth/src/web3.ts`) is a pure `window` sniff — EVM = `window.ethereum`,
Solana = `window.solana`/`solflare`/`backpack` — that returns the
`ENABLED_WALLET_CHAINS` with an injected provider. Derived from the enabled
set (one source of truth); testable via an injectable window (no DOM, no
connect, no I/O). Exported alongside `loginWithWalletChain`.
- **`SocialButtons` renders one entry.** The web3 provider now renders a single
"Connect Wallet" button (`data-wallet-connect`). On click, `onConnectWallet`
calls `detectWalletChains()`: exactly one injected chain → connect straight
(`startWallet(chain)`, no chooser); zero or many → reveal an inline chooser
(`.hanzo-id-wallet-chains`) of one button per enabled chain
(`data-chain=evm|solana`) so EITHER chain stays reachable. Same monochrome
`hanzo-id-social-btn` style as GitHub/GitLab/Google; the chooser is indented
under the entry. The per-chain path (`startWallet` → `loginWithWalletChain`)
is UNCHANGED — only the button that reaches it is merged.
- **Verified.** Auth unit tests green (3 new `detectWalletChains` cases:
single→[chain], both→[evm,solana], none→[]). Playwright against the dev SPA
(get-app-login intercepted so web3 resolves): one "Connect Wallet" renders;
click with no injected wallet → chooser shows Ethereum/EVM + Solana; click
with an injected `window.ethereum` → NO chooser, straight into the EVM flow.
- **Backend untouched.** Pure `id` SPA change (frontend), no IAM edit.
Ships as `ghcr.io/hanzoai/id:0.2.9`; deploy = bump the operator CR image
(`universe/infra/k8s/operator/crs/id.yaml`) by hand (id not in the
gitops-reconcile allowlist). NEVER restart ingress (TLS-outage hazard).
## The tenant catalog was never read — one key name, every host (0.2.21)
`/config.json` is served by `hanzoai/spa`, which derives its JSON key
**mechanically from the env var we supply**: `SPA_IAM_TENANT_CONFIG_JSON`
(ConfigMap `id-tenant-catalog`) becomes `iamTenantConfigJson`. The
`TenantConfig`→`OrgConfig` rename swept through this codebase and renamed the
READ too — `cfg.iamOrgConfigJson`, a key nothing emits. The fetch returned 200,
the field was `undefined`, `resolveOrg` fell back, and nothing logged. Verified
live: `https://hanzo.id/config.json` returns `iamTenantConfigJson`, and the SPA
was reading straight past it.
The env var is the interface and it is NOT ours to rename unilaterally — the
ConfigMap, the Deployment's `envFrom` and the spa image all speak it. Code
follows the interface, so the read moved to `catalogOf` in `pkgs/shared/src/
org.ts`, next to the resolver it feeds, pinned by tests that fail on the old
spelling.
Two live consequences, and the first is why this shipped with 0.2.20:
- **hanzo.id authenticated as the wrong app.** The catalog maps it to
`hanzo-console` (`enableSignUp: true`); with no catalog it fell to the built-in
`hanzo-id` (**false**). Federation PROVISIONS a local user for a new identity,
so a first-time GitHub user was refused "the application does not allow to sign
up new account" — the 0.2.20 fix working perfectly and still failing the exact
people it was for. `enableSignUp` itself needed no change; reading the catalog
did.
- **Every catalog-ONLY host resolved nothing.** osage.id, zoolabs.id,
id.zoo.network, id.lux.network, iam.lux.network, id.pars.network, id.bootno.de
and iam.hanzo.ai have no built-in entry, so they got `hostSkeleton` — empty
`clientId`, no application. That the resolver fails CLOSED there is why it leaked
no brand (the osage.id regression stayed fixed); it is also why it was silent.
Declared state was corrected to match: `hanzo-console` now lists
`https://hanzo.id/callback` in `init_data.json`. The live row already had it, and
`seed.go` deliberately keeps `redirectUris` off `appPolicyKeys` ("the registration
surface, which drifts legitimately and is owned elsewhere") so nothing reverts —
but redirectUris ARE applied on CREATE, and from 0.2.21 hanzo.id authenticates as
`hanzo-console` against that exact callback. Without the line, a REBUILT cluster
would come up unable to sign anyone in at hanzo.id.
The shape to remember: a rename is not finished at the language boundary. This
one crossed into an env var, a ConfigMap key and a base image's templating
convention, none of which the compiler or the type system can see — and the
failure mode was a successful fetch of a field that wasn't there.
**Follow-up, deliberately not taken here: `oauthCallbackOrigin` is now dead.**
Nothing reads it. It existed so the browser-built hop could target the shared
`iam.hanzo.ai/callback` OAuth client; with federation, IAM pins its own callback
from the TRUSTED request host (`federationBaseURL` → `resolveIssuer(c.Host())`),
so the browser never chooses a callback origin. What survives is the field in
`pkgs/shared/src/types.ts`, its passthrough in `org.ts`, two test fixtures, and
an entry on every catalog row in `universe/infra/k8s/id/configmap.yaml`. Its doc
comment still states the DELETED rule — "the provider hop MUST send
`redirect_uri=<oauthCallbackOrigin>/callback`" — which is exactly the sentence
that would talk the next reader into rebuilding the hop. Delete the field with
the next code change to this package; it was left alone today only because
touching it cuts another release of a live auth surface for zero behaviour
change, and the build refuses a commit that does not bump the version.
## Social sign-in goes through IAM, because it always did (0.2.20)
GitHub sign-in reached GitHub, succeeded there, and then dead-ended: the user
landed back on hanzo.id and was told to sign in first. That message is IAM's
`please sign in first` — a 401 from `/v1/iam/onboard`, reached with no session
and no bearer. The SPA had never obtained a token.
**The SPA was the relying party, and it cannot be one.** `social.ts` built the
IdP URL in the browser (`github.com/login/oauth/authorize`, `redirect_uri=
${origin}/callback`) against a contract copied from an IAM fork whose front end
no longer exists. GitHub then returned a GitHub code to the SPA's own
`/callback` — and **nothing can spend that code**: exchanging it needs the client
SECRET, which a browser must never hold, and IAM has no endpoint that takes a raw
provider code. `Callback.tsx` POSTed it to `/v1/iam/login`, which does password,
device approval and code minting, and knows nothing about providers. So every
social sign-in ended authenticated at GitHub and anonymous here.
**IAM already implements the whole flow and was never called.**
`internal/oidc/federation.go` is a complete OAuth2/OIDC relying party. Naming a
`provider` on the authorize endpoint IS the entry point: `authorizeHandler`
validates client_id, the EXACT redirect_uri and the PKCE policy, then
`beginFederation` resolves the provider, mints a single-use transaction, sets a
browser-binding cookie and sends the browser to the IdP. The IdP returns to
**IAM's** fixed callback, `/v1/iam/oauth/callback`, where IAM — holding the
secret — exchanges the code, links or provisions the user, and mints an IAM
authorization code bound to the original PKCE challenge, redirect_uri and nonce.
`OAuthAuthorizeRequest.provider` had been declared in the SPA's own types the
whole time; `authorize()` simply never emitted it. The fix is that one parameter,
plus deleting everything that existed to work around its absence.
**`provider` is the RECORD name — `provider-github`, never `github`.**
`federationProvider` matches `ProviderItem.Name` exactly, and `EnrichProviders`
resolves that same name to the record, so the two are one string by construction.
Verified live: `?provider=github` → `invalid_request: unknown or unavailable
provider`; `?provider=provider-github` → 302 to GitHub. A comment on
`providerKey` used to assert the opposite; it is now corrected in place. The bare
key is a DISPLAY key (icon, label, `provider_hint` matching) and nothing else.
**Two arms, and they are the two the password path already had.** The question is
only who owns the PKCE verifier (`SocialButtons.hop`):
- **An app sent the user here** (`redirect_uri` on the query) → re-enter authorize
with THAT app's request via `social.ts::authorizeRequest`, so IAM mints the code
against its client_id, redirect_uri and challenge and returns the browser
straight to it. The app holds the verifier; this portal is never in the return
path and never touches a token. Same branch as `Login.completeAfterAuth`.
- **A bare portal sign-in** → `createIam(...).signinRedirect({additionalParams:
{provider}})`. The SDK generates and persists the verifier in **localStorage**
(`txStorage`, keyed `hanzo_iam_code_verifier:<state>` — localStorage, not
session, deliberately, so it survives the full-page redirect), and
`handleCallback` reads back that exact slot. IAM returns the APP's state
(`federationMint` sets `state` from `AppState`, not its IdP-leg state), so the
SDK's state check matches.
Deleted, all of it dead once the browser stops being the relying party: the IdP
endpoint/scope table, `buildProviderAuthUrl`, `startProviderLogin`,
`isHoppableProvider`, `encodeState`/`decodeState`, `client.providerLogin`,
`ProviderExchangeRequest`, and `Callback.tsx`'s provider branch. `/callback` now
has ONE case — an ordinary IAM code — because a federated return is
indistinguishable from any other.
Verified in a real browser against live IAM (local bundle, `/config.json`
pointing `localhost` at `https://hanzo.id`; port 5173 because
`http://localhost:5173/callback` is registered on `hanzo-id`):
- bare-portal click → `hanzo.id/v1/iam/oauth/authorize?…&provider=provider-github`
→ 302 → GitHub, with `redirect_uri=https://hanzo.id/v1/iam/oauth/callback` and
the `hanzo_fed` cookie (`SameSite=Lax`, `path=/v1/iam/oauth/callback`, 600s);
a verifier slot appears keyed by state.
- app-initiated click → same, and NO verifier slot is created — proof the app's
own request was forwarded rather than the portal starting its own flow.
- return leg → `/callback?code=…&state=<stored>` consumes the slot and POSTs
`/v1/iam/oauth/token`, which answers `invalid_grant: invalid authorization
code` for a deliberately fake code. The exchange is wired; only the code was
false.
Not exercisable here: the GitHub login itself. See the registration note under
"Social providers" below — GitHub defers redirect_uri validation until after
sign-in, so an unauthenticated probe CANNOT tell a registered callback from an
unregistered one (a deliberately bogus URL returns the identical 302).
## Provider-hint auto-federation — click GitHub/Google downstream, land straight in the provider (0.2.6)
Clicking "Continue with GitHub/Google" on a downstream app (console.hanzo.ai)
used to bounce the user to the hanzo.id login FORM — the portal ignored the
provider the user already chose. Now it launches that provider immediately.
Three fixes. Points 1 and 2 still hold; point 3 and every mention of the "hop"
below are **superseded by 0.2.20** — the hop is gone, the auto-launch now starts
IAM's federation like the button does, and `method` is not a parameter of it
(IAM's federation callback links-or-provisions on its own).
1. **`Login.tsx` honors `provider_hint`.** The console SDK already appends
`&provider_hint=provider-github` (the IAM record `name`) to the authorize
redirect. A new `federate` phase: after a silent-SSO miss, if the hint is
present it renders a HEADLESS `<SocialButtons autoStart={hint}>` that
auto-runs the same hop the button runs — no form flash — and drops to the
form only if the hint matches no configured provider. Honor ONLY
`provider_hint`, never bare `provider=` (that carries the SSO SDK's
`<org>-iam` IDP hint — see the single-provider-state note below).
`matchProviderHint` (`social.ts`, pure + tested) maps the hint to a provider
by record name / key (`provider-github` or `github`, case-insensitive).
2. **`getAppLogin` validates against the DOWNSTREAM app's own redirect_uri.**
`getAppLogin(clientId, redirectUri?)` now sends the incoming OIDC
`redirect_uri` (read from the query), not the hardcoded
`${publicOrigin}/callback`. IAM validates it against the app's registered
list: `hanzo-cloud` (the console's client_id) registers
`console.hanzo.ai/auth/callback` but NOT `hanzo.id/callback`, so the old
hardcode made `CheckOAuthLogin` answer `status:error` ("Redirect URI …
doesn't exist in the allowed list") and the SPA dropped the whole response
(`status!=='ok'`→null) → NO providers resolved. This blocked the social
buttons for EVERY cross-app SSO read, not just auto-federation. `SocialButtons`
reads `redirect_uri` from `window.location.search` (same pattern as the wallet
path); absent (bare portal / device flow) it defaults to the portal callback.
3. **The interactive hop uses `method=signup` (find-or-create-login).** IAM runs
find-or-create only under `signup` (`controllers/auth.go:1041`: existing
3rd-party identity → sign in, else create). `signin` is the account-LINK
branch (`auth.go:1257`: `GetSessionUsername()==''` → `ResponseError("user
doesn't exist")`) — it needs a live session and errors on a fresh "Continue
with GitHub". The old `intent==='signup'?'signup':'signin'` sent `signin` from
the sign-in page; both intents now use `signup` (intent only changes button
copy). Latent bug — never exercised end-to-end before this.
Contracts locked in `pkgs/auth/src/{social,client}.test.ts`. Deploy: image
`ghcr.io/hanzoai/id:0.2.6`, operator CR `universe/infra/k8s/operator/crs/id.yaml`
(the `id` CR did not exist in-cluster before — id ran off the raw
`deployment.yaml`; applying the CR handed the Deployment to the operator, safe
because id carries no env, only `envFrom: id-tenant-catalog`). `id` is NOT in the
gitops-reconcile allowlist → apply the CR by hand. Rides on 0.2.5 forced-MFA.
## Silent SSO must be org-scoped — admin-guard god-mode fix (fixed 0.2.2 → 0.2.3)
`admin.hanzo.ai` (global-admin console) sits behind admin-guard, a Traefik
ForwardAuth that allows ONLY `owner == admin` tokens. Login rides
`client_id=hanzo-admin-guard` (org=admin). The 0.2.2 credential-form fix
(`LoginForm` posts `app.organization` from `get-app-login` → org=admin →
IAM resolves `admin/z`, owner=admin) is CORRECT and verified: posting
`organization=admin` to `/v1/iam/login` resolves the admin/* row and
`get-account` returns owner=admin.
But it "didn't take effect end-to-end" because `Login.tsx` (0.2.2) added a
`silentLogin` SSO fast-path (`canSilent = client_id && redirect_uri`) that,
on mount, minted an auth code from the AMBIENT `iam_session_id` session
REGARDLESS of that session's org. Real operators carry a hanzo/* session
(from hanzo.chat/console), so silentLogin minted an owner=hanzo code and
the guard bounced them to console.hanzo.ai — the org-scoped form was never
shown. Silent SSO shadowed the form fix.
Fix (`pkgs/auth/src/client.ts`, `silentLogin`): silent SSO may reuse the
ambient session ONLY when its user org == the app's org. `silentLogin` now
resolves the app org (`get-app-login`) + session owner (`get-account`,
new internal `sessionOwner()` helper) before minting; on no session or an
org mismatch it returns `{}` so `Login.tsx` falls back to the interactive
form (which authenticates in the app's own org). Same-org SSO (the common
case: hanzo session → hanzo app) still mints silently — no UX change.
Cross-org (hanzo session → admin-guard) → form → org=admin → owner=admin →
god-mode. Non-admins (e.g. Dave) never reach god-mode: the guard validates
`owner==admin` server-side, so this is availability (admins get IN), not a
privilege boundary — the fix is client-side and cannot admit a non-admin.
IAM (`iam:v1.31.14`) is unchanged; the SSO-ATO exact-match redirect fix
(d7648965) is untouched.
## Social login (GitHub/Google) — single-provider state + matched redirect_uri (fixed 0.1.24 → 0.1.25)
> **Superseded by 0.2.20.** Bug A (read the provider identity from the NESTED
> record) still holds and still ships. Everything below about the base64 `state`,
> `buildProviderAuthUrl` and `providerLogin` describes the browser-side IdP hop,
> which is gone — the browser no longer builds an IdP URL or handles a provider
> code, so neither bug can recur. Kept because it records how the flow was
> misdiagnosed twice: each fix made the SPA a slightly better relying party, when
> the SPA could never be one at all.
The social hop used to fail at the IAM `/callback` exchange — GitHub with
**"The provider: hanzo-iam does not exist"**, Google with "password or code is
incorrect". TWO independent bugs, both in the SPA's provider-name handling; the
IAM backend, the OAuth creds, and the registered redirect_uri were all fine.
**Bug A — wrong provider IDENTITY (fixed 0.1.24).** `parseAppLogin`
(`pkgs/auth/src/client.ts`) read the **outer** Casdoor app-provider LINK `name`
(`pkgs/auth/src/client.ts`) read the **outer** IAM app-provider LINK `name`
as the provider identity. `get-app-login` returns each provider as a link object
`{name, canSignIn, …, provider:{name, type, clientId, …}}`; the REAL identity is
the nested `provider.name` (e.g. `provider-github`), the name the backend
@@ -115,7 +617,38 @@ with `defaultProviders: [provider-github, provider-google, provider-web3,
provider-apple]` (or any subset) and leave each app's `providers: []`. All apps
inherit automatically — no per-app reconfiguration.
## Org-agnostic password login (fixed 0.1.23)
## Org-agnostic password login (0.1.23) — SUPERSEDED, do not restore
**This section describes behaviour that no longer works and must not be
reinstated.** It is kept because the reasoning below is what makes the current
design legible, and because someone reading only this section will otherwise
"fix" the apex login straight back into an outage.
iam2 removed cross-org resolution deliberately. It scopes every credential
lookup to one org and treats the collision this design leaned on as a defect —
`internal/registry/registry.go` names it "the F-2 bug where z@hanzo.ai collided
across admin and hanzo", because resolving across orgs coupled lockout counters
between rows and handed out a brute-force oracle on the SuperAdmin. An org-less
`POST /v1/iam/login` is now refused outright:
HTTP 200 {"status":"error","msg":"organization, username and password are required"}
Note the **200**. The form renders that as though the user's own password were
wrong, and every status-code monitor reads it as healthy. Left unfixed it killed
the bare sign-in on hanzo.id, lux.id, iam.hanzo.ai and pars.id at once while
looking green.
So `LoginForm` now resolves the app's own org via `get-app-login` and posts it on
BOTH entry points — the same thing the 0.2.2 fix below already established for
the downstream-app path. A global admin is no longer reached by omission; they
reach admin/* by signing into an admin-org app (e.g. `hanzo-admin-guard`), which
is the explicit path 0.2.2/0.2.3 describe. `client.login()` stays a pure
passthrough: it never invents an org, it only forwards one.
The original 0.1.23 note follows, for context only.
### (superseded) original text
The portal login is now **org-agnostic**: it no longer pins
`organization=<brand>` on `POST /v1/iam/login`. `LoginForm` passes
@@ -204,7 +737,7 @@ pars.id ──┘ │
per-brand OIDC issuer host: hanzo.id / lux.id /
zoo.id / pars.id (serves /.well-known + /v1/iam/*;
same Casdoor-fork backend, tenant-scoped by org)
same Hanzo IAM backend, tenant-scoped by org)
iam-* postgres in hanzo namespace
@@ -223,35 +756,54 @@ mirrors each `-id` app's provider config in
REST `login` and returns an auth code directly. Both honor a downstream
`redirect_uri`.
### Social providers — render only when configured; redirect via the "hop"
### Social providers — render only when configured; federate through IAM
`SocialButtons` renders ONLY providers IAM holds a REAL credential for
(`AppProvider.configured` = non-placeholder clientId). With the seed's
placeholders every social button is hidden, so a user never hits a dead-end;
they reappear automatically once real creds land. Clicking a configured OAuth
provider runs the **hop** (`social.ts::startProviderLogin`), which redirects
straight to the provider with a base64 `state` that round-trips the original
authorize request — matching the IAM (Casdoor) `getAuthUrl` contract. The
provider returns to `/callback`; `Callback.tsx` detects the provider state and
calls `client.providerLogin` to exchange the code at the IAM backend, then
follows the continue-URL (which re-enters `/callback` as the normal OIDC code).
(NOT `@hanzo/iam` `signinRedirect` — that loops back to the login page.)
(`AppProvider.configured` = non-placeholder clientId), so a placeholder-seeded
provider is hidden rather than dead-ending; it reappears once real creds land.
IAM applies the same rule server-side (`isConfigured`), so a hidden provider is
also refused at authorize — the two agree without sharing a table. Clicking a
configured provider names it on IAM's authorize endpoint and IAM runs the entire
IdP leg; see "Social sign-in goes through IAM" above. This browser never builds
an IdP URL, never sees a provider code, and holds no secret.
**To ENABLE real social login (the only remaining work):**
1. Register an OAuth app per provider (GitHub/Google) with callback
**`https://<brand>/v1/iam/callback`** AND the app authorize redirect
`https://<brand>/callback` (per brand host: hanzo.id, lux.id, pars.id …).
2. Put the client id/secret in KMS at **project `hanzo-iam`, env `prod`**, keys
`IAM_GITHUB_CLIENT_ID` / `IAM_GITHUB_CLIENT_SECRET` (and `IAM_GOOGLE_*`). The
`iam-kms-sync` KMSSecret (`universe/infra/k8s/iam/secret.yaml`) syncs that
path into `iam-secrets`; init_data.json substitutes `${IAM_GITHUB_CLIENT_ID}`
at deploy. The whole sync + env-ref chain already exists — today those keys
just hold placeholder values, so providers read as unconfigured (buttons
hidden). Replace the values; nothing else to wire.
3. The buttons appear automatically (no portal change). **Live-verify** the
round-trip reaches the provider and completes — the hop + exchange are wired
and unit-tested (`pkgs/auth/src/social.test.ts`) but can only be exercised
end-to-end once real creds exist.
**The callback the IdP must have registered is IAM's, not the SPA's.** One fixed
path, every provider, per brand host:
https://<brand>/v1/iam/oauth/callback ← register THIS at GitHub/Google
(`PathFederationCallback`; the origin is the brand's pinned issuer, resolved from
the trusted request host — `hanzo.id` federates to `hanzo.id/v1/iam/oauth/
callback`, lux.id to its own, and so on.) It is NOT `/v1/iam/callback` (that path
does not exist — an earlier revision of this file said so and was wrong) and NOT
`https://<brand>/callback`, which is the APP's authorize redirect and belongs to
the app's registered `redirectUris` in `init_data.json`, a different list for a
different leg.
Beware verifying this from outside: **GitHub defers redirect_uri validation until
after the user signs in**, so an unauthenticated request to
`github.com/login/oauth/authorize` 302s to the login page whether the callback is
registered or not — a deliberately unregistered URL behaves identically. The only
sound check is the GitHub App's own settings page.
Credentials live in KMS at project `hanzo-iam`, env `prod`, keys
`IAM_GITHUB_CLIENT_ID` / `IAM_GITHUB_CLIENT_SECRET` (and `IAM_GOOGLE_*`). The
`iam-kms-sync` KMSSecret (`universe/infra/k8s/iam/secret.yaml`) syncs that path
into `iam-secrets`; `init_data.json` substitutes `${IAM_GITHUB_CLIENT_ID}` at
deploy. That chain already works — GitHub and Google both carry real values
today; `provider-web3` and `provider-apple` are still placeholders and stay
hidden.
**`enableSignUp` gates a FIRST-TIME federated user.** Federation provisions a
local user when the identity is new, so an app with `enableSignUp:false` refuses
a first-time GitHub user with "the application does not allow to sign up new
account" — a sign-in failure that has nothing to do with the flow being wired.
Live: `hanzo-console` and `hanzo-app` are true; `hanzo-id` and `hanzo-cloud` are
false. `hanzo.id` sends `clientId: hanzo-console` (`universe/infra/k8s/id/
configmap.yaml`), so the portal is on a signup-permitting app. It is governed
declaratively in `universe/infra/k8s/iam/init_data.json` and reconciled every
boot (`iam/internal/seed/seed.go`, `appPolicyKeys`) — change it THERE, never by
an admin call, which the next boot would revert.
## Workspace
@@ -352,7 +904,7 @@ Custom providers: implement the `IDVProvider` interface in
## Backend
The Go IAM backend lives at `~/work/hanzo/iam` (Casdoor fork, module
The Go IAM backend lives at `~/work/hanzo/iam` (Hanzo IAM, module
`github.com/hanzoai/iam`, image `ghcr.io/hanzoai/iam`). All paths are under
the `/v1/iam` prefix — no legacy `/oauth/*`, no `/api/`. This portal talks
to it via:
@@ -372,3 +924,226 @@ to it via:
All hostnames talk to the same IAM backend — the org is carried in the
request body (`organization: <orgId>`), and the IAM backend tenant-scopes
on that.
## Truth flows git.hanzo.ai -> GitHub (2026-07-26)
`hanzoai/id` is CANONICAL on git.hanzo.ai (`mirror:false`, default `main`).
GitHub is a **push-mirror** of it, `sync_on_commit: true` with an 8h floor —
push here and GitHub follows on its own.
It used to be the reverse, and that was the bug: as a pull-mirror this repo
could run no CI at all, so `.hanzo/workflows` never fired and four commits
shipped zero images without a single red signal. Do not re-point the sync.
Builds publish to BOTH `oci.hanzo.ai/id` (ours, the destination) and
`ghcr.io/hanzoai/id` (kept so a rollback target always resolves), tagged from
`package.json` — a release IS a version bump. Deploying is a `spec.image.tag`
edit in `universe/infra/k8s/operator/crs/id.yaml`; CI must never patch that CR
itself, because Hanzo CD reverts it within ~90s.
---
## HANDOFF — 2026-08-03: social login, SSO, and the build path
State at handoff. Everything below was measured against production, not inferred.
Where a claim is unverified it says so.
### THE ONE-LINE ROOT CAUSE OF "DAYS OF LOGIN BUGS"
`pkgs/shared/src/org.ts` — `oauthCallbackOrigin` defaulted to `publicOrigin`, the
BRAND'S OWN host, and no catalog entry overrode it. So every property sent a
different `redirect_uri` to Google/GitHub:
hanzo.app -> hanzo.app/callback hanzo.chat -> hanzo.chat/callback
console -> console.hanzo.ai/... cloud -> cloud.hanzo.ai/...
Each provider holds ONE OAuth client with a FIXED list of authorized URIs, so
social login could work on at most ONE property. Google's own error payload,
base64-decoded, reads `redirect_uri_mismatch`. GitHub's reads "redirect_uri is
not associated with this application".
IT WAS NEVER KMS OR SECRETS. The Google client_id
(113591532635-s8pvqrebprkbndluhftddmdap4htvu1p.apps.googleusercontent.com)
reached Google intact every time. Do not go looking at KMS again.
FIXED in c153004: the default is now the org's hosted ID host, read out of
DEFAULT_TENANTS via `idOriginFor()` so the `.id` hosts stay declared once.
Verified by executing resolveOrg under tsx with a production-shaped catalog:
hanzo.app / hanzo.chat / console.hanzo.ai / cloud.hanzo.ai -> hanzo.id/callback
id.lux.network -> lux.id/callback
FIRST ATTEMPT WAS WRONG — recorded so nobody repeats it: defaulting to
`iamIssuer` does NOT work. `hostSkeleton()` derives the issuer from the REQUEST
HOST too, so it is per-brand for the same reason. It must be a per-ORG constant.
### LEFT TO DO, IN ORDER
1. CONFIRM `ghcr.io/hanzoai/id:0.2.22` EXISTS, then bump the chart.
`crane digest ghcr.io/hanzoai/id:0.2.22` — a digest is the ONLY proof; the
build job's exit code lied twice this session. Chart:
universe/charts/app/values/hanzo/id.yaml (currently 0.2.21). Pin tag AND
digest together — a stale digest wins over the tag and serves the old build.
Build was fired by pushing to forge (see §BUILD PATH) as pf-runner-yuylsijvdnpb.
2. REGISTER THE CALLBACK URI with both providers. One entry each, whole fleet:
- GitHub OAuth App -> Authorization callback URL
- Google Cloud project -> Authorized redirect URIs
GET THE EXACT STRING EMPIRICALLY: click "Continue with GitHub" at hanzo.id and
read `redirect_uri=` out of the address bar BEFORE the error page. Do not
trust a predicted string.
Until BOTH (1) and (2) are done, social login stays broken and the error text
is identical either way.
3. WALLET LOGIN IS A DATA FIX, NOT CODE. `provider-web3` has canSignUp=true and
canSignIn=false on 6 of 11 apps (hanzo-app, hanzo-base, hanzo-docs,
hanzo-insights, hanzo-o11y, hanzo-world); symmetric and correct on
hanzo-chat, hanzo-cloud, hanzo-console, hanzo-platform. So it is inconsistent
seeding, not policy. Anyone who signed up with a wallet cannot sign back in.
SocialButtons.tsx:162 renders the flags faithfully — do not "fix" the code.
Needs an admin bearer token: the portal session cookie gets 401 on
/v1/iam/applications.
4. MFA / AUTHENTICATORS / PASSKEYS — COMPLETELY UNVERIFIED. Surfaces exist
(MfaEnrollForm, OTPForm, SmsConsentNotice, django_otp in insights) but NO
flow was driven end to end. Email OTP, SMS OTP, TOTP and WebAuthn each need
a real browser pass. Do not report any of them as working without driving it.
5. PLATFORM BUILD ENQUEUE IS BROKEN (separate from the above).
POST platform.hanzo.ai/v1/runner -> 400 "organizationId is required (no
DEFAULT_BUILD_ORG_ID configured)"; supplying `hanzo`, `admin/hanzo`, or the
real UUID dfb7a19b-108f-5150-8131-7d207488bf48 all -> 500 "enqueue failed:
FOREIGN KEY constraint failed". Cause: platform's /data/data.db was LAST
WRITTEN Jul 28 (auxiliary.db is live today) — the org row the build_job FK
references does not exist. No sqlite3 in the pod.
CTO ruling: DELETE `DEFAULT_BUILD_ORG_ID` entirely. A fallback org is a second
way to do things and it let this fail silently for a week; every build belongs
to a real prepaid org. Make the enqueue say "org X has no build account"
instead of leaking a raw SQL constraint. Fix lives in ~/work/hanzo/platform.
api.hanzo.ai/v1/runner also returned 503 during this window.
6. o11y-mcp — DELETE IT, do not debrand it. Branch debrand/no-signoz (016535f)
removes SigNoz+ClickHouse from go.mod and builds clean, but it is the wrong
fix: o11y already declares 353 typed zip ops and zip projects each into an MCP
tool from the same declaration. Cloud's door at POST api.hanzo.ai/v1/mcp
serves 932 tools across 116 apps. A hand-rolled Go MCP server is a second way
to do MCP, and its staleness proved it — it was publishing DDL for tables
HIP-0132 dropped. Archive the repo; regenerate plugin/o11y/mcp.json from the
typed ops (it holds 12 entries where o11y declares 353).
7. LEDGER CONSOLIDATION -> hanzoai/ledger. Two live forks of one Formance root.
ledger-fi is the live lineage (8 real PRs, already zip, 6 commits behind
upstream). Registry rule is Hanzo->hanzoai, so hanzo-fi is a fourth org that
exists for one repo. RENAME THE LIVE ONE FIRST (hanzo-fi/ledger ->
hanzoai/ledger) so GitHub's redirect protects `go get`, THEN delete the dead
twin — never the reverse or the redirect is stranded. Its clickhouse-go is
Formance's own dep and leaves with the consolidation. NOTE: 0 of 1,444 changed
lines in internal/api are ours; do not rewrite vendored paths beyond go.mod.
8. ALSO DEPRECATED: archive `hanzoai/datastore-go` — imported by NOBODY while 20
repos use `github.com/hanzo-ds/go`. Confirmed deprecated by the CTO.
### SHIPPED AND LIVE (verified by response body, not status code)
- iam@3f86f1f5e — ONE SSO seam. Two independent fixes reconciled:
* session is CREATED on every grant shape. The bug: `if f.Type != "code"`
guarded sessions.Set, while client.ts:170 sends type=code for every OAuth
login — so the IdP forgot the human the instant they signed in and the
fully-built silent-SSO branch had nothing to read.
* session is USABLE without UI: prompt=none now returns the code with no UI,
or error=login_required TO THE REDIRECT_URI (never a rendered page).
VERIFIED LIVE on hanzo.id. Also landed: __Host- cookie prefix, max_age
enforcement, and id_token_hint signature verification (without it a silent
renewal could return a code for a DIFFERENT human through a callback the RP
already trusts — an identity swap with nothing on screen).
- id@f2fc1e4 — the callback fix + semver: id-shared 0.1.2, id-auth 0.1.7,
id-onboarding 0.1.2, id-idv 0.1.1, root 0.2.22. id-connect NOT bumped (does
not depend on id-shared).
- id@2fe32a7 — deleted ProviderButtons.tsx, a dead 2-provider list that was in
no barrel and imported by nobody, sitting next to the 4-provider SocialButtons.
- console v8.5.36 — hero h1 now 61.6px line box for 56px glyphs. It was
line-height 1.12px: a ONE-PIXEL box under 30px glyphs, so the heading
overflowed onto its own subtitle. Cause: react-native-web appends `px` to
numeric style values absent from its unitless list, and lineHeight is absent,
so {lineHeight:1.12} compiled to `1.12px`.
- console 4656c316f4 — react-native-svg 15.15.5. console main had been
UNBUILDABLE since the gui-8 bump: @hanzogui/lucide-icons-2@8.0.0 imports
react-native-svg while declaring it in NEITHER dependencies. That is why
v8.5.33/34/35 never existed in GHCR and two weeks of fixes never shipped.
- cloud.hanzo.ai login — client_id=hanzo-cloud (was hanzo-app, whose client
carries only hanzo.app/auth/callback).
- datastore 8Gi -> 12Gi — ClickHouse derives max_server_memory_usage from the
cgroup at 0.9, so 8Gi WAS the 7.20 GiB ceiling and the server was refusing
reads with MEMORY_LIMIT_EXCEEDED at 7.88 GiB RSS.
### BUILD PATH THAT WORKS
platform.hanzo.ai/v1/runner is broken (§5). USE THE NATIVE FORGE PUSH:
git push forge origin/main:main # git.hanzo.ai fires .hanzo/workflows/deploy.yml
Verified working this session: git-runner fleet 4/4 Running, and
build-console / build-docs / build-openapi all Completed within 30 min.
NOTE forge/main was 8 commits BEHIND origin/main for `id` — a working builder
would still have built the wrong tree. Check both remotes agree.
### METHOD NOTES — these caught real errors, four times
- A CHART BUMP IS NOT A DEPLOY; A MERGED COMMIT IS NOT PRODUCTION. Verify what
RUNS. I pinned console to v8.5.35, an image that NEVER EXISTED (GHCR 403'd
anonymously and I shipped anyway on "RollingUpdate fails safe" — it did fail
safe, and it also shipped nothing while reading as done).
- A GREEN TEST COMMAND IS NOT A GREEN TEST. `pnpm --filter @hanzo/id-shared test`
exits 0 having run NOTHING — that package declares no `test` script. Root has
`test: vitest run`, but vitest is not installed in the checkout. The org.ts
behaviour was verified by EXECUTING resolveOrg under tsx.
- `$?` AFTER A PIPE IS THE PIPE'S STATUS. Printed "build: 0" for a failed build.
Put echo $? on its own line.
- MEASURE origin/main, NEVER A LOCAL CHECKOUT, and run `git status -sb` first.
Stale checkouts produced five false findings, worst: hanzoai/iam called a beego
carrier while the local tree sat 388 commits behind on a dead branch that ships
its own DEPRECATED.md. node is 7,565 behind. Filter `// indirect` too.
- ROUTE/SYMBOL COUNTS IN A REPO PROVE NOTHING ABOUT WHAT SHIPS. bootnode was
ranked a top conversion target on 206 chi routes that ship in ZERO binaries
(its live API is Python/uvicorn). Use `go list -deps` against the binary the
Dockerfile builds, plus what runs in the cluster.
### SECURITY, OPEN
- NO FIRST-USE CONSENT on prompt=none: a signed-in victim top-level navigated to
authorize?client_id=<attacker>&prompt=none yields a code to that client's
registered redirect_uri. SameSite=Lax sends the cookie on a top-level GET and
Sec-Fetch cannot help — it is a genuine navigation. Bounded by MintFor's
tenancy rule, so blast radius depends on who may set IsShared. This is the
standard reason IdPs gate first use of a client behind consent.
- CORS EDGE, RE-CHECK: login.go's own comment records a proxy on the
hanzo.ai/hanzo.id zones once reflecting *.hanzo.ai with
Access-Control-Allow-Credentials:true, which would make the credential-less
mint reachable from any subdomain. ACAO measures as exactly https://hanzo.id
today; a hostile Origin was NOT tested. The SSO fix puts live sessions in far
more browsers, so this matters more now.
- hanzo_iam_access_token on domain hanzo.app is a full RS256 JWT in a
NON-HttpOnly, JS-readable cookie.
- Application.EnableSigninSession — declared at pkg/schema/application.go:146,
set TRUE on every app, READ BY NO CODE. Revive it as a real gate or delete it.
- The __Host- cookie rename INVALIDATES EVERY LIVE SESSION on deploy. One
re-login per human. Decide accept-vs-dual-read before rolling.
- CLEANUP OWED: qa-signup-probe-0803@hanzo.ai is a real account created to prove
signup works end to end; delete it. A DigitalOcean PAT (dop_v1_ff09e128…) was
pasted into the session transcript and is on disk — ROTATE IT.
### DEAD ENDS — do not re-derive
- insights.hanzo.ai is NOT broken. It is SSO-gated and the chain works:
/login -> /login/oidc/ -> hanzo.id/…authorize?client_id=hanzo-insights -> 200.
The "532 MIME errors, empty #root" report was an unauthenticated browser
following those 302s.
- THERE IS NO SHADCN TO KILL. Neither console main nor blue3/ui-shadcn-explicit
contains shadcn, radix or tailwind; main's only two matches are comments saying
the console deliberately is not the shadcn build. That branch is 906 behind /
10 ahead and REGRESSES deps (@hanzo/gui 7.3.0 vs ^8.0.0, @hanzo/iam ^0.13.6 vs
^0.21.2). ABANDON IT, do not merge. Tamagui-native is already true on main.
- arc is dead and arcd was removed. It was a systemd --user service on the spark
workstation, NOT in k8s; the k8s ARC removal happened 2026-07-29. Its 65 jobs
in 30 days were 65/65 FAILURES from one cron. hanzoai/ci's runner default was
NEVER arc — it is hanzo-build-linux-amd64, served by git-runner.
+2
View File
@@ -1,3 +1,5 @@
<p align="center"><img src=".github/hero.svg" alt="id" width="880"></p>
# @hanzo/id
White-label login + identity verification portal. One Vite SPA, four hosts
+15
View File
@@ -0,0 +1,15 @@
# @hanzo/id-account
The account portal: branded account-management pages at `account.hanzo.id`,
`account.lux.id`, … served by a Cloudflare Worker, authenticated with IAM access
tokens over the OAuth code exchange.
It lives HERE, in `hanzoai/id`, because it is identity UI — the same brands, the
same IAM, the same login redirect as `apps/web`. It used to live in
`hanzoai/account`, which is a Go module: one repo held two unrelated codebases
under one name (`main` was this Worker, the Go module survived only on a `go`
branch and its tags). That collision is what made
`github.com/hanzoai/account` unresolvable as a Go module from a clean checkout.
`hanzoai/account` is now the Go module and nothing else — see its README for the
billing-account rule it owns.
+13
View File
@@ -0,0 +1,13 @@
{
"name": "@hanzo/id-account",
"version": "1.0.0",
"private": true,
"scripts": {
"dev": "wrangler dev",
"deploy": "wrangler deploy"
},
"devDependencies": {
"wrangler": "^4.65.0"
},
"description": "Branded account-management pages for account.<brand>.id \u2014 a Cloudflare Worker."
}
+803
View File
@@ -0,0 +1,803 @@
/**
* Account Portal — Cloudflare Worker
*
* Serves branded account management pages for:
* - account.hanzo.id (Hanzo brand)
* - account.lux.id (Lux brand)
*
* Authentication: Uses IAM access tokens via OAuth code exchange.
* Users are redirected to their brand's login page if unauthenticated.
*/
const IAM_ORIGIN = 'https://iam.hanzo.ai';
// Brand configuration keyed by hostname
const BRANDS = {
'account.hanzo.id': {
name: 'Hanzo',
domain: 'hanzo.id',
loginUrl: 'https://hanzo.id/login',
bg: '#0a0a0a',
surface: '#111111',
border: '#222222',
accent: '#fd4444',
clientId: 'hanzo-app-client-id',
logo: `<svg viewBox="0 0 100 100" width="40" height="40" fill="none" xmlns="http://www.w3.org/2000/svg">
<path d="M50 5L90 27.5V72.5L50 95L10 72.5V27.5L50 5Z" stroke="#fd4444" stroke-width="4"/>
<path d="M30 35V65M70 35V65M30 50H70" stroke="#fd4444" stroke-width="4" stroke-linecap="round"/>
</svg>`,
},
'account.lux.id': {
name: 'Lux',
domain: 'lux.id',
loginUrl: 'https://lux.id/login',
bg: '#050508',
surface: '#0c0c10',
border: '#222222',
accent: '#ffffff',
clientId: 'lux-app-client-id',
logo: `<svg viewBox="0 0 100 100" width="40" height="40" fill="none" xmlns="http://www.w3.org/2000/svg">
<polygon points="50,10 90,75 10,75" stroke="white" stroke-width="3" fill="none"/>
<polygon points="50,30 72,68 28,68" stroke="white" stroke-width="2" fill="none"/>
</svg>`,
},
};
function getBrand(hostname) {
return BRANDS[hostname] || BRANDS['account.hanzo.id'];
}
function htmlResponse(html) {
return new Response(html, {
headers: {
'content-type': 'text/html;charset=UTF-8',
'cache-control': 'no-store',
'X-Content-Type-Options': 'nosniff',
'X-Frame-Options': 'DENY',
},
});
}
// Parse access token from cookie
function getToken(request) {
const cookie = request.headers.get('Cookie') || '';
const match = cookie.match(/account_token=([^;]+)/);
return match ? decodeURIComponent(match[1]) : null;
}
// Fetch user info from IAM using token
async function getUserInfo(token) {
const res = await fetch(`${IAM_ORIGIN}/v1/iam/userinfo`, {
headers: { Authorization: `Bearer ${token}` },
});
if (!res.ok) return null;
const data = await res.json();
return data;
}
// Fetch full user object for editing
async function getUser(token, owner, name) {
const res = await fetch(`${IAM_ORIGIN}/v1/iam/get-user?id=${encodeURIComponent(owner)}/${encodeURIComponent(name)}`, {
headers: { Authorization: `Bearer ${token}` },
});
if (!res.ok) return null;
const data = await res.json();
return data.data || null;
}
// Build the OAuth login URL for the brand. Points at the brand's branded
// /login page (front-door worker) with the OAuth params; that page renders
// the two-pane login and emits the canonical /v1/iam/* calls itself — no
// bare /oauth/authorize, no host leak to iam.hanzo.ai.
function buildLoginUrl(brand, callbackUrl) {
const params = new URLSearchParams({
client_id: brand.clientId,
redirect_uri: callbackUrl,
response_type: 'code',
scope: 'openid profile email',
state: 'account',
});
return `${brand.loginUrl}?${params.toString()}`;
}
// Exchange authorization code for access token
async function exchangeCode(code, callbackUrl, brand) {
const res = await fetch(`${IAM_ORIGIN}/v1/iam/oauth/token`, {
method: 'POST',
headers: { 'Content-Type': 'application/x-www-form-urlencoded' },
body: new URLSearchParams({
grant_type: 'authorization_code',
code,
redirect_uri: callbackUrl,
client_id: brand.clientId,
}).toString(),
});
if (!res.ok) return null;
const data = await res.json();
return data.access_token || null;
}
function renderAccountPage(brand, user, fullUser) {
const providers = [];
// Extract linked providers from user object
const providerFields = ['github', 'google', 'facebook', 'twitter', 'linkedin', 'discord', 'wechat', 'dingtalk'];
if (fullUser) {
for (const p of providerFields) {
if (fullUser[p] && fullUser[p] !== '') {
providers.push({ name: p, id: fullUser[p] });
}
}
// Check for MetaMask/Web3 wallet
if (fullUser.metamask && fullUser.metamask !== '') {
providers.push({ name: 'web3', id: fullUser.metamask });
} else if (fullUser.web3onboard && fullUser.web3onboard !== '') {
providers.push({ name: 'web3', id: fullUser.web3onboard });
}
}
return `<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Account - ${brand.name}</title>
<style>
* { margin: 0; padding: 0; box-sizing: border-box; }
body {
font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif;
background: ${brand.bg};
color: #fff;
min-height: 100vh;
}
.topbar {
border-bottom: 1px solid ${brand.border};
padding: 0.75rem 1.5rem;
display: flex;
align-items: center;
justify-content: space-between;
background: ${brand.surface};
}
.topbar-brand {
display: flex;
align-items: center;
gap: 0.75rem;
font-weight: 600;
font-size: 1.1rem;
}
.topbar-actions { display: flex; gap: 0.75rem; align-items: center; }
.topbar-actions a {
color: #999;
text-decoration: none;
font-size: 0.85rem;
padding: 0.4rem 0.8rem;
border-radius: 6px;
border: 1px solid ${brand.border};
}
.topbar-actions a:hover { color: #fff; border-color: #555; }
.topbar-actions .logout { color: #ff6b6b; border-color: #ff6b6b33; }
.topbar-actions .logout:hover { background: #ff6b6b11; }
.container {
max-width: 800px;
margin: 2rem auto;
padding: 0 1.5rem;
}
.section {
background: ${brand.surface};
border: 1px solid ${brand.border};
border-radius: 12px;
padding: 1.5rem;
margin-bottom: 1.5rem;
}
.section h2 {
font-size: 1.1rem;
margin-bottom: 1rem;
padding-bottom: 0.75rem;
border-bottom: 1px solid ${brand.border};
}
.field {
display: flex;
justify-content: space-between;
align-items: center;
padding: 0.6rem 0;
}
.field + .field { border-top: 1px solid ${brand.border}22; }
.field-label { color: #888; font-size: 0.85rem; min-width: 120px; }
.field-value { font-size: 0.92rem; }
.field-action {
color: ${brand.accent};
text-decoration: none;
font-size: 0.82rem;
cursor: pointer;
background: none;
border: 1px solid ${brand.accent}44;
padding: 0.3rem 0.6rem;
border-radius: 6px;
}
.field-action:hover { background: ${brand.accent}11; }
.provider-list { display: flex; flex-direction: column; gap: 0.5rem; }
.provider-item {
display: flex;
align-items: center;
justify-content: space-between;
padding: 0.6rem 0.8rem;
border: 1px solid ${brand.border};
border-radius: 8px;
background: ${brand.bg};
}
.provider-item .name {
text-transform: capitalize;
font-weight: 500;
}
.provider-item .id {
color: #888;
font-size: 0.8rem;
margin-left: 0.5rem;
}
.provider-item .unlink {
color: #ff6b6b;
font-size: 0.78rem;
cursor: pointer;
background: none;
border: 1px solid #ff6b6b33;
padding: 0.2rem 0.5rem;
border-radius: 4px;
}
.add-provider {
display: flex;
gap: 0.5rem;
margin-top: 0.75rem;
flex-wrap: wrap;
}
.add-btn {
display: flex;
align-items: center;
gap: 0.4rem;
padding: 0.4rem 0.8rem;
border: 1px dashed ${brand.border};
border-radius: 8px;
color: #999;
font-size: 0.82rem;
cursor: pointer;
background: none;
text-decoration: none;
}
.add-btn:hover { border-color: #555; color: #fff; }
.badge {
display: inline-block;
padding: 0.15rem 0.45rem;
border-radius: 4px;
font-size: 0.72rem;
font-weight: 500;
background: #1a472a;
color: #6ee7b7;
margin-left: 0.4rem;
}
.badge.unverified { background: #472a1a; color: #e7b76e; }
.danger-zone {
border-color: #ff6b6b33;
}
.danger-zone h2 { color: #ff6b6b; }
.msg {
display: none;
padding: 0.75rem;
border-radius: 8px;
margin-bottom: 1rem;
font-size: 0.85rem;
}
.msg.success { background: #1a472a; color: #6ee7b7; display: block; }
.msg.error { background: #472a1a; color: #e7b76e; display: block; }
.avatar-row {
display: flex;
align-items: center;
gap: 1rem;
padding: 0.6rem 0;
}
.avatar {
width: 64px;
height: 64px;
border-radius: 50%;
background: ${brand.border};
display: flex;
align-items: center;
justify-content: center;
font-size: 1.5rem;
font-weight: 700;
color: #fff;
overflow: hidden;
}
.avatar img { width: 100%; height: 100%; object-fit: cover; }
.modal-overlay {
display: none;
position: fixed;
inset: 0;
background: rgba(0,0,0,0.7);
z-index: 100;
align-items: center;
justify-content: center;
}
.modal-overlay.active { display: flex; }
.modal {
background: ${brand.surface};
border: 1px solid ${brand.border};
border-radius: 12px;
padding: 1.5rem;
width: 100%;
max-width: 420px;
}
.modal h3 { margin-bottom: 1rem; }
.modal input {
width: 100%;
border-radius: 8px;
border: 1px solid ${brand.border};
background: ${brand.bg};
color: #fff;
padding: 0.6rem 0.8rem;
font-size: 0.9rem;
margin-bottom: 0.75rem;
}
.modal-actions {
display: flex;
gap: 0.5rem;
justify-content: flex-end;
margin-top: 0.75rem;
}
.modal-actions button {
padding: 0.5rem 1rem;
border-radius: 6px;
border: 1px solid ${brand.border};
background: none;
color: #fff;
cursor: pointer;
font-size: 0.85rem;
}
.modal-actions .primary {
background: ${brand.accent};
color: ${brand.accent === '#ffffff' ? '#111' : '#fff'};
border-color: ${brand.accent};
}
@media (max-width: 640px) {
.container { padding: 0 1rem; margin: 1rem auto; }
.section { padding: 1rem; }
.field { flex-direction: column; align-items: flex-start; gap: 0.3rem; }
}
</style>
</head>
<body>
<nav class="topbar">
<div class="topbar-brand">
${brand.logo}
<span>${brand.name} Account</span>
</div>
<div class="topbar-actions">
<a href="https://${brand.domain}">Back to ${brand.name}</a>
<a href="https://${brand.domain}/logout" class="logout">Sign Out</a>
</div>
</nav>
<div class="container">
<div id="msg" class="msg"></div>
<div class="section">
<h2>Profile</h2>
<div class="avatar-row">
<div class="avatar">
${user.picture ? `<img src="${user.picture}" alt="Avatar">` : (user.name || 'U').charAt(0).toUpperCase()}
</div>
<div>
<div style="font-weight:600;font-size:1.1rem;">${user.preferred_username || user.name || 'User'}</div>
<div style="color:#888;font-size:0.85rem;">${user.email || ''}</div>
</div>
</div>
<div class="field">
<span class="field-label">Display Name</span>
<span class="field-value">${user.name || '—'}</span>
<button class="field-action" onclick="editField('name','${(user.name || '').replace(/'/g, "\\'")}')">Edit</button>
</div>
<div class="field">
<span class="field-label">Email</span>
<span class="field-value">
${user.email || '—'}
${user.email_verified ? '<span class="badge">Verified</span>' : '<span class="badge unverified">Unverified</span>'}
</span>
</div>
<div class="field">
<span class="field-label">Phone</span>
<span class="field-value">${user.phone || '—'}</span>
</div>
</div>
<div class="section">
<h2>Login Methods</h2>
<div class="provider-list">
<div class="provider-item">
<div>
<span class="name">Email & Password</span>
<span class="id">${user.email || 'Not set'}</span>
</div>
<button class="field-action" onclick="openPasswordModal()">Change Password</button>
</div>
${providers.map(p => `
<div class="provider-item">
<div>
<span class="name">${p.name === 'web3' ? 'Web3 Wallet' : p.name}</span>
<span class="id">${p.name === 'web3' ? p.id.slice(0, 6) + '...' + p.id.slice(-4) : p.id}</span>
</div>
<button class="unlink" onclick="unlinkProvider('${p.name}')">Unlink</button>
</div>`).join('')}
</div>
<div class="add-provider">
<a class="add-btn" href="https://${brand.domain}/v1/iam/oauth/authorize?client_id=${brand.clientId}&redirect_uri=${encodeURIComponent(`https://${brand.domain}/callback`)}&response_type=code&scope=openid+profile+email&provider=provider-google">
+ Google
</a>
<a class="add-btn" href="https://${brand.domain}/v1/iam/oauth/authorize?client_id=${brand.clientId}&redirect_uri=${encodeURIComponent(`https://${brand.domain}/callback`)}&response_type=code&scope=openid+profile+email&provider=provider-github">
+ GitHub
</a>
<button class="add-btn" onclick="linkWallet()">+ Web3 Wallet</button>
</div>
</div>
<div class="section">
<h2>Security</h2>
<div class="field">
<span class="field-label">Two-Factor Auth</span>
<span class="field-value">${fullUser && fullUser.totpSecret ? '<span class="badge">Enabled</span>' : 'Not enabled'}</span>
<button class="field-action" onclick="window.location.href='https://iam.hanzo.ai/account#mfa'">${fullUser && fullUser.totpSecret ? 'Manage' : 'Enable'}</button>
</div>
<div class="field">
<span class="field-label">Last Sign-in</span>
<span class="field-value">${fullUser && fullUser.lastSigninTime ? new Date(fullUser.lastSigninTime).toLocaleString() : '—'}</span>
</div>
<div class="field">
<span class="field-label">Last Sign-in IP</span>
<span class="field-value">${fullUser && fullUser.lastSigninIp ? fullUser.lastSigninIp : '—'}</span>
</div>
</div>
<div class="section danger-zone">
<h2>Danger Zone</h2>
<div class="field">
<span class="field-label">Delete Account</span>
<span class="field-value" style="color:#888;font-size:0.82rem;">Permanently delete your account and all data</span>
<button class="field-action" style="color:#ff6b6b;border-color:#ff6b6b44;" onclick="confirmDelete()">Delete Account</button>
</div>
</div>
</div>
<!-- Password Change Modal -->
<div class="modal-overlay" id="password-modal">
<div class="modal">
<h3>Change Password</h3>
<input type="password" id="old-password" placeholder="Current password" autocomplete="current-password">
<input type="password" id="new-password" placeholder="New password (min 8 characters)" autocomplete="new-password">
<input type="password" id="confirm-password" placeholder="Confirm new password" autocomplete="new-password">
<div id="pw-error" style="color:#ff8f8f;font-size:0.82rem;display:none;margin-bottom:0.5rem;"></div>
<div class="modal-actions">
<button onclick="closePasswordModal()">Cancel</button>
<button class="primary" onclick="changePassword()">Update Password</button>
</div>
</div>
</div>
<!-- Edit Field Modal -->
<div class="modal-overlay" id="edit-modal">
<div class="modal">
<h3 id="edit-title">Edit Field</h3>
<input type="text" id="edit-value">
<div class="modal-actions">
<button onclick="closeEditModal()">Cancel</button>
<button class="primary" onclick="saveField()">Save</button>
</div>
</div>
</div>
<script>
const TOKEN = document.cookie.match(/account_token=([^;]+)/)?.[1] ? decodeURIComponent(document.cookie.match(/account_token=([^;]+)/)[1]) : null;
const IAM = '${IAM_ORIGIN}';
const BRAND_DOMAIN = '${brand.domain}';
let editingField = null;
function showMsg(text, type) {
const el = document.getElementById('msg');
el.textContent = text;
el.className = 'msg ' + type;
setTimeout(() => { el.className = 'msg'; }, 5000);
}
function openPasswordModal() {
document.getElementById('password-modal').classList.add('active');
}
function closePasswordModal() {
document.getElementById('password-modal').classList.remove('active');
document.getElementById('old-password').value = '';
document.getElementById('new-password').value = '';
document.getElementById('confirm-password').value = '';
document.getElementById('pw-error').style.display = 'none';
}
async function changePassword() {
const oldPw = document.getElementById('old-password').value;
const newPw = document.getElementById('new-password').value;
const confirm = document.getElementById('confirm-password').value;
const errEl = document.getElementById('pw-error');
if (!oldPw || !newPw) { errEl.textContent = 'All fields required'; errEl.style.display = 'block'; return; }
if (newPw.length < 8) { errEl.textContent = 'Password must be at least 8 characters'; errEl.style.display = 'block'; return; }
if (newPw !== confirm) { errEl.textContent = 'Passwords do not match'; errEl.style.display = 'block'; return; }
try {
const res = await fetch('/v1/iam/set-password', {
method: 'POST',
headers: {
'Content-Type': 'application/x-www-form-urlencoded',
Authorization: 'Bearer ' + TOKEN,
},
body: new URLSearchParams({
userOwner: '${fullUser ? fullUser.owner : 'hanzo'}',
userName: '${fullUser ? fullUser.name : ''}',
oldPassword: oldPw,
newPassword: newPw,
}).toString(),
});
const data = await res.json();
if (data.status === 'ok') {
closePasswordModal();
showMsg('Password updated successfully', 'success');
} else {
errEl.textContent = data.msg || 'Failed to update password';
errEl.style.display = 'block';
}
} catch (e) {
errEl.textContent = 'Network error';
errEl.style.display = 'block';
}
}
function editField(field, currentValue) {
editingField = field;
document.getElementById('edit-title').textContent = 'Edit ' + field.charAt(0).toUpperCase() + field.slice(1);
document.getElementById('edit-value').value = currentValue;
document.getElementById('edit-modal').classList.add('active');
}
function closeEditModal() {
document.getElementById('edit-modal').classList.remove('active');
editingField = null;
}
async function saveField() {
if (!editingField) return;
const value = document.getElementById('edit-value').value;
try {
const userObj = { owner: '${fullUser ? fullUser.owner : 'hanzo'}', name: '${fullUser ? fullUser.name : ''}' };
if (editingField === 'name') userObj.displayName = value;
const res = await fetch('/v1/iam/update-user?id=${fullUser ? fullUser.owner + '/' + fullUser.name : ''}', {
method: 'POST',
headers: {
'Content-Type': 'application/json',
Authorization: 'Bearer ' + TOKEN,
},
body: JSON.stringify(userObj),
});
const data = await res.json();
if (data.status === 'ok') {
closeEditModal();
showMsg('Updated successfully. Refreshing...', 'success');
setTimeout(() => location.reload(), 1000);
} else {
showMsg(data.msg || 'Update failed', 'error');
}
} catch (e) {
showMsg('Network error', 'error');
}
}
async function unlinkProvider(provider) {
if (!confirm('Unlink ' + provider + ' from your account?')) return;
try {
var field = provider === 'web3' ? 'metamask' : provider;
var userObj = { owner: '${fullUser ? fullUser.owner : 'hanzo'}', name: '${fullUser ? fullUser.name : ''}' };
userObj[field] = '';
var res = await fetch('/v1/iam/update-user?id=${fullUser ? fullUser.owner + '/' + fullUser.name : ''}', {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify(userObj),
});
var data = await res.json();
if (data.status === 'ok') {
showMsg(provider + ' unlinked successfully', 'success');
setTimeout(function() { location.reload(); }, 1000);
} else {
showMsg(data.msg || 'Failed to unlink ' + provider, 'error');
}
} catch (e) {
showMsg('Network error: ' + e.message, 'error');
}
}
async function linkWallet() {
if (typeof window.ethereum === 'undefined') {
showMsg('Please install MetaMask or another Web3 wallet', 'error');
return;
}
try {
var accounts = await window.ethereum.request({ method: 'eth_requestAccounts' });
var address = accounts[0];
if (!address) { showMsg('No wallet account found', 'error'); return; }
// Sign a message to prove wallet ownership
var message = 'Link wallet ' + address + ' to ' + BRAND_DOMAIN + ' account for ${fullUser ? fullUser.name : 'user'}';
await window.ethereum.request({ method: 'personal_sign', params: [message, address] });
// Update user with wallet address
var userObj = { owner: '${fullUser ? fullUser.owner : 'hanzo'}', name: '${fullUser ? fullUser.name : ''}', metamask: address };
var res = await fetch('/v1/iam/update-user?id=${fullUser ? fullUser.owner + '/' + fullUser.name : ''}', {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify(userObj),
});
var data = await res.json();
if (data.status === 'ok') {
showMsg('Wallet linked: ' + address.slice(0, 6) + '...' + address.slice(-4), 'success');
setTimeout(function() { location.reload(); }, 1500);
} else {
showMsg(data.msg || 'Failed to link wallet', 'error');
}
} catch (err) {
if (err.code === 4001) return; // User rejected
showMsg(err.message || 'Failed to connect wallet', 'error');
}
}
function confirmDelete() {
if (!confirm('Are you sure? This action is permanent and cannot be undone.')) return;
if (!confirm('This will permanently delete your account and all associated data. Type OK to confirm.')) return;
showMsg('Account deletion requires verification. Redirecting to IAM...', 'error');
window.location.href = 'https://iam.hanzo.ai/account';
}
</script>
</body>
</html>`;
}
function renderLoginRedirectPage(brand, loginUrl) {
return `<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Sign In Required - ${brand.name}</title>
<meta http-equiv="refresh" content="2;url=${loginUrl}">
<style>
body {
font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif;
background: ${brand.bg};
color: #fff;
display: flex;
align-items: center;
justify-content: center;
min-height: 100vh;
margin: 0;
}
.card {
text-align: center;
padding: 2rem;
border: 1px solid ${brand.border};
border-radius: 12px;
background: ${brand.surface};
max-width: 400px;
}
a { color: ${brand.accent}; }
</style>
</head>
<body>
<div class="card">
${brand.logo}
<h2 style="margin-top:1rem;">Sign in to continue</h2>
<p style="color:#888;margin-top:0.5rem;">Redirecting to ${brand.name} login...</p>
<p style="margin-top:1rem;"><a href="${loginUrl}">Click here if not redirected</a></p>
</div>
</body>
</html>`;
}
export default {
async fetch(request, env) {
const url = new URL(request.url);
const hostname = url.hostname;
const pathname = url.pathname;
const brand = getBrand(hostname);
const callbackUrl = `https://${hostname}/callback`;
// Handle OAuth callback — exchange code for token
if (pathname === '/callback') {
const code = url.searchParams.get('code');
if (!code) {
return new Response(null, {
status: 302,
headers: { Location: buildLoginUrl(brand, callbackUrl) },
});
}
const token = await exchangeCode(code, callbackUrl, brand);
if (!token) {
return new Response(null, {
status: 302,
headers: { Location: buildLoginUrl(brand, callbackUrl) },
});
}
// Set token cookie and redirect to account page
return new Response(null, {
status: 302,
headers: {
Location: '/',
'Set-Cookie': `account_token=${encodeURIComponent(token)}; Path=/; HttpOnly; Secure; SameSite=Lax; Max-Age=86400`,
},
});
}
// Handle logout
if (pathname === '/logout') {
return new Response(null, {
status: 302,
headers: {
Location: `https://${brand.domain}`,
'Set-Cookie': 'account_token=; Path=/; Expires=Thu, 01 Jan 1970 00:00:00 GMT',
},
});
}
// Proxy IAM API calls (for client-side JS). Same-origin /v1/iam/* from
// the account page is forwarded to IAM; the access token is attached
// server-side from the HttpOnly cookie so it never rides in page JS.
if (pathname.startsWith('/v1/iam/')) {
const token = getToken(request);
const iamUrl = new URL(pathname + url.search, IAM_ORIGIN);
const headers = new Headers(request.headers);
headers.set('Host', new URL(IAM_ORIGIN).hostname);
if (token) headers.set('Authorization', `Bearer ${token}`);
const iamRes = await fetch(iamUrl.toString(), {
method: request.method,
headers,
body: request.method !== 'GET' ? request.body : undefined,
});
return new Response(iamRes.body, {
status: iamRes.status,
headers: {
'content-type': iamRes.headers.get('content-type') || 'application/json',
'cache-control': 'no-store',
},
});
}
// Check authentication for all other routes
const token = getToken(request);
if (!token) {
const loginUrl = buildLoginUrl(brand, callbackUrl);
return new Response(null, {
status: 302,
headers: { Location: loginUrl },
});
}
// Get user info
const user = await getUserInfo(token);
if (!user || !user.name) {
// Token expired or invalid — re-authenticate
const loginUrl = buildLoginUrl(brand, callbackUrl);
return new Response(null, {
status: 302,
headers: {
Location: loginUrl,
'Set-Cookie': 'account_token=; Path=/; Expires=Thu, 01 Jan 1970 00:00:00 GMT',
},
});
}
// Get full user object for detailed info
const fullUser = await getUser(token, user.owner || 'hanzo', user.preferred_username || user.name);
// Serve account page
return htmlResponse(renderAccountPage(brand, user, fullUser));
},
};
+12
View File
@@ -0,0 +1,12 @@
name = "account-portal"
main = "src/worker.js"
compatibility_date = "2024-01-01"
workers_dev = false
# Custom domains are managed via CF Workers Custom Domains API:
# account.hanzo.id -> zone hanzo.id
# account.lux.id -> zone lux.id
# DNS records and SSL certs are automatically provisioned.
[vars]
IAM_ORIGIN = "https://iam.hanzo.ai"
+5 -4
View File
@@ -1,8 +1,8 @@
{
"name": "@hanzo/id-web",
"private": true,
"version": "0.1.26",
"description": "Hanzo ID \u2014 white-label login / signup / IDV portal. Vite + React 19 + @hanzo/gui. Same image serves hanzo.id / lux.id / zoo.id / pars.id.",
"version": "0.1.34",
"description": "Hanzo ID \u2014 white-label login / signup / IDV portal. Vite + React 19, styled from @hanzo/design tokens. Same image serves hanzo.id / lux.id / zoo.id / pars.id.",
"type": "module",
"scripts": {
"dev": "vite",
@@ -13,8 +13,9 @@
"dependencies": {
"@crossmarkio/sdk": "^0.4.0",
"@hanzo/brand": "^1.3.0",
"@hanzo/gui": "^7.2.4",
"@hanzo/iam": "^0.13.1",
"@hanzo/design": "^0.4.9",
"@hanzo/event": "^0.3.11",
"@hanzo/iam": "^0.21.1",
"@hanzo/id-auth": "workspace:*",
"@hanzo/id-connect": "workspace:*",
"@hanzo/id-idv": "workspace:*",
+21 -21
View File
@@ -1,5 +1,5 @@
import { useEffect, useMemo, useState } from 'react'
import { loadBrand, parseCatalog, resolveTenant, type BrandContract, type TenantConfig } from '@hanzo/id-shared'
import { loadBrand, catalogOf, parseCatalog, resolveOrg, idBrandLabel, type BrandContract, type OrgConfig } from '@hanzo/id-shared'
import { createAuthClient } from '@hanzo/id-auth'
import { Portal } from './pages/Portal'
import { Login } from './pages/Login'
@@ -10,45 +10,45 @@ import { Onboarding } from './pages/Onboarding'
import { DeviceApproval } from './pages/DeviceApproval'
/**
* Top-level wiring. Resolves tenant + brand once on mount, then routes via
* Top-level wiring. Resolves org + brand once on mount, then routes via
* `window.location.pathname`. No router lib needed — this app is 5 pages,
* `<a href>` is enough. Adding paths is a switch case.
*/
export function App() {
const [tenant, setTenant] = useState<TenantConfig | null>(null)
const [org, setOrg] = useState<OrgConfig | null>(null)
const [brand, setBrand] = useState<BrandContract | null>(null)
const [error, setError] = useState<string | null>(null)
useEffect(() => {
let cancelled = false
async function boot() {
// The runtime serves the per-host tenant catalog at /config.json
// (templated from SPA_IAM_TENANT_CONFIG_JSON by the static server). Read
// it from there — NOT a `window.__ID_CATALOG__` global, which the runtime
// never injects (relying on it silently dropped every catalog-only host,
// e.g. osage.id, to the bundled Hanzo default). Fall back to the inlined
// global, then empty, so a host always resolves to something.
// The runtime serves the per-host org catalog at /config.json — NOT a
// `window.__ID_CATALOG__` global, which the runtime never injects
// (relying on it silently dropped every catalog-only host, e.g. osage.id,
// to the bundled Hanzo default). Fall back to the global, then empty, so a
// host always resolves to something.
//
// `catalogOf` owns which key that payload uses — the name is the server's,
// not ours (see its doc). Reading the wrong one is a silent total-catalog
// outage, so it is pinned by a test next to the resolver it feeds.
let catalogRaw: string | undefined
try {
const res = await fetch('/config.json', { cache: 'no-store' })
if (res.ok) {
const cfg = (await res.json()) as { iamTenantConfigJson?: string }
catalogRaw = cfg.iamTenantConfigJson
}
if (res.ok) catalogRaw = catalogOf(await res.json())
} catch {
// network/parse error → fall back below
}
if (!catalogRaw) {
catalogRaw = (window as unknown as { __ID_CATALOG__?: string }).__ID_CATALOG__
}
const t = resolveTenant(window.location.hostname, { catalog: parseCatalog(catalogRaw) })
const t = resolveOrg(window.location.hostname, { catalog: parseCatalog(catalogRaw) })
if (cancelled) return
setTenant(t)
setOrg(t)
try {
const b = await loadBrand(t.brandPackage)
if (cancelled) return
setBrand(b)
document.title = `Sign in — ${b.name}`
document.title = idBrandLabel(b, t.orgId)
const fav = document.getElementById('favicon') as HTMLLinkElement | null
if (fav && b.faviconUrl) fav.href = b.faviconUrl
} catch (e) {
@@ -61,10 +61,10 @@ export function App() {
}
}, [])
const client = useMemo(() => (tenant ? createAuthClient({ tenant }) : null), [tenant])
const client = useMemo(() => (org ? createAuthClient({ org }) : null), [org])
if (error) return <div className="hanzo-id-error">{error}</div>
if (!tenant || !brand || !client) return <div>Loading</div>
if (!org || !brand || !client) return <div>Loading</div>
const path = window.location.pathname
// Device-authorization approval (RFC 8628). Must precede the `/login` catch
@@ -74,7 +74,7 @@ export function App() {
if (path === '/login' || path.startsWith('/login/')) return <Login client={client} brand={brand} />
if (path === '/signup' || path.startsWith('/signup/')) return <Signup client={client} brand={brand} />
if (path === '/forget' || path === '/forgot' || path.startsWith('/forg')) return <Forgot client={client} brand={brand} />
if (path === '/callback' || path.startsWith('/callback/')) return <Callback tenant={tenant} brand={brand} />
if (path === '/onboarding' || path.startsWith('/onboarding/')) return <Onboarding tenant={tenant} brand={brand} />
return <Portal client={client} brand={brand} tenant={tenant} />
if (path === '/callback' || path.startsWith('/callback/')) return <Callback org={org} brand={brand} />
if (path === '/onboarding' || path.startsWith('/onboarding/')) return <Onboarding org={org} brand={brand} />
return <Portal client={client} brand={brand} org={org} />
}
+222
View File
@@ -0,0 +1,222 @@
/**
* The telemetry gate, tested where it is actually true or false: on the bytes
* the client puts on the wire.
*
* The defect this guards is not hypothetical and is not visible in review. The
* obvious way to keep an OAuth code out of telemetry — "send the pathname, never
* the href" — DOES NOT WORK against @hanzo/event, because `build()` stamps
* `url: window.location.href` onto every event it assembles regardless of the
* `path` the caller passed. A pageview from `/callback?code=…&state=…` therefore
* ships the authorization code while `path` reads a clean `/callback`, and the
* client's scrubber does not catch it: that scrubber redacts secret SHAPES
* (JWT, sk-/pk-/hk-, bearer, cloud keys, PAN) and an opaque authorization code
* is not one.
*
* So the gate is "do not emit from a route whose URL carries a credential", and
* the test below asserts BOTH halves: that gated routes emit nothing, and that
* the same setup ungated really does leak. The second half is what keeps this
* from decaying into a decorative assertion — if @hanzo/event ever stops putting
* the href on the wire, that case fails and this whole file can be revisited.
*/
import { test } from 'vitest'
import assert from 'node:assert/strict'
import fs from 'node:fs'
import path from 'node:path'
import { createAnalytics } from '@hanzo/event'
import { telemetryAllowed, consented } from './analytics'
// ── the route gate ──────────────────────────────────────────────────────────
test('auth-artifact routes are refused, funnel routes are not', () => {
// Carry a credential in the query string -> must never emit.
for (const p of [
'/callback',
'/callback/',
'/callback/anything',
'/login/oauth/device',
'/login/oauth/device/',
'/login/oauth/device/WDJB-MJHT',
]) {
assert.equal(telemetryAllowed(p), false, `${p} must not emit`)
}
// The funnel this exists to measure: arrival -> sign-in -> session.
for (const p of [
'/',
'/login',
'/login/',
'/signup',
'/forgot',
'/forget',
'/onboarding',
'/callbacks', // near-miss: a real route that merely starts the same way
'/login/oauth', // the device path is the specific one, not all of oauth
]) {
assert.equal(telemetryAllowed(p), true, `${p} must emit`)
}
})
/**
* The gate is written against literal paths, and App.tsx dispatches against its
* own. If someone adds a route that lands on the callback or device page, this
* fails rather than silently starting to ship codes — the same reason the token
* suite computes from what the bundle serves instead of trusting a list.
*/
test('every auth-artifact route App.tsx dispatches is covered by the gate', () => {
const app = fs.readFileSync(path.join(import.meta.dirname, 'App.tsx'), 'utf8')
// Route literals compared in App.tsx: path === '…' / path.startsWith('…').
const routes = [...app.matchAll(/path\s*(?:===\s*|\.startsWith\(\s*)'([^']+)'/g)].map((m) => m[1]!)
assert.ok(routes.length >= 10, `expected App.tsx route literals, found ${routes.length}`)
for (const r of routes) {
const isAuthArtifact = r.startsWith('/callback') || r.startsWith('/login/oauth/device')
if (isAuthArtifact) {
assert.equal(telemetryAllowed(r), false, `App.tsx routes ${r} to an auth-artifact page; gate it`)
}
}
// Both pages are actually reachable — the gate is not guarding dead routes.
assert.ok(routes.some((r) => r.startsWith('/callback')), 'App.tsx must route /callback')
assert.ok(
routes.some((r) => r.startsWith('/login/oauth/device')),
'App.tsx must route /login/oauth/device',
)
})
// ── consent ─────────────────────────────────────────────────────────────────
test('an explicit browser opt-out turns everything off', () => {
assert.equal(consented({ globalPrivacyControl: true }), false)
assert.equal(consented({ doNotTrack: '1' }), false)
assert.equal(consented({ doNotTrack: 'yes' }), false)
assert.equal(consented(), true)
assert.equal(consented({}), true)
assert.equal(consented({ globalPrivacyControl: false, doNotTrack: '0' }), true)
assert.equal(consented({ doNotTrack: null }), true)
})
// ── the wire ────────────────────────────────────────────────────────────────
const CODE = 'AUTHCODE_abc123XYZ'
const STATE = 'STATE_deadbeef'
const USER_CODE = 'WDJB-MJHT'
/** Installs the browser globals @hanzo/event reads, at a given location. */
function atLocation(href: string, pathname: string, search: string) {
const store: Record<string, string> = {}
const localStorage = {
getItem: (k: string) => store[k] ?? null,
setItem: (k: string, v: string) => void (store[k] = String(v)),
removeItem: (k: string) => void delete store[k],
}
const g = globalThis as Record<string, unknown>
g.window = {
location: { href, pathname, search, hostname: 'hanzo.id', origin: 'https://hanzo.id' },
addEventListener() {},
removeEventListener() {},
localStorage,
screen: { width: 1440, height: 900 },
}
g.document = {
referrer: '',
title: 'Sign in',
visibilityState: 'visible',
addEventListener() {},
removeEventListener() {},
}
g.localStorage = localStorage
g.screen = { width: 1440, height: 900 }
g.location = (g.window as { location: unknown }).location
}
function clearLocation() {
const g = globalThis as Record<string, unknown>
delete g.window
delete g.document
delete g.localStorage
delete g.screen
delete g.location
}
/** Runs the client exactly as mounted and returns everything it tried to send. */
function wireFrom(href: string, pathname: string, search: string, enabled: boolean): string {
atLocation(href, pathname, search)
try {
const sent: string[] = []
const client = createAnalytics({
product: 'id',
host: 'https://api.hanzo.ai',
ingestKey: 'pk-live-TESTKEY',
enabled,
transport: { send: (_url: string, body: string) => void sent.push(body) },
})
client.init()
client.pageview(pathname) // pathname only — the mitigation that is NOT enough
client.captureError(new Error('boom'))
client.flush()
return sent.join('')
} finally {
clearLocation()
}
}
test('a gated auth-artifact route puts nothing on the wire', () => {
const cb = wireFrom(
`https://hanzo.id/callback?code=${CODE}&state=${STATE}`,
'/callback',
`?code=${CODE}&state=${STATE}`,
telemetryAllowed('/callback'),
)
assert.equal(cb, '', 'the callback route must emit nothing at all')
assert.ok(!cb.includes(CODE), 'authorization code must never reach the wire')
assert.ok(!cb.includes(STATE), 'state must never reach the wire')
const dev = wireFrom(
`https://hanzo.id/login/oauth/device?user_code=${USER_CODE}`,
'/login/oauth/device',
`?user_code=${USER_CODE}`,
telemetryAllowed('/login/oauth/device'),
)
assert.equal(dev, '', 'the device route must emit nothing at all')
assert.ok(!dev.includes(USER_CODE), 'device user_code must never reach the wire')
})
/**
* The reason the gate exists. Passing a clean pathname is NOT what protects the
* code — if this ever stops leaking, @hanzo/event changed and the gate's
* justification should be re-read.
*/
test('without the gate, a clean pathname still leaks the code (why the gate exists)', () => {
const leaked = wireFrom(
`https://hanzo.id/callback?code=${CODE}&state=${STATE}`,
'/callback',
`?code=${CODE}&state=${STATE}`,
true, // ungated
)
assert.ok(leaked.includes(CODE), 'expected the ungated client to leak the code via `url`')
assert.ok(leaked.includes(STATE), 'expected the ungated client to leak the state via `url`')
assert.ok(leaked.includes('"path":"/callback"'), 'and to report a clean path while doing it')
})
test('funnel routes do report, and carry no credential', () => {
for (const p of ['/', '/login', '/signup', '/onboarding']) {
const wire = wireFrom(`https://hanzo.id${p}`, p, '', telemetryAllowed(p))
assert.ok(wire.includes('"$pageview"'), `${p} must report a pageview`)
assert.ok(wire.includes('"product":"id"'), `${p} must attribute to the id product`)
for (const secret of [CODE, STATE, USER_CODE]) {
assert.ok(!wire.includes(secret), `${p} must not carry ${secret}`)
}
}
})
test('an opted-out visitor emits nothing even on a funnel route', () => {
const wire = wireFrom(
'https://hanzo.id/login',
'/login',
'',
consented({ globalPrivacyControl: true }) && telemetryAllowed('/login'),
)
assert.equal(wire, '', 'GPC must suppress the whole client')
})
+138
View File
@@ -0,0 +1,138 @@
// Telemetry for the sign-in portal — pageviews and errors, anonymous, via the
// ONE @hanzo/event client (POST /v1/event, the front door cloud fans out into
// the web / product / error lenses). No page tag, no second SDK.
//
// This surface reported NOTHING before this file existed, which is why the
// arrival->session funnel had no denominator: hanzo.id is where every property's
// visitor lands, and none of it was attributable.
//
// It is also an AUTH surface, so what is NOT here is deliberate:
//
// - no identify(). Attribution here is anonymous; @hanzo/event stamps a
// per-browser `anonymousId` that survives sign-up, so the visitor's
// pre-signup pageviews still join to whoever they become once a
// post-auth surface (chat/console) identifies them. Reading the IAM
// subject would mean wiring this into the auth context for a join that
// already happens downstream.
// - no interaction autocapture (@hanzo/observe). Heat maps answer "where do
// they click"; the question this funnel exists to answer is "did they get a
// session", which pageviews answer completely. Autocapture on the login and
// signup forms is capture surface bought for no funnel signal.
// - no session replay, no input capture, no email/name.
import { useEffect, useState } from 'react'
import type { ReactNode } from 'react'
import { AnalyticsProvider as EventProvider, usePageview } from '@hanzo/event/react'
const HOST = 'https://api.hanzo.ai'
/**
* Publishable ingest key (pk-…), inlined by Vite from the build env. Write-only:
* it attributes a write to ONE org and mints no reading principal, which is what
* makes it safe in a bundle — and it is the ONLY thing that attributes a
* LOGGED-OUT visitor, which on a sign-in portal is nearly all of them.
*
* Absent is not a degraded mode: cloud takes an unkeyed beacon down the anonymous
* lane and files it under `$public`, a tenant this org cannot read, and answers
* 200 either way. The loss is silent on both ends, so the Dockerfile fails the
* build rather than letting an empty value ship. Never hardcode a value here.
*/
const INGEST_KEY = import.meta.env.VITE_PUBLISHABLE_KEY?.trim() || undefined
/**
* Routes whose URL carries an authentication artifact.
*
* `/callback` holds the OAuth authorization `code` and `state`; the device
* verification URI holds a `user_code`. Both sit in the QUERY STRING, and
* @hanzo/event stamps `url: window.location.href` onto every event it builds —
* independently of the `path` a caller passes. So passing a clean pathname does
* NOT keep the code out of the payload; only not emitting does. Measured against
* the real client, a pageview from `/callback?code=…&state=…` put both values on
* the wire in cleartext while `path` read a tidy `/callback`.
*
* The client's scrubber does not save this either — it redacts secret SHAPES
* (JWTs, sk-/pk-/hk-, bearer, cloud keys, PANs) and an opaque authorization code
* matches none of them.
*
* Neither route is a funnel step: both are transient machine hops that redirect
* onward within a tick. The funnel is `/` -> `/login` -> `/onboarding`, and every
* one of those still reports. Dropping these two costs no signal and removes the
* entire class of credential leak. See analytics.test.ts.
*/
const AUTH_ARTIFACT = /^\/(callback|login\/oauth\/device)(\/|$)/
/** telemetryAllowed reports whether a path may emit at all. Pure. */
export function telemetryAllowed(pathname: string): boolean {
return !AUTH_ARTIFACT.test(pathname)
}
/**
* consented honours an explicit browser opt-out — Global Privacy Control, then
* legacy Do-Not-Track. This is the whole consent surface, and it suppresses
* pageviews AND errors together: a visitor who opted out is not "mostly" off.
* Pure with respect to its argument so the policy is testable without a DOM.
*/
export function consented(nav?: {
globalPrivacyControl?: boolean
doNotTrack?: string | null
}): boolean {
if (!nav) return true
if (nav.globalPrivacyControl === true) return false
const dnt = nav.doNotTrack
return dnt !== '1' && dnt !== 'yes'
}
/** Reads the live opt-out signals off `navigator`, or none outside a browser. */
function browserConsent(): boolean {
if (typeof navigator === 'undefined') return true
return consented(navigator as Navigator & { globalPrivacyControl?: boolean })
}
/**
* Fires a pageview on SPA route changes.
*
* Inert today, and deliberately kept: this app navigates with
* `window.location.assign/replace`, so every route change is a fresh document
* and the provider's own initial pageview counts each page exactly once
* (`usePageview` skips its first mount for precisely that reason — it would
* otherwise double-count). `@tanstack/react-router` is a declared dependency
* that nothing imports; the day someone mounts it, navigation stops reloading
* the document and this is what keeps pageviews from silently going to zero.
*
* It is fed the PATHNAME, never `location.href`, and only when the path is
* allowed to emit — `usePageview` no-ops on a null path.
*/
function RouteViews() {
const [pathname, setPathname] = useState(() =>
typeof window === 'undefined' ? '/' : window.location.pathname,
)
useEffect(() => {
const sync = () => setPathname(window.location.pathname)
window.addEventListener('popstate', sync)
return () => window.removeEventListener('popstate', sync)
}, [])
usePageview(telemetryAllowed(pathname) ? pathname : null)
return null
}
/**
* Mounts the client. `enabled` is the single gate every plane reads — it stops
* init, enqueue, flush and the error handlers alike, so an off state emits
* nothing at all rather than emitting less.
*
* The gate is evaluated once per document, which is exact here BECAUSE
* navigation is full-page: the path a document is loaded at is the path it dies
* at, so there is no window in which a `/callback` load is measured under an
* earlier route's decision.
*/
export function Analytics({ children }: { children: ReactNode }) {
const pathname = typeof window === 'undefined' ? '/' : window.location.pathname
const enabled = browserConsent() && telemetryAllowed(pathname)
return (
<EventProvider config={{ product: 'id', host: HOST, ingestKey: INGEST_KEY, enabled }}>
<RouteViews />
{children}
</EventProvider>
)
}
+512 -125
View File
@@ -1,118 +1,415 @@
/* Hanzo ID — the styling layer for hanzo.id / lux.id / zoo.id / pars.id.
*
* TOKENS COME FROM @hanzo/design. Not one colour, radius or type size is
* invented here; every value below resolves to a design token, so a token change
* lands on all four brand portals AND on pay.hanzo.ai — the other half of the
* same sign-in-then-pay flow — at once.
*
* ARCHITECTURE RULE, learned the hard way. A component's surface must never
* depend on WHERE it is mounted. This file used to paint controls with the
* descendant selectors `form input {…}` and `.hanzo-id-btn, form button {…}`,
* so any control that escaped a <form> ancestor fell out of the stylesheet and
* rendered as raw UA chrome: the device-approval screen (2 inputs, 0 forms)
* showed a 31px beveled browser input next to correctly-styled 44px siblings.
* That is the same class of defect as a distributed component shipping utility
* class names with no CSS behind them. Every rule below is keyed to a CLASS the
* component itself carries — `.hanzo-id-input`, `.hanzo-id-btn`,
* `.hanzo-id-field`, `.hanzo-id-form` — and there are no element-descendant
* selectors for surface anywhere in this file.
*/
/* ONE import, the whole token layer. This file used to cherry-pick four of the
* nine token groups, which meant z, elevation, spacing, fonts and the element
* defaults simply did not exist here — and an absent group is invisible: an
* unresolved var() paints nothing and reports no error. @hanzo/iam's account
* menu alone reaches for --z-popover, --shadow-floating and --space-1..3, none
* of which the four-group subset carried.
*
* The reason for cherry-picking is gone: as of @hanzo/design 0.3.0 Geist is
* SELF-HOSTED inside the package (two variable woff2, SIL OFL-1.1), so
* tokens/fonts.css no longer makes a request to fonts.googleapis.com and the
* sign-in path can take the typeface along with the colours. */
@import '@hanzo/design/styles.css';
:root {
--brand: #ffffff;
--bg: #0a0a0a;
--fg: #fafafa;
--muted: #a3a3a3;
--border: #262626;
font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif;
color-scheme: dark;
}
* { box-sizing: border-box; }
html, body, #root { height: 100%; margin: 0; }
/* 100dvh, with 100vh left underneath it as the fallback for anything that does
not know the unit. On a phone `vh` resolves against the LARGEST viewport —
the one with the URL bar retracted — so a bar that is actually on screen
makes a "full height" page taller than the space it has, and a login page
with one card on it acquires a scrollbar and sits low in the window. `dvh`
is the height that is really there. */
body {
background: var(--bg);
color: var(--fg);
background: var(--background);
color: var(--foreground);
min-height: 100vh;
min-height: 100dvh;
display: flex;
flex-direction: column;
font-size: var(--text-base);
}
/* NO focus rule here any more — tokens/base.css ships
`:focus-visible{outline:2px solid var(--ring);outline-offset:2px}` to every
consumer, and as of 0.3.0 --ring is var(--neutral-500), which measures 4.43:1
on --background (WCAG 2.4.13 wants 3:1). The local override that painted the
ring --primary existed only because --ring was #333333 at 1.66:1; that was a
finding against @hanzo/design, it has been fixed there, so the workaround
goes. One focus indicator, defined once, in the design system. */
/* ── Page shell ────────────────────────────────────────────────────── */
.hanzo-id-page {
flex: 1;
/* #root is not a flex container, so `flex: 1` alone never stretches this to
the viewport — pin a min height so `main`'s justify-content:center has room
to vertically center the auth card (box-sizing:border-box folds in padding).
dvh for the reason given on `body`; the vh line is the fallback. */
min-height: 100vh;
min-height: 100dvh;
display: flex;
flex-direction: column;
max-width: 480px;
margin: 0 auto;
padding: 24px;
/* index.html declares `viewport-fit=cover`, which puts the page UNDER the
notch, the rounded corners and the home indicator. Declaring cover without
consuming the insets is strictly worse than not declaring it, and nothing
in this stylesheet consumed them: `max()` keeps the 24px gutter everywhere
it is already enough and only grows it where the hardware intrudes, so
nothing moves on a device with no insets (env() is 0px there). */
padding: max(24px, env(safe-area-inset-top)) max(24px, env(safe-area-inset-right))
max(24px, env(safe-area-inset-bottom)) max(24px, env(safe-area-inset-left));
}
.hanzo-id-brand-header {
padding: 16px 0 32px;
.hanzo-id-brand-header { padding: 16px 0 32px; }
/* The home link wraps a 32px mark but measured 32x18 — an inline <a> takes its
box from the LINE BOX of its own font, not from a replaced child, so the tap
target was SMALLER than the logo inside it. inline-flex gives the anchor the
mark's real box; the padding/negative-margin pair then grows the hit area to
the 44px floor while leaving the mark exactly where it was — the padding
offsets the margin, so the logo's painted position and the header's height
are both unchanged, and only the invisible target bleeds into the gutter. */
.hanzo-id-brand-header a {
display: inline-flex;
align-items: center;
box-sizing: content-box;
min-width: 32px;
min-height: 32px;
padding: 6px;
margin: -6px;
}
/* The lockup is 32px tall for EVERY brand, and it has to be said here.
BrandHeader writes `height={32}`, but that is a presentational attribute and
@hanzo/design's `:where(img,video){height:auto}` overrides it — so the logo
was sized by whatever the brand package happened to ship, not by this page.
The two assets differ in exactly the way that hides it: @hanzo/brand's SVG
carries a viewBox and no width/height, so it has no intrinsic size and landed
near the intended 32; @luxfi/brand's declares 1024x1024, so it took the full
column and rendered a 342px mark over lux.id's sign-in form. Same markup,
same CSS, opposite results, and only the brand nobody was looking at broke.
Pinning the height and letting width follow makes the header's own
declaration true again whatever a brand ships. */
.hanzo-id-brand-header img { height: 32px; width: auto; }
/* Text fallback when a brand ships no logo asset (see BrandHeader). */
.hanzo-id-wordmark {
font-size: 22px;
font-weight: 700;
letter-spacing: -0.02em;
color: var(--fg);
font-size: var(--text-xl);
font-weight: var(--weight-bold);
letter-spacing: var(--tracking-tight);
color: var(--text-primary);
}
.hanzo-id-page main {
flex: 1;
display: flex;
flex-direction: column;
justify-content: center;
gap: 24px;
}
.hanzo-id-page h1 { margin: 0; font-size: 28px; }
.hanzo-id-page .lede { color: var(--muted); margin: 0; }
form { display: flex; flex-direction: column; gap: 16px; }
form label { display: flex; flex-direction: column; gap: 6px; font-size: 14px; color: var(--muted); }
form input {
background: #111;
color: var(--fg);
border: 1px solid var(--border);
border-radius: 8px;
padding: 12px 14px;
font-size: 16px;
/* The type scale is the design scale, five rungs: xs / sm / base / xl / 2xl.
`h1` at --text-2xl is the SAME 24px as pay.hanzo.ai's `text-2xl` headings, so
a heading does not change size when the flow crosses between the two. */
.hanzo-id-page h1 {
margin: 0;
font-size: var(--text-2xl);
line-height: var(--leading-2xl);
letter-spacing: var(--tracking-tight);
}
form input:focus { outline: 2px solid var(--brand); outline-offset: -1px; }
.hanzo-id-page h2 { margin: 0; font-size: var(--text-xl); line-height: var(--leading-xl); }
.hanzo-id-page .lede { color: var(--muted-foreground); margin: 0; font-size: var(--text-base); }
.hanzo-id-btn, form button {
background: var(--fg);
color: var(--bg);
border: 0;
border-radius: 8px;
padding: 12px 16px;
font-size: 16px;
font-weight: 600;
/* ── Field ─────────────────────────────────────────────────────────── */
.hanzo-id-form { display: flex; flex-direction: column; gap: 16px; }
.hanzo-id-field {
display: flex;
flex-direction: column;
gap: 6px;
font-size: var(--text-sm);
color: var(--muted-foreground);
}
.hanzo-id-input {
/* `font: inherit` first: a bare <input>/<button> otherwise renders in the UA
face (Arial), which put two typefaces inside one 432px card — including on
the primary CTA. */
font: inherit;
font-size: var(--text-base);
background: var(--white-05);
color: var(--foreground);
/* A control's resting edge is --border-control. 0.4.2 cut it back to the alpha
ladder (.15) because a boundary that clears 3:1 on a near-black page IS a
mid-grey box, and a form of them reads as a wireframe. The contrast budget
went to --ring instead — the focus indicator is what a keyboard user
navigates by, it is the only boundary still held at 3:1, and unlike a
resting edge it is worth the loudness because it is transient. */
border: 1px solid var(--border-control);
/* 8px. 0.4.2 moved every control in the system to --radius-md, and tokens/
base.css draws a bare field at it; --radius-sm (6px) keeps its job on
genuinely small parts — badges, chips, menu rows — where 8px looks bubbly. */
border-radius: var(--radius-md);
padding: 0 14px;
min-height: 44px; /* the touch-target floor */
width: 100%;
}
.hanzo-id-input::placeholder { color: var(--text-disabled); }
.hanzo-id-input:disabled { opacity: 0.5; cursor: not-allowed; }
/* Password reveal. The button sits INSIDE the field's box rather than beside it,
so adding it costs no layout: the row is still one 44px control on one line,
at every width, and nothing below it moves. */
.hanzo-id-reveal { position: relative; display: flex; }
/* Reserve the button's width so a long password never runs under the eye. 14px
resting pad + 44px target = 58px, which is what the input gives back. */
.hanzo-id-reveal .hanzo-id-input { padding-right: 58px; }
.hanzo-id-revealbtn {
font: inherit;
position: absolute;
right: 0;
top: 0;
bottom: 0;
/* 44x44 — the same touch floor every other control here holds, and the reason
the button is sized rather than just padded: an icon button that is only as
big as its 20px glyph is a miss on a phone, which is the device this whole
feature is for. */
width: 44px;
display: flex;
align-items: center;
justify-content: center;
/* No box of its own: a second bordered surface inside the field would read as
a nested control. The glyph IS the affordance. */
background: none;
border: none;
border-radius: var(--radius-md);
padding: 0;
cursor: pointer;
color: var(--text-secondary);
transition: color 0.15s ease;
}
.hanzo-id-revealbtn:hover { color: var(--foreground); }
/* Pressed = the password is currently visible. Worth a permanent, non-hover
signal: the strongest cue that a credential is on screen should not depend on
where the pointer is, and on touch there is no hover at all. */
.hanzo-id-revealbtn[aria-pressed='true'] { color: var(--foreground); }
/* iOS Safari ZOOMS the viewport when a focused control's text is under 16px,
and it does not zoom back out — the user is left on a magnified, sideways-
scrolling sign-in page mid-credential. This rule used to be a COMMENT on
font-size above claiming `--text-base` prevented that; --text-base is
0.875rem = 14px, so the protection was never in effect on either credential
field. The type scale has no 16px rung and should not grow one — 16 is not a
design value here, it is the threshold in Safari's own zoom heuristic — so it
is written as the literal it is, and scoped to touch-primary pointers so the
desktop type ramp is untouched. */
@media (pointer: coarse) {
.hanzo-id-input { font-size: 16px; }
}
/* Checkbox: the UA control, sized up from its 13px intrinsic box and tinted with
the brand. Its touch target is the whole <label> row, which is why it is the
one control here that is not 44px itself. */
.hanzo-id-check {
font: inherit;
accent-color: var(--primary);
width: 18px;
height: 18px;
min-height: 0;
margin-top: 2px;
flex: none;
}
/* ── Button — ONE primitive, two modifiers ─────────────────────────── */
/* `.hanzo-id-btn` is filled (the primary action). `.ghost` is the secondary
surface — social sign-in, org rows, Skip/Back. `.row` spreads content for a
list row. There is no `.primary`: the base IS primary, so there is one and
only one way to write the default button. */
.hanzo-id-btn {
font: inherit;
font-size: var(--text-base);
font-weight: var(--weight-semibold);
display: inline-flex;
align-items: center;
justify-content: center;
gap: 10px;
background: var(--primary);
color: var(--primary-foreground);
border: 1px solid transparent;
/* Same 8px as the field it sits under. Two stacked controls in one 432px card
cannot round differently, and the system draws buttons at 8-10px too. */
border-radius: var(--radius-md);
padding: 10px 16px;
min-height: 44px;
cursor: pointer;
text-decoration: none;
text-align: center;
display: inline-block;
transition: background var(--duration-fast) var(--ease-out),
border-color var(--duration-fast) var(--ease-out);
}
.hanzo-id-btn[aria-disabled='true'], form button:disabled { opacity: 0.5; cursor: not-allowed; }
.hanzo-id-btn.primary { background: var(--brand); }
.hanzo-id-btn:disabled,
.hanzo-id-btn[aria-disabled='true'] { opacity: 0.5; cursor: not-allowed; }
.hanzo-id-btn svg { flex: none; }
/* The filled button had no hover at all. It has declared `transition:
background` since it was written and the system has shipped --primary-hover
the whole time, so the most-clicked control on the portal — Sign in, Continue,
Create account — was animating a property nothing ever changed. The ghost
variant below got its hover and this one was simply missed.
`:not(.ghost)` rather than relying on order: this selector and the ghost
hover both compute to (0,4,0), so with a bare `.hanzo-id-btn:hover` the two
would tie and source order would decide which surface a ghost button lifts
to. Saying which buttons are meant makes them unable to collide. */
.hanzo-id-btn:not(.ghost):hover:not(:disabled):not([aria-disabled='true']) {
background: var(--primary-hover);
}
.hanzo-id-btn.ghost {
background: var(--white-05);
color: var(--foreground);
border-color: var(--border-strong);
font-weight: var(--weight-medium);
}
/* Hover is a SURFACE lift, not a brighter edge. This used to also set
`border-color: var(--foreground)` — #ededed, 17.9:1 on --background — which
painted a near-solid-white 1px wireframe around the two most-hovered controls
on the page (Continue with GitHub / Continue with Google). Two problems with
moving the edge on hover, and the second is the one that matters:
1. It read as a wireframe, not as a refined surface. The house hairline is a
low-alpha step (--white-10/-15), ~4x quieter than #ededed.
2. Brightening an edge is THEME-HOSTILE. Every rung brighter than
--neutral-500 is worse on white: --neutral-400 measures 8.3:1 on black but
only 2.52:1 on white, which would fail the same WCAG 1.4.11 floor the
resting border is documented to hold in BOTH themes. There is no token
that gets brighter on dark and darker on light, so the edge must not
encode state at all.
So the border stays the constant control boundary (--border-strong, 4.43:1,
both themes) and the background carries the state — the lift the `transition`
on .hanzo-id-btn was already animating. */
.hanzo-id-btn.ghost:hover:not(:disabled) { background: var(--white-10); }
.hanzo-id-btn.row { justify-content: space-between; text-align: left; width: 100%; }
.hanzo-id-cta-row { display: flex; gap: 12px; }
.hanzo-id-footer-links { color: var(--muted); font-size: 14px; }
.hanzo-id-footer-links a { color: var(--fg); }
.hanzo-id-cta-row .hanzo-id-btn { flex: 1; }
.hanzo-id-linkbtn {
font: inherit;
background: none;
border: 0;
color: var(--text-primary);
font-size: var(--text-sm);
cursor: pointer;
text-align: left;
padding: 4px 0;
text-decoration: underline;
text-underline-offset: 4px;
}
.hanzo-id-footer-links { color: var(--muted-foreground); font-size: var(--text-sm); }
/* "Forgot password?" / "Create account" / "Sign in" / "Back to sign in" are the
only route changes on these pages that are not buttons, and they measured 17px
tall — the line box of their own text. Vertical padding on an INLINE box is
hit-tested but does not enter line-box height, so this buys the 44px target
with zero layout movement and no change to the sentences they sit inside.
(17 + 14 + 14 = 45.) The 14px bleed stays inside `main`'s 24px gap, so no two
targets overlap. Horizontal size already clears 44px on every one of them. */
.hanzo-id-footer-links a {
color: var(--text-primary);
padding: 14px 0;
margin: -14px 0;
}
/* A2P SMS consent disclosure (shown on phone/SMS surfaces). */
.hanzo-id-sms-consent { color: var(--muted); font-size: 12px; line-height: 1.5; }
.hanzo-id-sms-consent { color: var(--muted-foreground); font-size: var(--text-xs); line-height: var(--leading-relaxed); }
.hanzo-id-sms-consent p { margin: 0 0 6px; }
.hanzo-id-sms-consent-links { margin: 0; }
.hanzo-id-sms-consent a { color: var(--fg); }
.hanzo-id-sms-consent a { color: var(--text-primary); }
.hanzo-id-error {
background: #2d0a0a;
color: #ff7878;
border: 1px solid #5a1414;
/* ── Inline message ────────────────────────────────────────────────── */
/* Red is one of the two hues @hanzo/design permits, and these are the system's
own state tokens — the same #fca5a5 pay.hanzo.ai renders errors in. The
informational variant used to be a blue (#78b8ff) that exists nowhere in the
system; on a monochrome surface "information" is a neutral card. */
.hanzo-id-error,
.hanzo-id-info {
padding: 12px;
border-radius: 8px;
font-size: 14px;
border-radius: var(--radius-md);
font-size: var(--text-sm);
line-height: var(--leading-sm);
margin: 0;
}
.hanzo-id-error {
background: var(--state-error-bg);
color: var(--state-error-text);
border: 1px solid var(--state-error);
}
.hanzo-id-info {
background: #0a1f2d;
color: #78b8ff;
border: 1px solid #14385a;
padding: 12px;
border-radius: 8px;
background: var(--card);
color: var(--text-secondary);
border: 1px solid var(--white-10);
}
/* ── Device-authorization approval ─────────────────────────────── */
/* ── Loading ───────────────────────────────────────────────────────── */
/* `.hanzo-id-spinner` used to be a class with NO rule behind it: the loading
state measured 0px tall and was invisible on every portal. */
.hanzo-id-spinner {
width: 28px;
height: 28px;
margin: 0 auto;
border: 2px solid var(--white-15);
border-top-color: var(--primary);
border-radius: var(--radius-full);
animation: hanzo-id-spin 700ms linear infinite;
}
@keyframes hanzo-id-spin { to { transform: rotate(360deg); } }
/* ── Device-authorization approval ─────────────────────────────────── */
.hanzo-id-device main { gap: 18px; }
.hanzo-id-device-prompt { color: var(--muted); font-size: 14px; line-height: 1.5; margin: 0; }
.hanzo-id-device-prompt strong { color: var(--fg); }
.hanzo-id-device-code-field { display: flex; flex-direction: column; gap: 6px; }
.hanzo-id-device-code-field span { color: var(--muted); font-size: 13px; }
.hanzo-id-device-prompt {
color: var(--muted-foreground);
font-size: var(--text-sm);
line-height: var(--leading-relaxed);
margin: 0;
}
.hanzo-id-device-prompt strong { color: var(--text-primary); }
.hanzo-id-device-code {
font-family: ui-monospace, monospace;
font-size: 22px;
letter-spacing: 0.25em;
font-family: var(--font-mono);
font-size: var(--text-xl);
letter-spacing: var(--tracking-widest);
text-transform: uppercase;
}
.hanzo-id-device-confirm {
@@ -120,104 +417,194 @@ form input:focus { outline: 2px solid var(--brand); outline-offset: -1px; }
flex-direction: row;
align-items: flex-start;
gap: 10px;
color: var(--muted);
font-size: 13px;
line-height: 1.5;
}
.hanzo-id-device-confirm input {
width: auto;
margin-top: 2px;
flex: none;
color: var(--muted-foreground);
font-size: var(--text-sm);
line-height: var(--leading-relaxed);
}
/* ── Social / Web3 sign-in buttons ─────────────────────────────── */
/* ── Forced TOTP enrollment ────────────────────────────────────────── */
.hanzo-id-mfa-enroll { display: flex; flex-direction: column; gap: 16px; }
.hanzo-id-mfa-enroll h2 { margin: 0; font-size: var(--text-xl); }
.hanzo-id-mfa-qr {
align-self: center;
background: var(--pure-white);
padding: 12px;
border-radius: var(--radius-lg);
width: 220px;
height: 220px;
box-sizing: content-box;
}
.hanzo-id-mfa-qr svg { width: 100%; height: 100%; display: block; }
.hanzo-id-mfa-manual { font-size: var(--text-sm); color: var(--muted-foreground); }
.hanzo-id-mfa-manual summary { cursor: pointer; }
.hanzo-id-mfa-secret,
.hanzo-id-mfa-recovery code {
display: inline-block;
margin-top: 8px;
padding: 6px 10px;
background: var(--white-05);
border: 1px solid var(--white-10);
border-radius: var(--radius-sm);
font-family: var(--font-mono);
letter-spacing: var(--tracking-widest);
word-break: break-all;
}
.hanzo-id-mfa-recovery { font-size: var(--text-sm); color: var(--muted-foreground); line-height: var(--leading-relaxed); }
.hanzo-id-mfa-recovery code { letter-spacing: var(--tracking-normal); }
/* ── Social / Web3 sign-in ─────────────────────────────────────────── */
/* These are `.hanzo-id-btn.ghost` — there is no separate social-button surface. */
.hanzo-id-social { display: flex; flex-direction: column; gap: 10px; }
.hanzo-id-social-btn {
/* Wallet chain chooser — revealed under the single "Connect Wallet" button when
the injected chain is ambiguous. Indented so the EVM/Solana options read as
children of the wallet entry. */
.hanzo-id-wallet-chains {
display: flex;
align-items: center;
justify-content: center;
flex-direction: column;
gap: 10px;
background: #111;
color: var(--fg);
border: 1px solid var(--border);
border-radius: 8px;
padding: 11px 14px;
font-size: 15px;
font-weight: 500;
cursor: pointer;
margin-left: 4px;
padding-left: 12px;
border-left: 1px solid var(--white-10);
}
.hanzo-id-social-btn:hover { border-color: #3a3a3a; background: #161616; }
.hanzo-id-social-btn svg { flex: none; }
/* Labeled divider between social row and email form. */
.hanzo-id-divider {
display: flex;
align-items: center;
text-align: center;
color: var(--muted);
font-size: 13px;
color: var(--muted-foreground);
font-size: var(--text-sm);
}
.hanzo-id-divider::before,
.hanzo-id-divider::after {
content: '';
flex: 1;
height: 1px;
background: var(--border);
background: var(--white-10);
}
.hanzo-id-divider span { padding: 0 12px; }
/* ── Onboarding flow ───────────────────────────────────────────── */
/* ── Signed-in portal: the apps launcher ───────────────────────────── */
/* These rules used to be an inline `style={{…}}` object on Portal.tsx carrying
rgba(255,255,255,0.14), borderRadius 12, fontSize 13, fontWeight 600 and two
bare opacities — six invented values for facts the token layer already
states. They are classes now for the same reason as everything else in this
file: a surface must not depend on where it is mounted. */
.hanzo-id-portal main { width: 100%; max-width: 760px; }
.hanzo-id-apps {
display: grid;
grid-template-columns: repeat(auto-fill, minmax(210px, 1fr));
gap: var(--space-3);
margin-top: var(--space-6);
}
.hanzo-id-applink {
display: block;
padding: var(--space-4);
/* A tile whose affordance is carried by its label, not its edge — so this is
the decorative hairline rung, not the 3:1 control boundary. --white-15 is
the ladder rung the old rgba(255,255,255,0.14) was approximating. */
border: 1px solid var(--white-15);
border-radius: var(--radius-lg);
text-decoration: none;
color: inherit;
transition: border-color var(--duration-fast) var(--ease-out);
}
.hanzo-id-applink:hover { border-color: var(--border-strong); text-decoration: none; }
.hanzo-id-applink-name { display: flex; justify-content: space-between; font-weight: var(--weight-semibold); }
.hanzo-id-applink-name span[aria-hidden] { color: var(--text-tertiary); }
.hanzo-id-applink-desc { color: var(--muted-foreground); font-size: var(--text-sm); margin-top: var(--space-1); }
/* The account control is a control, not a bar: cap it so the trigger (and the
menu, which matches the trigger's width) reads at the size it does on every
other Hanzo surface instead of spanning the whole 760px column. */
.hanzo-id-portal-account { margin-top: var(--space-6); max-width: 260px; }
/* ── Onboarding flow ───────────────────────────────────────────────── */
.hanzo-id-onboarding { display: flex; flex-direction: column; gap: 24px; }
.hanzo-id-onboarding-head { display: flex; flex-direction: column; gap: 6px; }
.hanzo-id-onboarding-head h1 { margin: 0; font-size: 26px; }
.hanzo-id-onboarding-head h1,
.hanzo-id-onboarding-done h1 { margin: 0; font-size: var(--text-2xl); letter-spacing: var(--tracking-tight); }
.hanzo-id-onboarding-body { display: flex; flex-direction: column; gap: 16px; }
.hanzo-id-onboarding-done { display: flex; flex-direction: column; gap: 16px; }
.hanzo-id-onboarding-done h1 { margin: 0; font-size: 26px; }
.hanzo-id-stepdots { display: flex; gap: 8px; }
.hanzo-id-stepdots span {
flex: 1;
height: 4px;
border-radius: 2px;
background: var(--border);
border-radius: var(--radius-full);
background: var(--white-10);
}
.hanzo-id-stepdots span.on { background: var(--brand); }
.hanzo-id-stepdots span.on { background: var(--primary); }
.hanzo-id-org-list { list-style: none; margin: 0; padding: 0; display: flex; flex-direction: column; gap: 8px; }
.hanzo-id-org-row {
width: 100%;
display: flex;
align-items: center;
justify-content: space-between;
background: #111;
color: var(--fg);
border: 1px solid var(--border);
border-radius: 8px;
padding: 14px 16px;
font-size: 15px;
cursor: pointer;
text-align: left;
.hanzo-id-slug-preview {
color: var(--muted-foreground);
font-size: var(--text-sm);
margin: -8px 0 0;
font-family: var(--font-mono);
}
.hanzo-id-org-row:hover { border-color: #3a3a3a; background: #161616; }
.hanzo-id-org-slug { color: var(--muted); font-size: 13px; font-family: ui-monospace, monospace; }
.hanzo-id-linkbtn {
background: none;
border: 0;
color: var(--fg);
font-size: 14px;
cursor: pointer;
text-align: left;
padding: 4px 0;
text-decoration: underline;
}
.hanzo-id-slug-preview { color: var(--muted); font-size: 13px; margin: -8px 0 0; font-family: ui-monospace, monospace; }
.hanzo-id-onboarding-actions { display: flex; gap: 10px; flex-wrap: wrap; }
.hanzo-id-onboarding-actions .hanzo-id-btn { flex: 1; min-width: 120px; }
.hanzo-id-btn.ghost { background: transparent; color: var(--fg); border: 1px solid var(--border); }
.hanzo-id-btn.ghost:hover { border-color: #3a3a3a; }
.hanzo-id-summary { display: grid; grid-template-columns: auto 1fr; gap: 8px 16px; margin: 0; }
.hanzo-id-summary dt { color: var(--muted); font-size: 14px; }
.hanzo-id-summary dd { margin: 0; font-size: 14px; font-family: ui-monospace, monospace; }
.hanzo-id-summary dt { color: var(--muted-foreground); font-size: var(--text-sm); }
.hanzo-id-summary dd { margin: 0; font-size: var(--text-sm); font-family: var(--font-mono); }
/* Consent step */
.hanzo-id-consent { display: flex; flex-direction: column; gap: 16px; }
.hanzo-id-consent p { margin: 0; color: var(--muted-foreground); font-size: var(--text-sm); line-height: var(--leading-normal); }
.hanzo-id-consent-check {
display: flex;
align-items: flex-start;
gap: 10px;
padding: 12px;
border: 1px solid var(--border-control);
border-radius: var(--radius-lg);
cursor: pointer;
font-size: var(--text-sm);
}
.hanzo-id-consent-check input { margin-top: 2px; accent-color: var(--primary); }
/* Plan step — one card per catalog plan + the pay-as-you-go card */
.hanzo-id-plans { display: flex; flex-direction: column; gap: 10px; }
.hanzo-id-plan {
display: flex;
flex-direction: column;
align-items: flex-start;
gap: 4px;
min-height: 44px;
padding: 14px 16px;
border: 1px solid var(--border-control);
border-radius: var(--radius-lg);
background: var(--white-05);
color: var(--foreground);
font: inherit;
text-align: left;
cursor: pointer;
transition: background 0.15s ease;
}
.hanzo-id-plan:hover { background: var(--white-10); }
.hanzo-id-plan:disabled { opacity: 0.6; cursor: default; }
.hanzo-id-plan.popular { border-color: var(--border-selected); }
.hanzo-id-plan-badge {
align-self: flex-end;
margin: -4px 0 -18px;
padding: 2px 8px;
border-radius: var(--radius-full);
background: var(--primary);
color: var(--primary-foreground);
font-size: var(--text-xs);
font-weight: 600;
}
.hanzo-id-plans-empty { margin: 0; color: var(--muted-foreground); font-size: var(--text-sm); }
.hanzo-id-plan-name { font-weight: 600; }
.hanzo-id-plan-price { font-size: var(--text-sm); }
.hanzo-id-plan-price em { font-style: normal; color: var(--muted-foreground); }
.hanzo-id-plan-desc { color: var(--muted-foreground); font-size: var(--text-sm); }
+10 -1
View File
@@ -1,6 +1,7 @@
import { StrictMode } from 'react'
import { createRoot } from 'react-dom/client'
import { App } from './App'
import { Analytics } from './analytics'
import { registerProvider } from '@hanzo/id-idv'
import { createStubProvider } from '@hanzo/id-idv/providers/stub'
import './app.css'
@@ -12,6 +13,14 @@ const root = document.getElementById('root')
if (!root) throw new Error('#root missing')
createRoot(root).render(
<StrictMode>
<App />
{/*
Telemetry wraps App rather than living inside it: the pageview must be
recorded for the arrival itself, including the loads where `/config.json`
or the brand package fails and App renders nothing but an error. Those are
exactly the visits worth counting.
*/}
<Analytics>
<App />
</Analytics>
</StrictMode>,
)
+15 -6
View File
@@ -5,7 +5,7 @@
* visual essentials (name, logo, accent). The split-view login's marketing
* panel and the post-login apps launcher need richer, org-specific copy —
* ported verbatim from the frozen `legacy-nextjs` design (`staticBranding`
* content + `orgApps`). Keyed by `tenant.orgId` so it stays decoupled from
* content + `orgApps`). Keyed by `org.orgId` so it stays decoupled from
* hostname switches; unknown orgs fall back to `hanzo`.
*/
@@ -67,14 +67,23 @@ const MARKETING: Record<string, Marketing> = {
},
}
// The launcher lists PRODUCTS a person opens, not every host we run.
//
// Hanzo is three: App (build), Chat (talk), Cloud (the platform + its API).
// "Console" is NOT a fourth — it is Cloud's former name, and console.hanzo.ai
// now redirects to cloud.hanzo.ai, so listing both showed one product twice
// under two names and sent half the traffic through a redirect. It is gone;
// nothing here links to console.hanzo.ai.
//
// Analytics, Platform and Storage came out with it: s3.hanzo.ai answers a bare
// XML AccessDenied to a browser (it is an S3 API endpoint, not a page), and the
// other two are surfaces inside Cloud rather than products of their own. A
// launcher that lands you on an error page teaches people the tiles are broken.
const APPS: Record<string, readonly AppLink[]> = {
hanzo: [
{ name: 'Console', href: 'https://console.hanzo.ai', description: 'Observability & traces' },
{ name: 'App', href: 'https://hanzo.app', description: 'Build with AI' },
{ name: 'Chat', href: 'https://hanzo.chat', description: 'AI chat interface' },
{ name: 'Cloud', href: 'https://cloud.hanzo.ai', description: 'AI model API' },
{ name: 'Analytics', href: 'https://analytics.hanzo.ai', description: 'Web analytics' },
{ name: 'Platform', href: 'https://platform.hanzo.ai', description: 'PaaS deployments' },
{ name: 'Storage', href: 'https://s3.hanzo.ai', description: 'S3-compatible storage' },
{ name: 'Cloud', href: 'https://cloud.hanzo.ai', description: 'Models, compute & API' },
],
lux: [
{ name: 'Bridge', href: 'https://bridge.lux.network', description: 'Cross-chain bridge' },
+24 -59
View File
@@ -1,77 +1,42 @@
import { useEffect, useState } from 'react'
import type { BrandContract, TenantConfig } from '@hanzo/id-shared'
import { createIam, createAuthClient, decodeState } from '@hanzo/id-auth'
import type { BrandContract, OrgConfig } from '@hanzo/id-shared'
import { createIam } from '@hanzo/id-auth'
import { BrandHeader } from '../components/BrandHeader'
/**
* OAuth/OIDC callback.
* OAuth/OIDC callback — the portal's OWN PKCE return, and only that.
*
* Two kinds of return land here:
* 1. The portal's own OIDC PKCE return (password / SDK `signinRedirect`) —
* completed by the `@hanzo/iam` SDK's `handleCallback`, which reads back
* the exact PKCE verifier/state it stored.
* 2. A SOCIAL provider return (GitHub/Google), where `social.ts` sent the
* user out with a base64 `state` that encodes the original authorize
* request. We detect that, exchange the provider `code` at the IAM backend
* (`client.providerLogin`), and follow the continue-URL it returns — which
* re-enters this callback as case (1). (Pending live verification; only
* reachable once real provider creds are seeded.)
* One kind of return lands here: an IAM authorization code for a flow this
* portal started (password, wallet, or a federated provider begun through
* `signinRedirect`). The `@hanzo/iam` SDK's `handleCallback` completes it,
* reading back the exact PKCE verifier and state it stored.
*
* Routing after the OIDC exchange:
* - A downstream app left its target in `post_login_redirect` → forward tokens.
* There is no second, social-specific case. A federated sign-in returns from the
* IdP to IAM's OWN callback (`/v1/iam/oauth/callback`), which does the code
* exchange server-side and sends the browser back here with an ordinary IAM code
* — indistinguishable from any other. The page used to carry a branch that
* decoded a base64 provider `state` and posted the raw IdP code back to IAM; no
* endpoint ever accepted that, and nothing can produce that state any more.
*
* Routing after the exchange:
* - A non-OIDC "come back here" target left in `post_login_redirect` (device
* approval) → forward tokens there.
* - A bare portal sign-in → `/onboarding`.
*
* An app that sent the user here for a code never reaches this page at all: that
* flow re-enters IAM's authorize endpoint and IAM redirects straight to the app.
*/
/** Decode a social-provider `state` (URL-safe base64 of the authorize query). */
function decodeProviderState(state: string | null): URLSearchParams | null {
if (!state) return null
try {
const decoded = decodeState(state)
const params = new URLSearchParams(decoded.replace(/^\?/, ''))
// A provider-login state always carries application + provider markers.
if (params.get('provider') && params.get('application')) return params
} catch {
// not base64 → an SDK/OIDC state, not a provider return
}
return null
}
export function Callback({ tenant, brand }: { tenant: TenantConfig; brand: BrandContract }) {
export function Callback({ org, brand }: { org: OrgConfig; brand: BrandContract }) {
const [error, setError] = useState<string | null>(null)
useEffect(() => {
const search = new URLSearchParams(window.location.search)
const providerState = decodeProviderState(search.get('state'))
// Case (2): social provider return → exchange the provider code, then follow
// the continue-URL back into case (1).
if (providerState && search.get('code')) {
const client = createAuthClient({ tenant })
const oidcQuery = decodeState(search.get('state')!)
client
.providerLogin({
application: providerState.get('application') ?? '',
provider: providerState.get('provider') ?? '',
code: search.get('code') ?? '',
oidcQuery,
method: providerState.get('method') ?? 'signin',
})
.then((r) => {
if (r.redirectUrl) window.location.replace(r.redirectUrl)
else setError(r.error ?? 'Sign-in failed')
})
.catch((e) => setError(String(e)))
return
}
// Case (1): the portal's own OIDC PKCE return.
const iam = createIam(tenant)
const iam = createIam(org)
iam
.handleCallback(window.location.href)
.then((tok) => {
const target = sessionStorage.getItem('post_login_redirect')
sessionStorage.removeItem('post_login_redirect')
if (target) {
// Forward tokens to whichever app initiated this flow.
// Forward tokens to the page that sent the user to sign in.
const url = new URL(target, window.location.origin)
url.searchParams.set('access_token', tok.accessToken)
if (tok.refreshToken) url.searchParams.set('refresh_token', tok.refreshToken)
@@ -83,7 +48,7 @@ export function Callback({ tenant, brand }: { tenant: TenantConfig; brand: Brand
window.location.replace('/onboarding')
})
.catch((e) => setError(String(e)))
}, [tenant])
}, [org])
return (
<div className="hanzo-id-page hanzo-id-callback">
+109 -39
View File
@@ -1,23 +1,32 @@
import { useEffect, useState, type ReactNode } from 'react'
import type { BrandContract } from '@hanzo/id-shared'
import { LoginForm, SocialButtons, type AuthClient } from '@hanzo/id-auth'
import { LoginForm, SocialButtons, type AuthClient, type DeviceInfoResult } from '@hanzo/id-auth'
import { BrandHeader } from '../components/BrandHeader'
/**
* RFC 8628 device-authorization approval (`/login/oauth/device`).
*
* The terminal leg of `dev login --device-auth`: the CLI shows a short
* `user_code` and sends the human here (the IAM `verification_uri`;
* `verification_uri_complete` adds `?user_code=<code>`). The human signs in to
* the SAME issuer, confirms the code matches what their device shows, and
* approves — which flips the device code's `UserSignIn=true` so the CLI's token
* poll completes.
* The terminal leg of `hanzo login`: the CLI shows a short `user_code` and sends
* the human here (the IAM `verification_uri`; `verification_uri_complete`
* appends the code as a PATH segment, `/login/oauth/device/<code>` — IAM builds
* it that way because that is the route this page is registered on, and
* `readUserCode` accepts the `?user_code=` query form too). The human signs in
* to the SAME issuer, confirms the code matches what their device shows, and
* approves — which binds their identity onto the pending row (`Token.User`,
* owner/name) so the CLI's token poll stops answering `authorization_pending`
* and mints. There is no `UserSignIn` flag; an empty `User` IS "not yet
* approved".
*
* Auth is reused, never reimplemented: not-signed-in renders the normal
* `<LoginForm>` + `<SocialButtons>`; once the issuer session cookie is set the
* page reads it back from `/v1/iam/get-account` and shows the confirm step.
* Approval rides that session cookie (`client.approveDevice`), so no token ever
* touches the URL or logs.
*
* The screen exists to answer ONE question — which application am I authorizing?
* — so the application it names is read from the code (`client.deviceInfo`) and
* from nowhere else. Until IAM has named one there is no name on screen and no
* button to press.
*/
type Phase =
@@ -55,14 +64,33 @@ function scrubUrl() {
export function DeviceApproval({ client, brand }: { client: AuthClient; brand: BrandContract }) {
const [phase, setPhase] = useState<Phase>({ s: 'checking' })
const [userCode, setUserCode] = useState(() => readUserCode())
// Anti-phishing gate: the `?user_code=` prefill is DISPLAY-ONLY. A signed-in
// victim who lands here from a crafted `verification_uri_complete` link must
// NOT be able to approve an attacker's device with one click — they have to
// explicitly affirm the code matches the one their OWN device shows. The
// prefill cannot tick this box, so it can never auto-approve.
const [confirmed, setConfirmed] = useState(false)
// The code is prefilled from `?user_code=` and stays EDITABLE, which is the
// anti-phishing property that matters: approving is an explicit click on a
// code the human can read and correct against what their own device shows.
// There was also a "I started this sign-in" checkbox in front of that click.
// No device page anyone actually uses has one — Google, GitHub and AWS all
// show the code and an Approve button — and a tickbox is not evidence: a
// victim being walked through a crafted link ticks it as readily as they
// click Approve. It bought nothing and cost every honest user a step.
const [error, setError] = useState<string | null>(null)
const appLabel = client.tenant.appName
// WHICH application is asking — the whole point of this screen, and the one
// thing the page cannot know on its own. It used to render `org.appName`: this
// PORTAL's own branding, a static per-org string, so a sign-in started by
// hanzo-cli was approved under a screen reading "hanzo-console". The client is
// a property of the CODE (it lives on the pending row and is what the backend
// actually approves), so it is read from the code — `client.deviceInfo`,
// IAM `POST /v1/iam/oauth/device/info`.
//
// That read is session-gated and answers with one opaque refusal for unknown /
// expired / already-approved, so it is no oracle for hunting live codes: it
// tells a caller strictly less than the approval that same caller could already
// attempt.
//
// null = not resolved yet. NOTHING is rendered in its place — no fallback name,
// no portal name, no guess. Naming the wrong party is the defect being fixed
// here, and a screen that names none is strictly better than one that lies.
const [app, setApp] = useState<DeviceInfoResult | null>(null)
const named = app?.ok ? app : null
// Resolve the issuer session: signed in → confirm, else → sign-in form. Reads
// same-origin from `/v1/iam/get-account` (cookie session; the brand `*.id`
@@ -70,7 +98,7 @@ export function DeviceApproval({ client, brand }: { client: AuthClient; brand: B
useEffect(() => {
scrubUrl()
let alive = true
fetch(new URL('/v1/iam/get-account', client.tenant.iamUrl).toString(), {
fetch(new URL('/v1/iam/get-account', client.org.iamUrl).toString(), {
credentials: 'include',
headers: { Accept: 'application/json' },
})
@@ -90,7 +118,36 @@ export function DeviceApproval({ client, brand }: { client: AuthClient; brand: B
return () => {
alive = false
}
}, [client.tenant.iamUrl])
}, [client.org.iamUrl])
// Ask WHICH application the code belongs to. Needs both halves of what the
// endpoint is gated on: the issuer session (every phase past the check has one
// except `signin`, and the boolean keeps consent/approving from re-asking) and
// a code to ask about.
//
// The debounce is what makes a hand-typed code work: each keystroke is a
// different code, and a partial one is not a real code — without it the human
// watches IAM's refusal flash at them while they are still typing.
const signedIn = phase.s !== 'checking' && phase.s !== 'signin'
const blank = userCode.trim().length === 0
useEffect(() => {
if (!signedIn || blank) return
let alive = true
const t = setTimeout(() => {
client.deviceInfo(userCode).then((r) => {
if (!alive) return
// The session lapsed between the get-account check and this read. The
// signin phase preserves the code in `returnTo`, so the human lands back
// here with it intact.
if (!r.ok && r.loginRequired) setPhase({ s: 'signin' })
else setApp(r)
})
}, 250)
return () => {
alive = false
clearTimeout(t)
}
}, [client, userCode, signedIn, blank])
async function approve() {
setError(null)
@@ -145,18 +202,33 @@ export function DeviceApproval({ client, brand }: { client: AuthClient; brand: B
const busy = phase.s === 'approving'
const consent = phase.s === 'consent'
const email = phase.s === 'confirm' || phase.s === 'consent' ? phase.email : undefined
// ONE place shows a failure, whichever leg produced it: the approval itself, or
// the lookup that has to name an application before an approval is offered.
const failure = error ?? (app && !app.ok ? app.error : null)
return (
<Shell brand={brand}>
<h1>Approve this device</h1>
{email ? <p className="lede">Signed in as {email}</p> : null}
{/* The application is named ONLY once IAM has confirmed it — the clientId
alongside the display name, so a technical human can check it reads
`hanzo-cli` exactly and not something that merely looks like it. Until
then the sentence says a device, because that is all the page knows. */}
<p className="hanzo-id-device-prompt">
You are about to authorize <strong>{appLabel}</strong> to sign in on a device. Approve
ONLY if this code matches the one shown on that device.
{named ? (
<>
<strong>{named.displayName}</strong> (<code>{named.clientId}</code>) is asking to
sign in as you.
</>
) : (
'A device is asking to sign in as you.'
)}{' '}
Approve ONLY if the code below matches the one shown on that device, and only if you
started this sign-in yourself.
</p>
<label className="hanzo-id-device-code-field">
<label className="hanzo-id-field">
<span>Device code</span>
<input
type="text"
@@ -166,40 +238,38 @@ export function DeviceApproval({ client, brand }: { client: AuthClient; brand: B
spellCheck={false}
autoComplete="one-time-code"
aria-label="Device code"
className="hanzo-id-device-code"
className="hanzo-id-input hanzo-id-device-code"
value={userCode}
onChange={(e) => setUserCode(e.target.value)}
// A name — and a failure — belongs to a CODE. Edit the code and both are
// dropped in the same commit, so no name is ever left on screen for a
// frame beside a code it was not confirmed for.
onChange={(e) => {
setUserCode(e.target.value)
setApp(null)
setError(null)
}}
placeholder="e.g. K7M4P2QH"
disabled={busy}
/>
</label>
{consent ? (
{consent && named ? (
<p className="hanzo-id-info">
{brand.name} needs your consent to continue. By approving you grant {appLabel} access to
your profile.
<strong>{named.displayName}</strong> needs your consent to continue. By approving
you grant the device showing this code access to your profile.
</p>
) : null}
<label className="hanzo-id-device-confirm">
<input
type="checkbox"
checked={confirmed}
onChange={(e) => setConfirmed(e.target.checked)}
disabled={busy}
/>
<span>
I started this sign-in on my own device, and this code matches the one it shows.
</span>
</label>
{error ? <p role="alert" className="hanzo-id-error">{error}</p> : null}
{failure ? <p role="alert" className="hanzo-id-error">{failure}</p> : null}
<div className="hanzo-id-cta-row">
<button
type="button"
className="hanzo-id-btn primary"
disabled={busy || userCode.trim().length === 0 || !confirmed}
// Nothing is approved until IAM has named what is being approved. An
// unresolved or refused lookup leaves no button to press, rather than a
// button that authorizes an unnamed party.
className="hanzo-id-btn"
disabled={busy || !named}
onClick={approve}
>
{busy ? 'Approving' : consent ? 'Approve & grant access' : 'Approve'}
@@ -0,0 +1,40 @@
import { describe, expect, it } from 'vitest'
import { readFileSync } from 'node:fs'
import { join } from 'node:path'
/**
* The signup funnel has to reach signup.
*
* hanzo.app's "Get started" forwards `signup=true` on the authorize request and
* this page ignored it, so a net-new customer landed on "Sign in to Hanzo ID"
* with empty credentials and had to spot the small "Create account" link. The
* app was already doing its part; the IdP dropped the hint.
*
* Source text rather than a mount: the assertions are about which BRANCH exists
* and its ordering against silent SSO, and a render test would need the whole
* auth client stubbed to say much less.
*/
const src = readFileSync(join(__dirname, 'Login.tsx'), 'utf8')
describe('Login honors a registration hint', () => {
it('reads both spellings of the hint', () => {
expect(src).toMatch(/sp\.get\('signup'\)\s*===\s*'true'/)
// The OIDC-standard spelling, so a compliant client works without knowing ours.
expect(src).toMatch(/sp\.get\('screen_hint'\)\s*===\s*'signup'/)
})
it('sends the whole OIDC request across, or registration cannot return the user', () => {
// client_id, redirect_uri, state and the PKCE challenge live in the search
// string; a bare '/signup' strands the new account with nowhere to go back to.
expect(src).toMatch(/replace\(`\/signup\$\{window\.location\.search\}`\)/)
})
it('loses to silent SSO instead of short-circuiting it', () => {
// Someone whose browser already holds an issuer session HAS an account.
// The hint must sit in the fallback, so `canSilent` still wins the initial
// phase — sending a returning customer to registration is worse than
// ignoring the hint entirely.
expect(src).toMatch(/const fallback = providerHint \? 'federate' : wantsSignup \? 'register' : 'form'/)
expect(src).toMatch(/useState<[^>]*>\(\s*canSilent \? 'silent' : fallback,?\s*\)/)
})
})
+158 -12
View File
@@ -1,26 +1,66 @@
import { useEffect, useState } from 'react'
import type { BrandContract } from '@hanzo/id-shared'
import { LoginForm, SocialButtons, type AuthClient } from '@hanzo/id-auth'
import { idBrandLabel, type BrandContract } from '@hanzo/id-shared'
import {
LoginForm,
MfaEnrollForm,
OTPForm,
SocialButtons,
mfaChannelOf,
type AuthClient,
type LoginResponse,
} from '@hanzo/id-auth'
import { BrandHeader } from '../components/BrandHeader'
import { clientIdFrom } from '../route'
export function Login({ client, brand }: { client: AuthClient; brand: BrandContract }) {
const sp = new URLSearchParams(window.location.search)
const redirectUri = sp.get('redirect_uri') ?? undefined
const state = sp.get('state') ?? undefined
const clientIdOverride = sp.get('client_id') ?? undefined
const clientIdOverride = clientIdFrom(window.location.search, window.location.pathname)
const codeChallenge = sp.get('code_challenge') ?? undefined
const codeChallengeMethod = (sp.get('code_challenge_method') as 'S256' | 'plain' | null) ?? undefined
const nonce = sp.get('nonce') ?? undefined
// A provider the user already chose upstream (the console sends
// `?provider_hint=provider-github` when they click "Continue with GitHub"
// over there). With no live session we launch that provider straight away
// instead of showing this form — so the click lands directly in the social
// flow, never bouncing the user to a second login page. We honor ONLY
// `provider_hint`, never a bare `provider=` (the SSO SDK uses that for its
// `<org>-iam` IDP hint — a different meaning).
const providerHint = sp.get('provider_hint') ?? undefined
// TRUE single sign-on. When an app sent the user here for an authorization
// code (client_id + redirect_uri present) AND the browser already holds an
// issuer session from an earlier sign-in (the `iam_session_id` cookie), mint
// the code from that session and redirect straight back — no form, no
// credential re-entry. Only fall back to the interactive form when there is
// no live session. A bare portal visit (no client_id/redirect_uri) has
// nowhere to redirect, so it shows the form immediately as before.
// credential re-entry. With no live session we fall back to auto-launching the
// hinted provider if one was named, else the interactive form. A bare portal
// visit (no client_id/redirect_uri) has nowhere to redirect, so it shows the
// form immediately as before.
// A caller that sent the user here to REGISTER should get registration.
// hanzo.app's "Get started" forwards `signup=true` and this page ignored it,
// so every net-new customer met a sign-in form with empty credentials and had
// to notice the small "Create account" link to get past it — the signup funnel
// never reached signup. `screen_hint=signup` is the OIDC-standard spelling of
// the same request, so both are honored.
//
// It LOSES to silent SSO, deliberately. A browser already holding an issuer
// session belongs to someone who has an account, and sending them to
// registration would be worse than ignoring the hint. So it is the fallback
// when there is no session, never a short-circuit ahead of one.
const wantsSignup = sp.get('signup') === 'true' || sp.get('screen_hint') === 'signup'
const canSilent = !!clientIdOverride && !!redirectUri
const [phase, setPhase] = useState<'silent' | 'form'>(canSilent ? 'silent' : 'form')
const fallback = providerHint ? 'federate' : wantsSignup ? 'register' : 'form'
const [phase, setPhase] = useState<'silent' | 'federate' | 'form' | 'register'>(
canSilent ? 'silent' : fallback,
)
// null = show the credential form; otherwise IAM returned an MFA signal and
// we render the matching step instead of navigating on.
const [mfa, setMfa] = useState<LoginResponse | null>(null)
const [challengeError, setChallengeError] = useState<string | null>(null)
const clientId = clientIdOverride ?? client.org.clientId
useEffect(() => {
if (!canSilent) return
@@ -40,11 +80,11 @@ export function Login({ client, brand }: { client: AuthClient; brand: BrandContr
if (r.redirectUrl) {
window.location.assign(r.redirectUrl)
} else {
setPhase('form')
setPhase(fallback)
}
})
.catch(() => {
if (!cancelled) setPhase('form')
if (!cancelled) setPhase(fallback)
})
return () => {
cancelled = true
@@ -53,7 +93,36 @@ export function Login({ client, brand }: { client: AuthClient; brand: BrandContr
// eslint-disable-next-line react-hooks/exhaustive-deps
}, [])
if (phase === 'silent') {
// The credential check succeeded (or MFA was satisfied). For a downstream
// OIDC request, re-enter authorize with the now-established IAM session so it
// mints the code; for a bare portal sign-in, land on onboarding.
function completeAfterAuth() {
if (redirectUri) {
window.location.href = client.authorize({
clientId,
redirectUri,
state: state ?? '',
codeChallenge,
codeChallengeMethod,
})
} else {
window.location.href = '/onboarding'
}
}
// Registration lives on its own page, so this is a real navigation rather than
// a branch in the render. `replace`, not `assign`: Back from the signup form
// must return to whatever sent the user here, not to a login page that would
// immediately bounce forward again. The whole search string travels — the
// client_id, redirect_uri, state and PKCE challenge are what let registration
// return the new user to the app that asked for them.
useEffect(() => {
if (phase === 'register') {
window.location.replace(`/signup${window.location.search}`)
}
}, [phase])
if (phase === 'silent' || phase === 'register') {
return (
<div className="hanzo-id-page hanzo-id-login">
<BrandHeader brand={brand} />
@@ -64,11 +133,82 @@ export function Login({ client, brand }: { client: AuthClient; brand: BrandContr
)
}
// Auto-launch the hinted provider. `SocialButtons` is headless here — it
// resolves the app config and runs the hop; we show a busy state meanwhile,
// and drop to the form only if the hint matched no configured provider.
if (phase === 'federate') {
return (
<div className="hanzo-id-page hanzo-id-login">
<BrandHeader brand={brand} />
<main aria-busy="true">
<p>Signing you in</p>
<SocialButtons
client={client}
clientIdOverride={clientIdOverride}
intent="signin"
postLoginRedirect={redirectUri}
autoStart={providerHint}
onAutoStartResolved={(started) => {
if (!started) setPhase('form')
}}
/>
</main>
</div>
)
}
if (mfa?.mfaStage === 'enroll') {
return (
<div className="hanzo-id-page hanzo-id-login">
<BrandHeader brand={brand} />
<main>
<MfaEnrollForm client={client} onComplete={completeAfterAuth} />
</main>
</div>
)
}
if (mfa?.mfaStage === 'challenge') {
const iamType = mfa.mfaTypes?.[0] ?? 'app'
async function onChallenge(code: string) {
setChallengeError(null)
const res = await client.mfaChallenge({
mfaType: iamType,
passcode: code,
clientId,
application: client.org.appName,
organization: client.org.orgId,
redirectUri,
state,
codeChallenge,
codeChallengeMethod,
})
if (res.error) {
setChallengeError(res.error)
} else if (res.redirectUrl) {
window.location.href = res.redirectUrl
} else {
completeAfterAuth()
}
}
return (
<div className="hanzo-id-page hanzo-id-login">
<BrandHeader brand={brand} />
<main>
<h1>Two-factor authentication</h1>
<p className="lede">Enter the code from your authenticator app to finish signing in.</p>
{challengeError ? <p role="alert" className="hanzo-id-error">{challengeError}</p> : null}
<OTPForm channel={mfaChannelOf(iamType)} onSubmit={onChallenge} />
</main>
</div>
)
}
return (
<div className="hanzo-id-page hanzo-id-login">
<BrandHeader brand={brand} />
<main>
<h1>Sign in to {brand.name}</h1>
<h1>Sign in to {idBrandLabel(brand, client.org.orgId)}</h1>
<SocialButtons
client={client}
clientIdOverride={clientIdOverride}
@@ -83,9 +223,15 @@ export function Login({ client, brand }: { client: AuthClient; brand: BrandContr
codeChallenge={codeChallenge}
codeChallengeMethod={codeChallengeMethod}
nonce={nonce}
onMfaRequired={setMfa}
/>
<p className="hanzo-id-footer-links">
<a href="/forget">Forgot password?</a> · <a href="/signup">Create account</a>
{/* Carry the OIDC request across. These are full page loads, so a bare
href drops the client_id, redirect_uri, state and PKCE challenge the
app sent — and registration then has nothing to return the new user
to. `Signup` reads exactly these params. */}
<a href={`/forget${window.location.search}`}>Forgot password?</a> ·{' '}
<a href={`/signup${window.location.search}`}>Create account</a>
</p>
</main>
</div>
+98 -23
View File
@@ -1,15 +1,19 @@
import { useMemo } from 'react'
import type { BrandContract, TenantConfig } from '@hanzo/id-shared'
import { useEffect, useMemo, useState } from 'react'
import type { BrandContract, OrgConfig } from '@hanzo/id-shared'
import { createIam } from '@hanzo/id-auth'
import { OnboardingFlow, createOnboardingService, type OnboardingState } from '@hanzo/id-onboarding'
import { getConnector } from '@hanzo/id-connect/connectors'
import { BrandHeader } from '../components/BrandHeader'
/** Fleet default pay origin; a white-label brand overrides via catalog `payUrl`. */
const DEFAULT_PAY_URL = 'https://pay.hanzo.ai'
/**
* Post-login onboarding page.
*
* Reached after a bare portal sign-in (no downstream `redirect_uri`). Mounts
* the `@hanzo/id-onboarding` flow (org → project → wallet) wired to:
* the `@hanzo/id-onboarding` flow (org → project → wallet → consent → plan)
* wired to:
*
* - the IAM session token: read from the same `@hanzo/iam` PKCE client the
* Callback stored it on, so the onboarding writes ride the logged-in
@@ -18,39 +22,110 @@ import { BrandHeader } from '../components/BrandHeader'
* the onboarding pkg stays wallet-agnostic. Absent injected provider →
* the wallet step is skip-only.
*
* On completion it lands on the portal home (`/`); a downstream app that
* wanted a token would have carried `redirect_uri` and never reached here.
* NEVER REPEATS: completion is recorded on the USER (Properties, via
* saveOnboarding) — so before mounting the flow this page reads it back and,
* if the user already finished onboarding on ANY browser, goes straight to
* the portal. The read failing open (network blip → run the flow again) is
* deliberate: repeating is annoying, silently skipping a required step is
* worse.
*
* On completion it routes by the plan choice — the platform is prepay-only,
* so a plan goes to the pay cart and pay-as-you-go goes to the top-up flow.
* A downstream app that wanted a token would have carried `redirect_uri` and
* never reached here.
*/
export function Onboarding({ tenant, brand }: { tenant: TenantConfig; brand: BrandContract }) {
const iam = useMemo(() => createIam(tenant), [tenant])
export function Onboarding({ org, brand }: { org: OrgConfig; brand: BrandContract }) {
const iam = useMemo(() => createIam(org), [org])
const payUrl = org.payUrl || DEFAULT_PAY_URL
const service = useMemo(
() =>
createOnboardingService({
iamUrl: tenant.iamUrl,
orgId: tenant.orgId,
iamUrl: org.iamUrl,
orgId: org.orgId,
getAccessToken: () => iam.getValidAccessToken(),
}),
[tenant, iam],
[org, iam],
)
function onComplete(_state: OnboardingState) {
// Land on the authenticated portal (apps launcher), NOT the bare hero.
// The marker makes the portal treat the just-established session as authed
// even before the cross-request get-account read settles.
window.location.replace('/?signed_in=1')
// null = still checking; false = run the flow; true = already done, leaving.
const [alreadyDone, setAlreadyDone] = useState<boolean | null>(null)
useEffect(() => {
let alive = true
;(async () => {
// TOKEN BOOTSTRAP. A password/form sign-in mints a SESSION COOKIE and
// lands here directly — the PKCE SDK holds no token, so every onboarding
// WRITE (update-user needs a bearer; the cookie alone is refused, and
// rightly — a cookie-authed write is a CSRF surface) answered 401 and
// the funnel dead-ended at consent. With a live session, authorize is
// the silent-SSO branch: signinRedirect bounces through IAM with no UI,
// /callback stores the token and returns to /onboarding. One bounce per
// session, guarded, so a broken mint degrades to the read-only 401
// instead of a redirect loop.
const token = await iam.getValidAccessToken().catch(() => null)
if (!alive) return
if (!token) {
const guard = 'onboarding.token_bounce'
if (!sessionStorage.getItem(guard)) {
sessionStorage.setItem(guard, '1')
void iam.signinRedirect()
return
}
} else {
sessionStorage.removeItem('onboarding.token_bounce')
}
try {
const { completedAt } = await service.readOnboarding()
if (!alive) return
if (completedAt) {
setAlreadyDone(true)
window.location.replace('/?signed_in=1')
return
}
} catch {
// Read failing open (network blip → run the flow) is deliberate.
}
if (alive) setAlreadyDone(false)
})()
return () => {
alive = false
}
}, [service, iam])
function onComplete(state: OnboardingState) {
// Prepay-only funnel: the last step recorded a choice, now act on it.
// - a plan slug → the pay cart, seats + payment there (price is the
// catalog's — commerce recomputes server-side, the slug is enough)
// - pay as you go → the top-up flow ($5 minimum, all methods)
// The plan choice is already persisted on the user, so bouncing off the
// payment page never re-enters onboarding.
const choice = state.planChoice
if (choice === 'payg') {
window.location.replace(`${payUrl}/onboard`)
} else if (choice) {
window.location.replace(`${payUrl}/cart?plan=${encodeURIComponent(choice)}`)
} else {
// No recorded choice (should not happen — the plan step requires one):
// land on the authenticated portal rather than a dead end.
window.location.replace('/?signed_in=1')
}
}
return (
<div className="hanzo-id-page hanzo-id-onboarding-page">
<BrandHeader brand={brand} />
<main>
<OnboardingFlow
service={service}
brandName={brand.name}
connectWallet={connectInjectedWallet}
onComplete={onComplete}
/>
{alreadyDone === false ? (
<OnboardingFlow
service={service}
brandName={brand.name}
connectWallet={connectInjectedWallet}
onComplete={onComplete}
payUrl={payUrl}
/>
) : (
<div className="hanzo-id-spinner" aria-label="Loading" />
)}
</main>
</div>
)
@@ -60,8 +135,8 @@ export function Onboarding({ tenant, brand }: { tenant: TenantConfig; brand: Bra
* EVM wallet connector backed by @hanzo/id-connect (EIP-6963 multi-injection,
* viem under the hood). Returns the checksummed 0x address, or null when the
* user cancels or no injected EVM wallet is present. The onboarding wallet step
* only needs the address (it stores it via update-user?columns=web3onboard), so
* we connect and return account.address — no signature round-trip here.
* only needs the address (it stores it via a full-row read-merge-write in the
* service), so we connect and return account.address — no signature round-trip.
*/
async function connectInjectedWallet(): Promise<string | null> {
try {
+43 -46
View File
@@ -1,5 +1,7 @@
import { useEffect, useState } from 'react'
import type { BrandContract, TenantConfig } from '@hanzo/id-shared'
import type { IamIdentity } from '@hanzo/iam/react'
import { UserMenu, resolveIdentity } from '@hanzo/iam/react'
import type { BrandContract, OrgConfig } from '@hanzo/id-shared'
import type { AuthClient } from '@hanzo/id-auth'
import { Login } from './Login'
import { BrandHeader } from '../components/BrandHeader'
@@ -8,7 +10,7 @@ import { appsFor, billingFor } from '../marketing'
type Auth =
| { s: 'loading' }
| { s: 'anon' }
| { s: 'authed'; name?: string; email?: string }
| { s: 'authed'; identity: IamIdentity | null }
/**
* Root portal (`/`). The portal IS the login surface, not a marketing hero:
@@ -19,7 +21,7 @@ type Auth =
* - signed in → the apps launcher (the org's apps) + billing / sign-out.
*
* Auth is read same-origin from `/v1/iam/get-account` (cookie session;
* `tenant.iamUrl` is the brand's own `*.id` host, so this is first-party and
* `org.iamUrl` is the brand's own `*.id` host, so this is first-party and
* the session cookie rides along). The `?signed_in=1` marker set by the
* bare-login / onboarding-complete redirect is the authoritative "just
* authenticated" signal when the cookie read hasn't propagated yet.
@@ -27,18 +29,18 @@ type Auth =
export function Portal({
client,
brand,
tenant,
org,
}: {
client: AuthClient
brand: BrandContract
tenant: TenantConfig
org: OrgConfig
}) {
const [auth, setAuth] = useState<Auth>({ s: 'loading' })
useEffect(() => {
let alive = true
const justSignedIn = new URLSearchParams(window.location.search).get('signed_in') === '1'
fetch(new URL('/v1/iam/get-account', tenant.iamUrl).toString(), {
fetch(new URL('/v1/iam/get-account', org.iamUrl).toString(), {
credentials: 'include',
headers: { Accept: 'application/json' },
})
@@ -47,23 +49,26 @@ export function Portal({
if (!alive) return
const d = b.data as Record<string, unknown> | undefined
if (b.status === 'ok' && d && typeof d === 'object') {
setAuth({ s: 'authed', name: str(d.displayName) ?? str(d.name), email: str(d.email) })
// `resolveIdentity` is the SAME name/avatar/initials resolution every
// Hanzo surface shows, so the portal cannot disagree with the console
// about who you are — and it never falls back to a raw uuid.
setAuth({ s: 'authed', identity: resolveIdentity(d, {}) })
} else {
setAuth(justSignedIn ? { s: 'authed' } : { s: 'anon' })
setAuth(justSignedIn ? { s: 'authed', identity: null } : { s: 'anon' })
}
})
.catch(() => {
if (alive) setAuth(justSignedIn ? { s: 'authed' } : { s: 'anon' })
if (alive) setAuth(justSignedIn ? { s: 'authed', identity: null } : { s: 'anon' })
})
return () => {
alive = false
}
}, [tenant.iamUrl])
}, [org.iamUrl])
if (auth.s === 'loading') {
return (
<div className="hanzo-id-page" style={{ minHeight: '40vh' }}>
<div className="hanzo-id-spinner" style={{ borderTopColor: brand.accentColor ?? '#fff' }} />
<div className="hanzo-id-spinner" style={{ borderTopColor: brand.accentColor ?? 'var(--primary)' }} />
</div>
)
}
@@ -72,54 +77,46 @@ export function Portal({
if (auth.s === 'anon') return <Login client={client} brand={brand} />
// Signed in: the apps launcher.
const apps = appsFor(tenant.orgId)
const billingUrl = billingFor(tenant.orgId)
const logoutUrl = client.logout(undefined, `${tenant.publicOrigin}/login`)
const apps = appsFor(org.orgId)
const billingUrl = billingFor(org.orgId)
// signOut, not logout: the latter only builds the IdP URL and leaves this
// browser's `hanzo_iam_*` keys in place, so the token string outlived the
// session it named. Called on click, not at render — it clears storage.
return (
<div className="hanzo-id-page hanzo-id-portal">
<BrandHeader brand={brand} />
<main style={{ width: '100%', maxWidth: 760 }}>
<main>
<h1>Your {brand.name} apps</h1>
{auth.email ? <p className="lede">{auth.email}</p> : null}
<div
style={{
display: 'grid',
gridTemplateColumns: 'repeat(auto-fill, minmax(210px, 1fr))',
gap: 12,
marginTop: 24,
}}
>
<div className="hanzo-id-apps">
{apps.map((a) => (
<a
key={a.name}
href={a.href}
style={{
display: 'block',
padding: '16px 18px',
border: '1px solid rgba(255,255,255,0.14)',
borderRadius: 12,
textDecoration: 'none',
color: 'inherit',
}}
>
<div style={{ fontWeight: 600, display: 'flex', justifyContent: 'space-between' }}>
<a key={a.name} className="hanzo-id-applink" href={a.href}>
<div className="hanzo-id-applink-name">
<span>{a.name}</span>
<span aria-hidden style={{ opacity: 0.5 }}></span>
<span aria-hidden></span>
</div>
<div style={{ opacity: 0.6, fontSize: 13, marginTop: 4 }}>{a.description}</div>
<div className="hanzo-id-applink-desc">{a.description}</div>
</a>
))}
</div>
<div style={{ marginTop: 28, display: 'flex', gap: 18 }}>
<a className="hanzo-id-linkbtn" href={billingUrl}>Billing</a>
<a className="hanzo-id-linkbtn" href={logoutUrl}>Sign out</a>
{/* The ONE account control. This was a hand-rolled "Billing / Sign out"
link row; every Hanzo surface mounts @hanzo/iam's UserMenu instead,
so identity, billing and sign-out read and behave identically here,
on hanzo.chat and in the console. The portal's session is its own
cookie read rather than an IamProvider, which is exactly what the
`identity` / `isAuthenticated` / `onSignOut` overrides are for.
No `brand` prop: omitting `markSvg` would put the HANZO mark on
lux.id and zoo.id, and this one image serves all four portals. */}
<div className="hanzo-id-portal-account">
<UserMenu
identity={auth.identity}
isAuthenticated
usageUrl={billingUrl}
usageLabel="Billing"
onSignOut={() => { window.location.href = client.signOut(`${org.publicOrigin}/login`) }}
/>
</div>
</main>
</div>
)
}
function str(v: unknown): string | undefined {
return typeof v === 'string' && v.length > 0 ? v : undefined
}
+69 -3
View File
@@ -1,12 +1,69 @@
import { useEffect, useState } from 'react'
import type { BrandContract } from '@hanzo/id-shared'
import { SignupForm, SocialButtons, type AuthClient } from '@hanzo/id-auth'
import { BrandHeader } from '../components/BrandHeader'
import { clientIdFrom } from '../route'
export function Signup({ client, brand }: { client: AuthClient; brand: BrandContract }) {
const sp = new URLSearchParams(window.location.search)
const inviteCode = sp.get('invite') ?? undefined
const clientIdOverride = sp.get('client_id') ?? undefined
const clientIdOverride = clientIdFrom(window.location.search, window.location.pathname)
const redirectUri = sp.get('redirect_uri') ?? undefined
// The same downstream OIDC request `Login` reads. Registration ends in a
// sign-in, so it needs the whole request — not just the client and its
// callback — or the minted code carries no PKCE binding and no state.
const state = sp.get('state') ?? undefined
const codeChallenge = sp.get('code_challenge') ?? undefined
const codeChallengeMethod = (sp.get('code_challenge_method') as 'S256' | 'plain' | null) ?? undefined
const nonce = sp.get('nonce') ?? undefined
// Only ask for credentials an account can actually be made with. IAM refuses
// signup on an app with `enableSignUp:false` — 48 of 51 hanzo applications
// today — and it refuses at SUBMIT, so this page used to take an email and a
// password and only then answer "the application does not allow to sign up
// new account". The provider buttons dead-end the same way: federation
// PROVISIONS a local user for a new identity, so a first-time GitHub sign-up
// hits the same gate after a whole round trip through GitHub.
//
// `enableSignUp` is already on `AppLogin` and already fetched — SignupForm
// reads the same row inside onSubmit. Reading it here instead is what turns
// the refusal from a surprise into a state.
//
// It FAILS OPEN, deliberately: an unreadable app config renders the form, and
// the server still refuses. This is honesty about a known answer, not a gate —
// the gate is `internal/oidc/signup.go` and must stay the only one.
const [open, setOpen] = useState(true)
useEffect(() => {
let cancelled = false
client
.getAppLogin(clientIdOverride, redirectUri)
.then((app) => {
if (!cancelled && app) setOpen(app.enableSignUp)
})
.catch(() => {})
return () => {
cancelled = true
}
}, [client, clientIdOverride, redirectUri])
if (!open) {
return (
<div className="hanzo-id-page hanzo-id-signup">
<BrandHeader brand={brand} />
<main>
<h1>Create your {brand.name} account</h1>
<p className="hanzo-id-info">
This application does not accept new accounts. If you already have
one, sign in below.
</p>
<p className="hanzo-id-footer-links">
<a href={`/login${window.location.search}`}>Sign in</a>
</p>
</main>
</div>
)
}
return (
<div className="hanzo-id-page hanzo-id-signup">
<BrandHeader brand={brand} />
@@ -18,9 +75,18 @@ export function Signup({ client, brand }: { client: AuthClient; brand: BrandCont
intent="signup"
postLoginRedirect={redirectUri}
/>
<SignupForm client={client} inviteCode={inviteCode} />
<SignupForm
client={client}
inviteCode={inviteCode}
clientIdOverride={clientIdOverride}
redirectUri={redirectUri}
state={state}
codeChallenge={codeChallenge}
codeChallengeMethod={codeChallengeMethod}
nonce={nonce}
/>
<p className="hanzo-id-footer-links">
Already have an account? <a href="/login">Sign in</a>
Already have an account? <a href={`/login${window.location.search}`}>Sign in</a>
</p>
</main>
</div>
+36
View File
@@ -0,0 +1,36 @@
import { test } from 'vitest'
import assert from 'node:assert/strict'
import { clientIdFrom } from './route'
test('the OAuth query shape wins — it is the one carrying a redirect_uri', () => {
assert.equal(clientIdFrom('?client_id=hanzo-chat', '/signup'), 'hanzo-chat')
// Both present: the query is the real request; the segment is decoration.
assert.equal(clientIdFrom('?client_id=hanzo-app', '/signup/hanzo-chat'), 'hanzo-app')
})
test('the plain-link path shape is read — this is the bug hanzo.chat hit', () => {
// Three live components in hanzoai/chat link exactly here. App.tsx routes
// `/signup/` and `/login/`, so the shape was accepted and then dropped: the
// page fell back to the host default and created the account under a
// DIFFERENT application than the button that asked for it.
assert.equal(clientIdFrom('', '/signup/hanzo-chat'), 'hanzo-chat')
assert.equal(clientIdFrom('', '/login/hanzo-chat'), 'hanzo-chat')
assert.equal(clientIdFrom('', '/signup/hanzo-cloud'), 'hanzo-cloud')
})
test('anything that is not one <org>-<app> segment falls back to the host default', () => {
// These must stay undefined, or the page would authenticate as whatever junk
// was in the URL — the fallback is the host's declared app, which is correct.
for (const p of [
'/signup', // bare page, the ordinary case
'/login',
'/signup/', // trailing slash, no segment
'/signup/nodash', // not <org>-<app>
'/signup/a/b', // deeper path
'/signup/Hanzo-Chat', // ids are lower-case
'/signup/../admin',
'/',
]) {
assert.equal(clientIdFrom('', p), undefined, p)
}
})
+33
View File
@@ -0,0 +1,33 @@
/**
* Which application a visitor arrived FOR.
*
* Two shapes reach these pages and both are deliberate:
*
* /signup?client_id=hanzo-chat the OAuth shape, carrying the whole
* request (redirect_uri, state, PKCE)
* /signup/hanzo-chat the plain-link shape, for a marketing
* "Sign up" that starts no OAuth request
*
* `App.tsx` has always routed the second one — `path.startsWith('/signup/')`
* and `'/login/'` are in its switch — so the shape is accepted by design. The
* PAGES then read only `?client_id`, so the segment was matched and thrown
* away: hanzo.chat links to `hanzo.id/signup/hanzo-chat` from three components,
* and every one of them landed on a page that had silently fallen back to the
* host's default app. An account created there belongs to a different
* application than the button that asked for it.
*
* Query wins when both are present: it is the OAuth request, and it is the one
* that carries a redirect_uri to return through.
*/
export function clientIdFrom(search: string, pathname: string): string | undefined {
const fromQuery = new URLSearchParams(search).get('client_id')
if (fromQuery) return fromQuery
// Exactly one segment after the page, and it must look like an IAM client id
// (`<org>-<app>`, the estate's one naming rule). Anything else — a deeper
// path, an encoded slash, junk — resolves to undefined and the caller falls
// back to the host default, which is the behaviour that was there before.
const seg = pathname.split('/').filter(Boolean)
if (seg.length !== 2) return undefined
return /^[a-z0-9]+(-[a-z0-9]+)+$/.test(seg[1]!) ? seg[1] : undefined
}
+131
View File
@@ -0,0 +1,131 @@
/**
* Every token this surface references must RESOLVE.
*
* An undefined CSS custom property paints nothing and reports no error, so this
* whole class of defect survives review: `var(--surface-1)` and
* `var(--shadow-lg)` shipped in @hanzo/iam's account menu against a token layer
* that defines neither, and the menu rendered transparent. "It is declared" was
* never evidence — nor was "it type-checks", because the reference is built at
* runtime from a string and is invisible to both the compiler and grep.
*
* So the gate is resolution, and it is computed from what the bundle ACTUALLY
* serves: it walks app.css's @import graph into the installed @hanzo/design,
* collects the tokens those files declare, then asserts that every var(--x)
* anywhere under src/ — plus every token @hanzo/iam paints its menu with — is
* in that set.
*
* This fails if someone cherry-picks token groups again (the four-of-nine
* subset this file used to import left --z-*, --shadow-* and --space-* out),
* if @hanzo/design renames or drops a token, or if a component starts asking
* for a token in @hanzo/brand's vocabulary instead of @hanzo/design's.
*/
import { test } from 'vitest'
import assert from 'node:assert/strict'
import fs from 'node:fs'
import path from 'node:path'
const SRC = path.join(import.meta.dirname, '.')
/**
* Both packages restrict `exports`, so resolve them by walking up for the
* node_modules directory rather than by asking Node for a subpath it refuses.
*/
function pkgRoot(name: string): string {
for (let d = SRC; d !== path.dirname(d); d = path.dirname(d)) {
const p = path.join(d, 'node_modules', name)
if (fs.existsSync(path.join(p, 'package.json'))) return fs.realpathSync(p)
}
throw new Error(`${name} is not installed`)
}
const DESIGN = pkgRoot('@hanzo/design')
const read = (p: string) => fs.readFileSync(p, 'utf8')
const declaredIn = (css: string) => [...css.matchAll(/(--[a-zA-Z0-9-]+)\s*:/g)].map((m) => m[1])
const referencedIn = (css: string) => [...css.matchAll(/var\(\s*(--[a-zA-Z0-9-]+)/g)].map((m) => m[1])
/** Follow @import from an entry stylesheet into the @hanzo/design package. */
function tokenFiles(entry: string, seen = new Set<string>()): string[] {
for (const m of read(entry).matchAll(/@import\s+(?:url\()?['"]([^'"]+)['"]/g)) {
const spec = m[1]
const abs = spec.startsWith('@hanzo/design/')
? path.join(DESIGN, spec.slice('@hanzo/design/'.length))
: path.resolve(path.dirname(entry), spec)
if (seen.has(abs) || !fs.existsSync(abs)) continue
seen.add(abs)
tokenFiles(abs, seen)
}
return [...seen]
}
/** Every .css/.ts/.tsx under src/, so inline `var(--x)` in a component counts. */
function sources(dir: string, out: string[] = []): string[] {
for (const e of fs.readdirSync(dir, { withFileTypes: true })) {
const p = path.join(dir, e.name)
if (e.isDirectory()) sources(p, out)
// Skip this file: its own doc comment quotes `var(--x)`.
else if (/\.(css|tsx?)$/.test(e.name) && p !== import.meta.filename) out.push(p)
}
return out
}
const served = tokenFiles(path.join(SRC, 'app.css'))
const available = new Set(served.flatMap((f) => declaredIn(read(f))))
const files = sources(SRC)
const local = new Set(files.flatMap((f) => declaredIn(read(f))))
/**
* Coverage is judged by TOKENS, not by files. This used to assert that each
* `tokens/<group>.css` was itself reached through @import, which stopped being
* true at @hanzo/design 0.4.x: gen-tokens.mjs now FLATTENS all ten groups into
* styles.css so a bundler never has to resolve those subpaths. Nothing was
* dropped — the files still ship, they are just inlined — so asking "is every
* token this group declares actually served?" catches the cherry-picking this
* gate exists for, and survives however the package chooses to assemble itself.
*/
test('app.css serves the whole @hanzo/design token layer, not a subset', () => {
const groups = fs.readdirSync(path.join(DESIGN, 'tokens')).filter((f) => f.endsWith('.css'))
const missing = groups.filter((g) => {
const declared = [...new Set(declaredIn(read(path.join(DESIGN, 'tokens', g))))]
return declared.length > 0 && !declared.every((t) => available.has(t))
})
assert.deepEqual(missing, [], `token groups authored by @hanzo/design but never served here: ${missing.join(', ')}`)
/* base.css is the odd group and token coverage cannot see it: it ships the
ELEMENT DEFAULTS (the control, the focused control, the scrollbar) as 23
:where() rules, and the single token it declares — --border — is declared by
colors.css too. It is also the only group that opens `@layer base`, so that
is the exact marker for "the defaults are actually being served". */
assert.ok(
served.some((f) => /@layer\s+base/.test(read(f))),
'the element defaults from tokens/base.css are not served'
)
})
test('every token this surface references is defined', () => {
const unresolved = new Map<string, string[]>()
for (const f of files) {
for (const name of referencedIn(read(f))) {
if (available.has(name) || local.has(name)) continue
const at = unresolved.get(name) ?? []
at.push(path.relative(SRC, f))
unresolved.set(name, at)
}
}
assert.deepEqual(
[...unresolved].map(([n, at]) => `${n} (${[...new Set(at)].join(', ')})`),
[],
)
})
test('every token @hanzo/iam paints the account menu with is defined', () => {
// The menu is a distributed component: it emits its own stylesheet at
// runtime, so its token references never appear in this repo's source and no
// amount of grepping here would find them. Read them out of the shipped
// bundle instead — literal `var(--x)` plus the names its tok() helper builds.
const iam = read(path.join(pkgRoot('@hanzo/iam'), 'dist/react.js'))
const names = new Set([
...referencedIn(iam),
...[...iam.matchAll(/\btok\(\s*["']([a-zA-Z0-9-]+)["']/g)].map((m) => `--${m[1]}`),
])
const unresolved = [...names].filter((n) => !available.has(n)).sort()
assert.deepEqual(unresolved, [])
})
+16
View File
@@ -0,0 +1,16 @@
/// <reference types="vite/client" />
interface ImportMetaEnv {
/**
* Publishable event-ingest key (pk-…), inlined at build time from the
* PUBLISHABLE_KEY build-arg (KMS `deploy/PUBLISHABLE_KEY`, env `prod`).
* Declared so a typo reads as a type error rather than as `any` — Vite's
* ImportMetaEnv carries a string index signature, so an undeclared
* `import.meta.env.VITE_EVENT_INGEST_KEZ` would type-check and ship empty.
*/
readonly VITE_PUBLISHABLE_KEY: string
}
interface ImportMeta {
readonly env: ImportMetaEnv
}
+3 -2
View File
@@ -2,9 +2,10 @@
"extends": "../../tsconfig.base.json",
"compilerOptions": {
"outDir": "dist",
"baseUrl": ".",
"paths": {
"@/*": ["src/*"]
"@/*": [
"./src/*"
]
},
"types": ["vite/client"]
},
+103
View File
@@ -0,0 +1,103 @@
# Hanzo ID — the ONE file that says how this repo builds, gates and ships.
#
# The only workflow is `.hanzo/workflows/cicd.yml`, ~10 lines that import
# hanzoai/ci; everything real is here, and platform.hanzo.ai reads this same
# file. It replaces TWO image lanes that had drifted apart:
#
# .hanzo/workflows/deploy.yml the one that worked — buildx, two registry
# logins, a version read out of package.json and
# a post-push manifest check. Every line of it is
# something the reusable already does.
# .github/workflows/docker.yml a second lane on a plane with no runners for
# our labels. Already reduced to an echo, but it
# kept the shape of a build lane alive in a
# directory this forge cannot even read.
#
# Neither gated anything. This repo has 13 test files and 160 assertions and CI
# ran none of them, which is how `pkgs/shared/src/org.test.ts` sat RED on main
# while the code it tested was correct.
# Gates run DIRECTLY on the runner, not inside the image build. hanzoai/ci
# provisions Node 22 and enables corepack; corepack then reads `packageManager`
# from package.json, so the pnpm this uses is the SAME pnpm the Dockerfile
# activates — stated once, in package.json, rather than pinned again here.
#
# Three gates, because there are three different ways this repo breaks and one
# combined gate would report all of them as the same failure.
test:
- name: install
# --frozen-lockfile, deliberately stricter than the Dockerfile's
# `--frozen-lockfile=false`. The image build must not be blocked by a
# lockfile that drifted; CI is exactly where that drift should be caught.
run: |
set -e
corepack enable
pnpm install --frozen-lockfile
- name: typecheck
# Vite does NOT typecheck — `vite build` transpiles and strips types, so a
# type error ships a green image. This is the only thing in the pipeline
# that reads the types across all 7 workspace packages.
run: |
set -e
pnpm -r tc
- name: unit
# vitest over pkgs/**/src and apps/**/src (vitest.config.ts). 160 tests,
# including the org resolver that decides which brand a host authenticates
# as and which redirect_uri every social hop sends — the surface that
# produced days of "login is broken" and cannot be verified by looking at it.
run: |
set -e
pnpm test
images:
- name: id
context: .
dockerfile: Dockerfile
repo: ghcr.io/hanzoai/id
platforms: [linux/amd64]
# Fetched from KMS (`deploy/PUBLISHABLE_KEY`, env `prod`) and passed as
# --build-arg PUBLISHABLE_KEY. The reusable fails closed on an empty value,
# and the Dockerfile gates the shape and then asserts the key actually landed
# in the bundle — because a build that is green with an unattributed bundle
# is the failure mode here, not a build that goes red.
#
# This is the ONE spelling the fleet carries: KMS holds the plain name, each
# Dockerfile re-exports it with the prefix its bundler inlines.
build_secrets: [PUBLISHABLE_KEY]
# The version comes from package.json — hanzoai/ci's bin/imgver reads it as
# the declared floor, exactly as deploy.yml's `node -p require('./package.
# json').version` did, so a release is still a version bump and nothing about
# how this image is named changes.
#
# One difference worth stating: imgver takes max(declared, published)+1 patch
# when the declared version is ALREADY at the registry, where deploy.yml
# failed the build instead ("bump the version to cut a release"). Both refuse
# to put a second digest under a tag someone may already be pinning; this one
# publishes the next patch rather than going red on a docs commit.
#
# oci.hanzo.ai is still written to — the reusable crane-copies the exact tag
# set after the ghcr push. The PATH changes: deploy.yml pushed
# `oci.hanzo.ai/id`, the reusable writes the org-qualified
# `oci.hanzo.ai/hanzoai/id`, which is the fleet convention. Nothing live
# pulls either one; charts/app/values/hanzo/id.yaml pins
# `ghcr.io/hanzoai/id` by tag AND digest.
# No `deploy:` ON PURPOSE, and this is deploy.yml's own rule, not a new one.
#
# id is governed by Hanzo CD. The tag it runs is declared in
# hanzoai/universe (charts/app/values/hanzo/id.yaml, tag + digest together), and
# the in-cluster reconcile restores that within ~60-90s. A CI-side `kubectl
# patch` therefore CANNOT stick — and the reason to refuse it is not that it
# fails, it is that it LOOKS like it works: the patch applies, the pod rolls, and
# CD quietly puts the old tag back a minute later.
#
# It would also be actively wrong here. The reusable's deploy step guards a
# semver pin against a transient branch build by reading the current tag out of
# `infra/k8s/operator/crs/<svc>.yaml` — a path this service does not use — so the
# guard would not fire and a `sha-<short>-amd64` tag would be written over a
# reviewed semver+digest pin on the live sign-in surface for every property in
# the fleet.
#
# Build and deploy are separate concerns: this emits an immutable image, git
# declares desired state, CD applies it. To ship a build, set image.tag AND
# image.digest in universe and push.
+7 -3
View File
@@ -1,15 +1,19 @@
{
"name": "@hanzo/id",
"private": true,
"version": "0.1.25",
"description": "Hanzo ID \u2014 white-label login + identity verification portal (Vite + @hanzo/gui)",
"version": "0.2.28",
"description": "Hanzo ID white-label login + identity verification portal (Vite + @hanzo/gui)",
"packageManager": "pnpm@10.15.0",
"scripts": {
"build": "pnpm -r build",
"dev": "pnpm --filter @hanzo/id-web dev",
"tc": "pnpm -r tc",
"test": "vitest run",
"test:watch": "vitest",
"clean": "bash scripts/clean.sh"
},
"devDependencies": {
"typescript": "^5.9.3"
"typescript": "^5.9.3",
"vitest": "^3.2.4"
}
}
+4 -4
View File
@@ -1,6 +1,6 @@
{
"name": "@hanzo/id-auth",
"version": "0.1.2",
"version": "0.1.8",
"description": "Composable login / signup / OTP / OAuth-PKCE flows on top of @hanzo/iam. UI primitives in @hanzo/gui.",
"license": "BSD-3-Clause",
"type": "module",
@@ -16,13 +16,13 @@
"src"
],
"scripts": {
"tc": "tsc --noEmit",
"test": "node --test --experimental-strip-types 'src/**/*.test.ts'"
"tc": "tsc --noEmit"
},
"dependencies": {
"@hanzo/id-connect": "workspace:*",
"@hanzo/id-shared": "workspace:*",
"@hanzo/iam": "^0.13.1"
"@hanzo/iam": "^0.21.1",
"@paulmillr/qr": "^0.3.0"
},
"peerDependencies": {
"react": ">=19",
+156
View File
@@ -0,0 +1,156 @@
/**
* MFA wiring tests — pure, no network (fetch is mocked via `fetchImpl`).
* Run with: pnpm --filter @hanzo/id-auth test
*
* Locks the wire contract verified live against iam.hanzo.ai:
* - login answers a forced-MFA org with `data:"RequiredMfa"` (enroll) or
* `data:"NextMfa"` + the challenge list — named `mfa` first, legacy
* `data2`; both decode until the legacy slot is deleted. STRINGS, never a
* boolean.
* - the `/v1/iam/mfa/setup/*` calls carry EVERY param on the query string with
* an EMPTY body (the one shape IAM's authz self-match + controller accept).
* - the challenge re-POSTs `/v1/iam/login` with `{mfaType,passcode}` and NO
* username, riding the MFA session cookie.
*/
import { test } from 'vitest'
import assert from 'node:assert/strict'
import type { OrgConfig } from '@hanzo/id-shared'
import { createAuthClient, mfaChannelOf, MFA_TOTP } from './client.ts'
const TENANT: OrgConfig = {
orgId: 'hanzo',
iamUrl: 'https://hanzo.id',
iamIssuer: 'https://hanzo.id',
clientId: 'hanzo-id',
appName: 'hanzo-id',
publicOrigin: 'https://hanzo.id',
brandPackage: '@hanzo/brand',
}
type Call = { url: string; init: RequestInit }
function mockFetch(body: unknown, calls: Call[]): typeof fetch {
return (async (input: string | URL, init?: RequestInit) => {
calls.push({ url: String(input), init: init ?? {} })
return new Response(JSON.stringify(body), { status: 200, headers: { 'content-type': 'application/json' } })
}) as unknown as typeof fetch
}
test('login → RequiredMfa maps to an enroll signal (not a redirect)', async () => {
const calls: Call[] = []
const client = createAuthClient({ org: TENANT, fetchImpl: mockFetch({ status: 'ok', data: 'RequiredMfa' }, calls) })
const res = await client.login({
identifier: 'davelorenzini@gmail.com',
password: 'x',
clientId: 'hanzo-id',
application: 'hanzo-id',
organization: 'hanzo',
})
assert.equal(res.mfaRequired, true)
assert.equal(res.mfaStage, 'enroll')
assert.equal(res.redirectUrl, undefined, 'must NOT short-circuit to /onboarding')
})
const CHALLENGE = [{ mfaType: 'app', enabled: true }, { mfaType: 'sms', enabled: true }]
// Both spellings decode until IAM's envelope rename lands everywhere and the
// legacy slot is deleted: named `mfa` (new), untyped `data2` (legacy), and
// named-first precedence when a transitional server sends both.
test.each([
['named mfa', { status: 'ok', data: 'NextMfa', mfa: CHALLENGE }],
['legacy data2', { status: 'ok', data: 'NextMfa', data2: CHALLENGE }],
['mfa wins over data2', { status: 'ok', data: 'NextMfa', mfa: CHALLENGE, data2: [{ mfaType: 'email', enabled: true }] }],
])('login → NextMfa maps to a challenge signal and carries the allowed types (%s)', async (_spelling, body) => {
const calls: Call[] = []
const client = createAuthClient({ org: TENANT, fetchImpl: mockFetch(body, calls) })
const res = await client.login({
identifier: 'davelorenzini@gmail.com',
password: 'x',
clientId: 'hanzo-id',
application: 'hanzo-id',
organization: 'hanzo',
})
assert.equal(res.mfaStage, 'challenge')
assert.deepEqual(res.mfaTypes, ['app', 'sms'])
})
test('mfaInitiate puts owner/name/mfaType on the query string with an empty body', async () => {
const calls: Call[] = []
const data = { secret: 'BOUYRUSHJCEDDB33', url: 'otpauth://totp/Hanzo:x?secret=BOUYRUSHJCEDDB33', recoveryCodes: ['rc-1'] }
const client = createAuthClient({ org: TENANT, fetchImpl: mockFetch({ status: 'ok', data }, calls) })
const setup = await client.mfaInitiate({ owner: 'hanzo', name: 'davelorenzini@gmail.com' })
assert.equal(setup.secret, 'BOUYRUSHJCEDDB33')
assert.equal(setup.mfaType, MFA_TOTP)
assert.deepEqual(setup.recoveryCodes, ['rc-1'])
const u = new URL(calls[0].url)
assert.equal(u.pathname, '/v1/iam/mfa/setup/initiate')
assert.equal(u.searchParams.get('owner'), 'hanzo')
assert.equal(u.searchParams.get('name'), 'davelorenzini@gmail.com')
assert.equal(u.searchParams.get('mfaType'), 'app')
assert.equal(calls[0].init.method, 'POST')
assert.equal(calls[0].init.body, undefined, 'body must be empty for authz self-match')
assert.equal(calls[0].init.credentials, 'include')
})
test('mfaVerify carries owner/name (for authz) + secret + passcode on the query', async () => {
const calls: Call[] = []
const client = createAuthClient({ org: TENANT, fetchImpl: mockFetch({ status: 'ok', data: 'OK' }, calls) })
const r = await client.mfaVerify({ owner: 'hanzo', name: 'dave@x', secret: 'SEC', passcode: '123456' })
assert.equal(r.ok, true)
const u = new URL(calls[0].url)
assert.equal(u.pathname, '/v1/iam/mfa/setup/verify')
assert.equal(u.searchParams.get('owner'), 'hanzo')
assert.equal(u.searchParams.get('secret'), 'SEC')
assert.equal(u.searchParams.get('passcode'), '123456')
assert.equal(u.searchParams.get('mfaType'), 'app')
})
test('mfaVerify surfaces an IAM error instead of throwing', async () => {
const calls: Call[] = []
const client = createAuthClient({ org: TENANT, fetchImpl: mockFetch({ status: 'error', msg: 'wrong passcode' }, calls) })
const r = await client.mfaVerify({ owner: 'hanzo', name: 'dave@x', secret: 'SEC', passcode: '000000' })
assert.equal(r.ok, false)
assert.equal(r.error, 'wrong passcode')
})
test('mfaEnable echoes the recovery code back on the query', async () => {
const calls: Call[] = []
const client = createAuthClient({ org: TENANT, fetchImpl: mockFetch({ status: 'ok', data: 'OK' }, calls) })
const r = await client.mfaEnable({ owner: 'hanzo', name: 'dave@x', secret: 'SEC', recoveryCode: 'rc-1' })
assert.equal(r.ok, true)
const u = new URL(calls[0].url)
assert.equal(u.pathname, '/v1/iam/mfa/setup/enable')
assert.equal(u.searchParams.get('recoveryCodes'), 'rc-1')
assert.equal(u.searchParams.get('secret'), 'SEC')
})
test('mfaChallenge re-POSTs /v1/iam/login with mfaType/passcode and NO username', async () => {
const calls: Call[] = []
// code flow: data is the freshly minted auth code
const client = createAuthClient({ org: TENANT, fetchImpl: mockFetch({ status: 'ok', data: 'AUTHCODE' }, calls) })
const res = await client.mfaChallenge({
mfaType: 'app',
passcode: '654321',
clientId: 'hanzo-id',
application: 'hanzo-id',
organization: 'hanzo',
redirectUri: 'https://app.example/cb',
state: 'st',
})
const sent = JSON.parse(String(calls[0].init.body)) as Record<string, unknown>
assert.equal(new URL(calls[0].url).pathname, '/v1/iam/login')
assert.equal(sent.mfaType, 'app')
assert.equal(sent.passcode, '654321')
assert.equal(sent.username, undefined, 'challenge must not send a username')
assert.equal(calls[0].init.credentials, 'include')
assert.equal(res.redirectUrl, 'https://app.example/cb?code=AUTHCODE&state=st')
})
test('mfaChannelOf maps IAM types to UI channels', () => {
assert.equal(mfaChannelOf('app'), 'totp')
assert.equal(mfaChannelOf('sms'), 'sms')
assert.equal(mfaChannelOf('email'), 'email')
assert.equal(mfaChannelOf('anything-else'), 'totp')
})
+483 -80
View File
@@ -1,7 +1,7 @@
import { test } from 'node:test'
import { test } from 'vitest'
import assert from 'node:assert/strict'
import { createAuthClient } from './client.ts'
import type { TenantConfig } from '@hanzo/id-shared'
import type { OrgConfig } from '@hanzo/id-shared'
// A capturing fetch double: records the URL + parsed JSON body of the last call
// and returns a canned IAM "ok" response. No network.
@@ -20,7 +20,34 @@ function capturingFetch() {
return { calls, fetchImpl }
}
function tenant(overrides: Partial<TenantConfig> = {}): TenantConfig {
// A routing fetch double for the silent-SSO org gate: silentLogin resolves the
// app's org (`/v1/iam/get-app-login`) and the ambient session's owner
// (`/v1/iam/get-account`) BEFORE minting a code (`/v1/iam/login`). This lets a
// test set the app org + session owner independently and assert whether the mint
// leg ran. `sessionOwner: null` models "no live session" (get-account errors).
function routingFetch(opts: { appOrg: string; sessionOwner: string | null; code?: string }) {
const calls: { url: string; body: Record<string, unknown> }[] = []
const json = (payload: unknown) =>
new Response(JSON.stringify(payload), { status: 200, headers: { 'Content-Type': 'application/json' } })
const fetchImpl: typeof fetch = async (input, init) => {
const url = typeof input === 'string' ? input : input.toString()
let body: Record<string, unknown> = {}
if (init?.body && typeof init.body === 'string') body = JSON.parse(init.body)
calls.push({ url, body })
if (url.includes('/get-app-login')) {
return json({ status: 'ok', data: { name: 'app', organization: opts.appOrg, providers: [] } })
}
if (url.includes('/get-account')) {
return opts.sessionOwner
? json({ status: 'ok', data: { owner: opts.sessionOwner, name: 'z' } })
: json({ status: 'error', msg: 'please sign in first' })
}
return json({ status: 'ok', data: opts.code ?? 'AUTHCODE' })
}
return { calls, fetchImpl }
}
function org(overrides: Partial<OrgConfig> = {}): OrgConfig {
return {
orgId: 'hanzo',
iamUrl: 'https://hanzo.id',
@@ -34,27 +61,37 @@ function tenant(overrides: Partial<TenantConfig> = {}): TenantConfig {
}
}
// THE FIX: with loginOrg unset, the portal must NOT pin the brand org — it omits
// `organization` so IAM resolves the user cross-org (a global admin → the admin
// org / full session; a brand user → their own org). Pinning `hanzo` here is the
// live bug that truncates a global admin to one org.
test('login OMITS organization when loginOrg is unset (org-agnostic resolution)', async () => {
// `client.login` is a PURE PASSTHROUGH for `organization`: it sends what the
// caller gave it and omits the key when there is nothing to send. That is the
// contract these two tests pin, and it is unchanged.
//
// What DID change is whose job it is to supply one. This used to be deliberate
// omission — IAM resolved the user cross-org so a colliding identity
// (z@hanzo.ai exists in both `admin` and `hanzo`) landed on admin/* with a full
// multi-org session. iam2 removed that on purpose, treating the collision as a
// defect ("the F-2 bug where z@hanzo.ai collided across admin and hanzo": it
// coupled lockout counters across rows and gave a brute-force oracle on the
// superadmin), and now REFUSES an org-less login. So LoginForm resolves the
// app's own org via get-app-login and always passes one. Do not re-add an
// omit-the-org path here expecting the server to figure it out — it will not,
// and it fails with an HTTP 200 that reads like a wrong password.
test('login omits organization when the caller supplies none', async () => {
const { calls, fetchImpl } = capturingFetch()
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
await client.login({
identifier: 'z@hanzo.ai',
password: 'pw',
clientId: 'hanzo-console',
application: 'hanzo-console',
// organization intentionally not provided (LoginForm passes tenant.loginOrg)
// organization intentionally not provided LoginForm now always resolves one
})
assert.equal(calls.length, 1)
assert.equal(
'organization' in calls[0]!.body,
false,
'organization must be absent from the body so IAM runs cross-org resolution',
'organization must be absent when the caller supplies none — the client never invents one',
)
// The identity + app still ride the request.
assert.equal(calls[0]!.body.username, 'z@hanzo.ai')
@@ -63,9 +100,9 @@ test('login OMITS organization when loginOrg is unset (org-agnostic resolution)'
// An empty-string org is treated the same as unset (defensive: a catalog might
// emit "").
test('login OMITS organization when it is an empty string', async () => {
test('login omits organization when it is an empty string', async () => {
const { calls, fetchImpl } = capturingFetch()
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
await client.login({
identifier: 'z@hanzo.ai',
password: 'pw',
@@ -79,7 +116,7 @@ test('login OMITS organization when it is an empty string', async () => {
// A brand that DELIBERATELY scopes its portal to one org can still force it.
test('login INCLUDES organization when one is explicitly provided', async () => {
const { calls, fetchImpl } = capturingFetch()
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
await client.login({
identifier: 'someone',
password: 'pw',
@@ -95,7 +132,7 @@ test('login INCLUDES organization when one is explicitly provided', async () =>
// for the SSO path too).
test('app SSO (redirectUri present) uses type=code and still omits organization', async () => {
const { calls, fetchImpl } = capturingFetch()
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
await client.login({
identifier: 'z@hanzo.ai',
password: 'pw',
@@ -112,7 +149,7 @@ test('app SSO (redirectUri present) uses type=code and still omits organization'
// Signup MUST still carry a concrete org — you cannot create a user in "no org".
test('signup STILL sends organization (unchanged — create needs a concrete org)', async () => {
const { calls, fetchImpl } = capturingFetch()
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
await client.signup({
email: 'new@hanzo.ai',
password: 'pw',
@@ -123,7 +160,83 @@ test('signup STILL sends organization (unchanged — create needs a concrete org
assert.equal(calls[0]!.body.organization, 'hanzo')
})
// REGRESSION (the `hanzo-iam does not exist` social-login bug): a Casdoor
// REGRESSION (every new customer was stranded on the portal): IAM's signup is
// CREATE-ONLY — it sets no session and mints no code. Signup must therefore end
// in a real sign-in, or the app that sent the user waits forever for a code.
test('signup COMPLETES the OIDC request — create, then sign in, then redirect back with the code', async () => {
const { calls, fetchImpl } = capturingFetch()
const client = createAuthClient({ org: org(), fetchImpl })
const res = await client.signup({
email: 'new@hanzo.ai',
password: 'pw',
clientId: 'hanzo-app',
application: 'hanzo-app',
organization: 'hanzo',
redirectUri: 'https://hanzo.app/auth/callback',
state: 'xyz',
codeChallenge: 'CHALLENGE',
codeChallengeMethod: 'S256',
})
// Two legs, in order: create the row, then authenticate it.
assert.equal(calls.length, 2)
assert.match(calls[0]!.url, /\/v1\/iam\/signup/)
assert.match(calls[1]!.url, /\/v1\/iam\/login/)
// The sign-in leg carries the downstream request, so the code is PKCE-bound.
assert.match(calls[1]!.url, /code_challenge=CHALLENGE/)
assert.match(calls[1]!.url, /code_challenge_method=S256/)
assert.match(calls[1]!.url, /type=code/)
assert.equal(calls[1]!.body.username, 'new@hanzo.ai')
// And the caller is handed a destination BACK AT THE APP — never the portal's
// own /onboarding, which is where the create-only response used to land.
assert.equal(res.redirectUrl, 'https://hanzo.app/auth/callback?code=AUTHCODE&state=xyz')
})
// `autoSignin` was posted for its name and dropped on the floor: the Go
// signupForm has no such field, so it never signed anyone in. Do not post a flag
// the server does not read — it is what made this look like it worked.
test('signup does NOT post autoSignin (IAM has no such field)', async () => {
const { calls, fetchImpl } = capturingFetch()
const client = createAuthClient({ org: org(), fetchImpl })
await client.signup({
email: 'new@hanzo.ai',
password: 'pw',
clientId: 'hanzo-app',
application: 'hanzo-app',
organization: 'hanzo',
})
assert.equal('autoSignin' in calls[0]!.body, false)
})
// IAM refuses with HTTP 200 + status:"error", so the status code proves nothing.
// A refused create must surface the reason and must NOT go on to try a login.
test('a refused signup surfaces the reason and never attempts a sign-in', async () => {
const seen: string[] = []
const fetchImpl: typeof fetch = async (input) => {
seen.push(typeof input === 'string' ? input : input.toString())
return new Response(JSON.stringify({ status: 'error', msg: 'email already exists', data: null }), {
status: 200,
headers: { 'Content-Type': 'application/json' },
})
}
const client = createAuthClient({ org: org(), fetchImpl })
const res = await client.signup({
email: 'taken@hanzo.ai',
password: 'pw',
clientId: 'hanzo-app',
application: 'hanzo-app',
organization: 'hanzo',
redirectUri: 'https://hanzo.app/auth/callback',
})
assert.equal(res.error, 'email already exists')
assert.equal(res.redirectUrl, undefined)
assert.equal(seen.length, 1)
assert.match(seen[0]!, /\/v1\/iam\/signup/)
})
// REGRESSION (the `hanzo-iam does not exist` social-login bug): an IAM
// app-provider LINK can carry an outer `name` that is NOT the provider record's
// name (some seeds label it `<org>-iam`). The provider's real identity is the
// nested `provider.name` the backend resolves on the social hop. getAppLogin
@@ -152,7 +265,7 @@ test('getAppLogin uses the nested provider record name, not the outer link label
},
],
})
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
const app = await client.getAppLogin('hanzo-console')
assert.ok(app, 'app login resolved')
assert.equal(app!.providers.length, 1)
@@ -163,37 +276,38 @@ test('getAppLogin uses the nested provider record name, not the outer link label
assert.equal(gh.configured, true, 'a real (non-placeholder) clientId is configured')
})
// The social code exchange must reuse the provider's REGISTERED callback host
// (oauthCallbackOrigin), not the brand host (publicOrigin). IAM forwards this
// redirect_uri verbatim to the provider's token endpoint, which requires it to
// match the authorize hop or the exchange fails `invalid_grant`. When a brand
// portal (hanzo.id) shares the iam.hanzo.ai OAuth client these two differ.
test('providerLogin posts redirectUri from oauthCallbackOrigin (matches the hop), with the single provider + code', async () => {
const { calls, fetchImpl } = capturingFetch()
const client = createAuthClient({
tenant: tenant({ publicOrigin: 'https://hanzo.id', oauthCallbackOrigin: 'https://iam.hanzo.ai' }),
fetchImpl,
})
const r = await client.providerLogin({
application: 'hanzo-console',
provider: 'provider-google',
code: 'goog_code_xyz',
oidcQuery:
'?client_id=hanzo-console&redirect_uri=https%3A%2F%2Fconsole.hanzo.ai%2Fauth%2Fiam%2Fcallback&response_type=code&scope=openid&state=rp1',
method: 'signin',
})
assert.equal(calls.length, 1)
assert.equal(
calls[0]!.body.redirectUri,
'https://iam.hanzo.ai/callback',
'redirect_uri derives from oauthCallbackOrigin (the hop), never publicOrigin',
// Federation is entered by NAMING the provider on IAM's own authorize endpoint.
// The `provider` field has always been on OAuthAuthorizeRequest; `authorize`
// never emitted it, which is why social sign-in had no server side at all.
test('authorize emits the provider record name, so IAM federates instead of showing its login', () => {
const client = createAuthClient({ org: org({ iamUrl: 'https://hanzo.id' }) })
const url = new URL(
client.authorize({
clientId: 'hanzo-console',
redirectUri: 'https://hanzo.id/callback',
state: 'rp1',
codeChallenge: 'C1',
codeChallengeMethod: 'S256',
provider: 'provider-github',
}),
)
assert.equal(calls[0]!.body.provider, 'provider-google')
assert.equal(calls[0]!.body.code, 'goog_code_xyz')
// The upstream OIDC params ride the query so IAM continues the original authorize.
assert.match(calls[0]!.url, /client_id=hanzo-console/)
assert.match(calls[0]!.url, /state=rp1/)
assert.equal(r.redirectUrl, 'AUTHCODE')
assert.equal(url.pathname, '/v1/iam/oauth/authorize')
// The RECORD name, never the bare key: federationProvider matches
// ProviderItem.Name exactly (live, `provider=github` is refused).
assert.equal(url.searchParams.get('provider'), 'provider-github')
// The app's own request is what IAM binds the minted code to.
assert.equal(url.searchParams.get('client_id'), 'hanzo-console')
assert.equal(url.searchParams.get('redirect_uri'), 'https://hanzo.id/callback')
assert.equal(url.searchParams.get('code_challenge'), 'C1')
assert.equal(url.searchParams.get('code_challenge_method'), 'S256')
})
test('authorize without a provider stays the ordinary hosted-login request', () => {
const client = createAuthClient({ org: org({ iamUrl: 'https://hanzo.id' }) })
const url = new URL(
client.authorize({ clientId: 'hanzo-console', redirectUri: 'https://hanzo.id/callback', state: 'rp1' }),
)
assert.equal(url.searchParams.get('provider'), null)
})
// When there is NO nested record (degenerate seed), fall back to the outer label
@@ -204,7 +318,7 @@ test('getAppLogin falls back to the outer name when no nested provider record',
organization: 'hanzo',
providers: [{ name: 'provider-google', canSignIn: true, canSignUp: true, provider: null }],
})
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
const app = await client.getAppLogin('hanzo-console')
assert.ok(app)
assert.equal(app!.providers[0]!.name, 'provider-google')
@@ -212,10 +326,12 @@ test('getAppLogin falls back to the outer name when no nested provider record',
// TRUE SSO — the silent leg. silentLogin carries NO credentials: IAM mints the
// code from the existing issuer session (cookie sent via credentials:include).
// It builds the redirect back to the app from the minted code + state.
test('silentLogin posts NO credentials and redirects with the minted code', async () => {
const { calls, fetchImpl } = capturingFetch()
const client = createAuthClient({ tenant: tenant(), fetchImpl })
// It builds the redirect back to the app from the minted code + state. The mint
// runs ONLY when the ambient session's org matches the app's org (same-org SSO,
// the common case: a hanzo session signing into a hanzo app).
test('silentLogin (same-org session) mints the code and redirects, carrying NO credentials', async () => {
const { calls, fetchImpl } = routingFetch({ appOrg: 'hanzo', sessionOwner: 'hanzo' })
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.silentLogin({
clientId: 'hanzo-console',
@@ -225,50 +341,83 @@ test('silentLogin posts NO credentials and redirects with the minted code', asyn
codeChallenge: 'chal',
})
assert.equal(calls.length, 1)
// The mint leg (POST /v1/iam/login, type=code) ran after the org gate passed.
const mint = calls.find((c) => c.body.type === 'code')
assert.ok(mint, 'mint leg ran for a same-org session')
// No credentials of any kind — this is session-only.
assert.equal('username' in calls[0]!.body, false, 'no username in silent login')
assert.equal('password' in calls[0]!.body, false, 'no password in silent login')
assert.equal('provider' in calls[0]!.body, false, 'no provider hop in silent login')
// The body carries the code intent + the target application only.
assert.equal(calls[0]!.body.type, 'code')
assert.equal(calls[0]!.body.application, 'hanzo-console')
assert.equal('username' in mint!.body, false, 'no username in silent login')
assert.equal('password' in mint!.body, false, 'no password in silent login')
assert.equal('provider' in mint!.body, false, 'no provider hop in silent login')
assert.equal(mint!.body.application, 'hanzo-console')
// OAuth params ride the query so IAM mints a code for the right client + PKCE.
assert.match(calls[0]!.url, /clientId=hanzo-console/)
assert.match(calls[0]!.url, /code_challenge=chal/)
// The capturing fetch returns data:'AUTHCODE' -> a fully-formed app redirect.
assert.match(mint!.url, /clientId=hanzo-console/)
assert.match(mint!.url, /code_challenge=chal/)
// The mint returns data:'AUTHCODE' -> a fully-formed app redirect.
assert.equal(
r.redirectUrl,
'https://console.hanzo.ai/auth/iam/callback?code=AUTHCODE&state=st1',
)
})
// No live session: IAM answers status:error -> silentLogin surfaces { error }
// so Login.tsx falls back to the interactive form (never a dead end).
test('silentLogin returns { error } when there is no session (form fallback)', async () => {
const fetchImpl: typeof fetch = async () =>
new Response(JSON.stringify({ status: 'error', msg: 'please sign in first' }), {
status: 200,
headers: { 'Content-Type': 'application/json' },
})
const client = createAuthClient({ tenant: tenant(), fetchImpl })
// THE ADMIN-GUARD FIX: silent SSO must NOT reuse a session that belongs to a
// DIFFERENT org than the app being signed into. An operator with an ambient
// hanzo/* session hitting the admin-guard (org=admin) must fall through to the
// interactive form (which authenticates in the admin org and resolves the
// admin/* identity) — NOT silently mint a code from the hanzo session (which
// would confer owner=hanzo and shadow the fix). No mint leg runs; no redirect.
test('silentLogin (cross-org session) does NOT mint — falls back to the form', async () => {
const { calls, fetchImpl } = routingFetch({ appOrg: 'admin', sessionOwner: 'hanzo' })
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.silentLogin({
clientId: 'hanzo-admin-guard',
application: 'hanzo-admin-guard',
redirectUri: 'https://admin.hanzo.ai/__guard/callback',
state: 'st1',
codeChallenge: 'chal',
})
assert.equal(r.redirectUrl, undefined, 'no silent redirect for a cross-org session')
assert.equal(calls.some((c) => c.body.type === 'code'), false, 'the mint leg must NOT run')
})
// Same-org SSO still holds when BOTH are the admin org: an operator already
// signed in as admin/* silently re-enters the admin console.
test('silentLogin (same admin-org session) mints for the admin-guard', async () => {
const { calls, fetchImpl } = routingFetch({ appOrg: 'admin', sessionOwner: 'admin' })
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.silentLogin({
clientId: 'hanzo-admin-guard',
application: 'hanzo-admin-guard',
redirectUri: 'https://admin.hanzo.ai/__guard/callback',
state: 'st1',
})
assert.ok(calls.find((c) => c.body.type === 'code'), 'mint leg ran for a same-org admin session')
assert.equal(r.redirectUrl, 'https://admin.hanzo.ai/__guard/callback?code=AUTHCODE&state=st1')
})
// No live session: silentLogin returns an empty response (no mint) so Login.tsx
// falls back to the interactive form (never a dead end).
test('silentLogin returns empty (no mint) when there is no session (form fallback)', async () => {
const { calls, fetchImpl } = routingFetch({ appOrg: 'hanzo', sessionOwner: null })
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.silentLogin({
clientId: 'hanzo-console',
application: 'hanzo-console',
redirectUri: 'https://console.hanzo.ai/auth/iam/callback',
})
assert.equal(r.redirectUrl, undefined)
assert.equal(r.error, 'please sign in first')
assert.equal(calls.some((c) => c.body.type === 'code'), false, 'no mint without a session')
})
// ── Device-authorization approval (RFC 8628) ─────────────────────────────────
// approveDevice rides the issuer SESSION (like silentLogin): NO credentials in
// the body, `type:device` + the userCode IAM keys its DeviceAuthMap on, plus the
// tenant application/organization for the app lookup. On {status:ok} the device
// org application/organization for the app lookup. On {status:ok} the device
// code is approved (UserSignIn=true) and the CLI's token poll succeeds.
test('approveDevice posts type=device + normalized userCode + tenant app/org, NO credentials', async () => {
test('approveDevice posts type=device + normalized userCode + org app/org, NO credentials', async () => {
const { calls, fetchImpl } = capturingFetch()
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.approveDevice('K7M4P2QH')
@@ -290,7 +439,7 @@ test('approveDevice posts type=device + normalized userCode + tenant app/org, NO
// TO uppercase so the lookup matches — case-insensitive entry, exact-match send.
test('approveDevice uppercases and strips spaces/dashes before sending', async () => {
const { calls, fetchImpl } = capturingFetch()
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
await client.approveDevice(' k7m4-p2qh ')
assert.equal(calls[0]!.body.userCode, 'K7M4P2QH')
})
@@ -298,7 +447,7 @@ test('approveDevice uppercases and strips spaces/dashes before sending', async (
// An empty/blank code never hits the network — fail fast with a clear message.
test('approveDevice rejects an empty code without calling fetch', async () => {
const { calls, fetchImpl } = capturingFetch()
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.approveDevice(' ')
assert.equal(calls.length, 0)
assert.equal(r.ok, false)
@@ -312,7 +461,7 @@ test('approveDevice surfaces the IAM error message', async () => {
status: 200,
headers: { 'Content-Type': 'application/json' },
})
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.approveDevice('K7M4P2QH')
assert.equal(r.ok, false)
assert.equal(r.error, 'UserCode Expired')
@@ -326,9 +475,263 @@ test('approveDevice maps the consent-required branch to { required: true }', asy
status: 200,
headers: { 'Content-Type': 'application/json' },
})
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.approveDevice('K7M4P2QH')
assert.equal(r.ok, false)
assert.equal(r.required, true)
assert.equal(r.error, undefined)
})
// ── Which application is this code for? (deviceInfo) ─────────────────────────
// A one-call double for `POST /v1/iam/oauth/device/info`: records what the
// request actually was (URL, method, credentials, body) and answers with
// `payload`.
function deviceInfoFetch(payload: unknown) {
const calls: {
url: string
method?: string
credentials?: RequestCredentials
body?: string
}[] = []
const fetchImpl: typeof fetch = async (input, init) => {
calls.push({
url: typeof input === 'string' ? input : input.toString(),
method: init?.method,
credentials: init?.credentials,
body: typeof init?.body === 'string' ? init.body : undefined,
})
return new Response(JSON.stringify(payload), {
status: 200,
headers: { 'Content-Type': 'application/json' },
})
}
return { calls, fetchImpl }
}
// THE REGRESSION THIS FILE EXISTS FOR. The approval page used to render
// `org.appName` — the PORTAL's own branding, the static `hanzo-console` this
// test's org() is configured with — so a device sign-in started by `hanzo-cli`
// was approved under a screen naming a different application. The name must come
// off the RESPONSE, which is the code's own application, and never off the org
// config; asserting both is what keeps the two from being confused again.
test('deviceInfo names the RESPONSE client, never the portal org appName', async () => {
const { calls, fetchImpl } = deviceInfoFetch({
status: 'ok',
data: { clientId: 'hanzo-cli', displayName: 'Hanzo CLI' },
})
const cfg = org()
const client = createAuthClient({ org: cfg, fetchImpl })
const r = await client.deviceInfo('K7M4P2QH')
assert.equal(r.ok, true)
assert.equal(r.ok && r.clientId, 'hanzo-cli')
assert.equal(r.ok && r.displayName, 'Hanzo CLI')
// The portal is hanzo-console. Nothing about it may reach the result.
assert.equal(cfg.appName, 'hanzo-console')
assert.notEqual(r.ok && r.clientId, cfg.appName)
assert.notEqual(r.ok && r.displayName, cfg.appName)
// A session-cookie POST at the /v1/ device-info path. The user_code is the one
// secret in this flow, so it rides the BODY: a request line is copied into
// ingress and proxy access logs where a body is not, and this page ships
// scrubUrl() precisely to keep the code out of URLs.
assert.equal(calls.length, 1)
assert.equal(calls[0]!.url, 'https://hanzo.id/v1/iam/oauth/device/info')
assert.equal(calls[0]!.method, 'POST')
assert.equal(calls[0]!.credentials, 'include')
assert.equal(calls[0]!.body, JSON.stringify({ userCode: 'K7M4P2QH' }))
assert.equal(calls[0]!.url.includes('K7M4P2QH'), false)
})
// Same normalization as the approval: a code transcribed lower-cased or with
// dashes must resolve to the same row IAM minted, or the page would refuse to
// name an application that is perfectly live.
test('deviceInfo uppercases and strips spaces/dashes into the body', async () => {
const { calls, fetchImpl } = deviceInfoFetch({
status: 'ok',
data: { clientId: 'hanzo-cli', displayName: 'Hanzo CLI' },
})
const client = createAuthClient({ org: org(), fetchImpl })
await client.deviceInfo(' k7m4-p2qh ')
assert.equal(calls[0]!.url, 'https://hanzo.id/v1/iam/oauth/device/info')
assert.equal(calls[0]!.body, JSON.stringify({ userCode: 'K7M4P2QH' }))
})
// An empty code names nothing and never hits the network.
test('deviceInfo rejects an empty code without calling fetch', async () => {
const { calls, fetchImpl } = deviceInfoFetch({ status: 'ok', data: {} })
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.deviceInfo(' ')
assert.equal(calls.length, 0)
assert.equal(r.ok, false)
assert.equal(r.ok === false && r.loginRequired, undefined)
})
// IAM `CodeLoginRequired`: the session lapsed. Flagged separately from a refusal
// because the page's answer is to sign the human in and come back, not to give up.
test('deviceInfo flags login_required distinctly from a refusal', async () => {
const { fetchImpl } = deviceInfoFetch({
status: 'error',
msg: 'please sign in first',
code: 'login_required',
})
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.deviceInfo('K7M4P2QH')
assert.equal(r.ok, false)
assert.equal(r.ok === false && r.loginRequired, true)
assert.equal(r.ok === false && r.error, 'please sign in first')
})
// The ONE opaque refusal IAM answers for unknown / expired / already-approved —
// surfaced verbatim, carrying no loginRequired, so the page shows it and offers
// no approval. Distinguishing those three would be an oracle for hunting the
// 40-bit user_code; the client must not invent a distinction either.
test('deviceInfo surfaces the opaque refusal verbatim and does not name an app', async () => {
const { fetchImpl } = deviceInfoFetch({
status: 'error',
msg: 'the user code is invalid or expired',
})
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.deviceInfo('K7M4P2QH')
assert.equal(r.ok, false)
assert.equal(r.ok === false && r.error, 'the user code is invalid or expired')
assert.equal(r.ok === false && r.loginRequired, undefined)
})
// The org-boundary refusal is a plain refusal too: surfaced, not special-cased.
test('deviceInfo surfaces the wrong-org refusal', async () => {
const { fetchImpl } = deviceInfoFetch({
status: 'error',
msg: 'your organization may not approve this device sign-in',
})
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.deviceInfo('K7M4P2QH')
assert.equal(r.ok, false)
assert.equal(r.ok === false && r.error, 'your organization may not approve this device sign-in')
})
// An HTML error page from a proxy is not an application name. It must fail,
// never resolve to a blank or guessed one.
test('deviceInfo fails on a non-JSON response', async () => {
const fetchImpl: typeof fetch = async () =>
new Response('<html>502 Bad Gateway</html>', {
status: 502,
headers: { 'Content-Type': 'text/html' },
})
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.deviceInfo('K7M4P2QH')
assert.equal(r.ok, false)
assert.match(String(r.ok === false && r.error), /non-JSON/)
})
// A network failure resolves — never rejects — so the page renders the failure
// instead of tearing down on an unhandled rejection.
test('deviceInfo resolves an error when fetch throws', async () => {
const fetchImpl: typeof fetch = async () => {
throw new Error('offline')
}
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.deviceInfo('K7M4P2QH')
assert.equal(r.ok, false)
assert.match(String(r.ok === false && r.error), /offline/)
})
// A 200 that names no client is not a name. Falling back to ANY local string here
// is what produced the original defect, so an absent clientId is a failure.
test('deviceInfo refuses an ok response with no clientId', async () => {
const { fetchImpl } = deviceInfoFetch({ status: 'ok', data: { displayName: 'Hanzo CLI' } })
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.deviceInfo('K7M4P2QH')
assert.equal(r.ok, false)
})
// IAM already falls back to the app's name when DisplayName is empty; if one ever
// arrives blank anyway, the label is the server-confirmed clientId — never the portal's.
test('deviceInfo falls back to the confirmed clientId when displayName is empty', async () => {
const { fetchImpl } = deviceInfoFetch({ status: 'ok', data: { clientId: 'hanzo-cli', displayName: '' } })
const client = createAuthClient({ org: org(), fetchImpl })
const r = await client.deviceInfo('K7M4P2QH')
assert.equal(r.ok, true)
assert.equal(r.ok && r.displayName, 'hanzo-cli')
})
// getAppLogin's redirectUri is validated by IAM against the app's REGISTERED
// list. A cross-app SSO read (the console's `hanzo-cloud` viewed from hanzo.id)
// MUST send the downstream app's OWN redirect_uri — the portal's `/callback` is
// not in that app's list, so hardcoding it makes IAM drop the response and no
// social buttons resolve. Absent, it defaults to the portal's own callback.
test('getAppLogin sends the passed redirect_uri, and defaults to the portal callback when omitted', async () => {
const urls: string[] = []
const fetchImpl: typeof fetch = async (input) => {
urls.push(typeof input === 'string' ? input : input.toString())
return new Response(
JSON.stringify({ status: 'ok', data: { name: 'hanzo-cloud', organization: 'hanzo', providers: [] } }),
{ status: 200, headers: { 'Content-Type': 'application/json' } },
)
}
const client = createAuthClient({ org: org(), fetchImpl })
// Cross-app read: the console's registered redirect_uri rides through verbatim.
await client.getAppLogin('hanzo-cloud', 'https://console.hanzo.ai/auth/callback')
const u1 = new URL(urls[0]!)
assert.equal(u1.searchParams.get('clientId'), 'hanzo-cloud')
assert.equal(u1.searchParams.get('redirectUri'), 'https://console.hanzo.ai/auth/callback')
// Bare/own read: no redirect_uri → default to the portal's own /callback.
await client.getAppLogin('hanzo-id')
const u2 = new URL(urls[1]!)
assert.equal(u2.searchParams.get('redirectUri'), 'https://hanzo.id/callback')
})
// Sign-out has to clear THIS BROWSER, not just end the session at the IdP.
//
// The bug: Portal navigated straight to client.logout(), which builds the
// RP-initiated logout URL and nothing else. The server really did revoke the
// token (measured against prod: the leftover token 401s "invalid or revoked"),
// but every `hanzo_iam_*` key survived — so the token STRING outlived the
// session it named, and anything treating that key's presence as "signed in"
// still believed you were.
test('signOut clears every hanzo_iam_* key, in BOTH storages, and returns the IdP URL', () => {
const store = () => {
const m = new Map<string, string>()
return {
get length() { return m.size },
key: (i: number) => [...m.keys()][i] ?? null,
getItem: (k: string) => m.get(k) ?? null,
setItem: (k: string, v: string) => void m.set(k, v),
removeItem: (k: string) => void m.delete(k),
clear: () => m.clear(),
_keys: () => [...m.keys()],
}
}
const ss = store(), ls = store()
const g = globalThis as Record<string, unknown>
const [ps, pl] = [g.sessionStorage, g.localStorage]
g.sessionStorage = ss
g.localStorage = ls
try {
// Three of the SDK's keys, plus a neighbour that must SURVIVE — otherwise
// "it cleared everything" would pass this test just as well.
for (const s of [ss, ls]) {
s.setItem('hanzo_iam_access_token', 'x')
s.setItem('hanzo_iam_expires_at', 'x')
s.setItem('hanzo_iam_code_verifier:abc', 'x')
s.setItem('theme', 'dark')
}
const client = createAuthClient({
org: { orgId: 'hanzo', iamUrl: 'https://hanzo.id', iamIssuer: 'https://hanzo.id',
clientId: 'hanzo-console', appName: 'hanzo-console',
publicOrigin: 'https://hanzo.id', brandPackage: '@hanzo/brand' } as never,
})
const url = client.signOut('https://hanzo.id/login')
assert.deepEqual(ss._keys(), ['theme'], 'sessionStorage')
assert.deepEqual(ls._keys(), ['theme'], 'localStorage')
assert.match(url, /\/v1\/iam\/oauth\/logout\?/)
assert.match(url, /post_logout_redirect_uri=https%3A%2F%2Fhanzo\.id%2Flogin/)
} finally {
g.sessionStorage = ps
g.localStorage = pl
}
})
+421 -136
View File
@@ -1,23 +1,36 @@
import type { TenantConfig } from '@hanzo/id-shared'
import type { OrgConfig } from '@hanzo/id-shared'
import type {
AppLogin,
AppProvider,
DeviceApprovalResult,
DeviceInfoResult,
ForgotRequest,
LoginRequest,
LoginResponse,
MfaChallengeRequest,
MfaChannel,
MfaIdentity,
MfaSetup,
OAuthAuthorizeRequest,
SignupRequest,
SilentLoginRequest,
TokenResponse,
} from './types'
/** IAM's TOTP MFA type constant (`object.TotpType`). */
export const MFA_TOTP = 'app'
/** Map an IAM MFA type to the {@link MfaChannel} the OTP UI renders a label for. */
export function mfaChannelOf(iamType: string): MfaChannel {
return iamType === 'sms' ? 'sms' : iamType === 'email' ? 'email' : 'totp'
}
/**
* Composable IAM client.
*
* Stateless wrapper around the canonical IAM REST surface (Casdoor-compat
* paths under `/v1/iam/*` and the OIDC paths under `/v1/iam/oauth/*`). One
* client instance per tenant. The portal creates one in `createRoot()`;
* Stateless wrapper around the canonical IAM REST surface (paths under
* `/v1/iam/*` and the OIDC paths under `/v1/iam/oauth/*`). One
* client instance per org. The portal creates one in `createRoot()`;
* downstream pages call `.login()`, `.signup()`, `.forgot()`, `.authorize()`
* directly.
*
@@ -34,7 +47,7 @@ import type {
* past the redirect.
*/
export interface AuthClient {
readonly tenant: TenantConfig
readonly org: OrgConfig
login(req: LoginRequest): Promise<LoginResponse>
/**
* Silent single-sign-on: mint an authorization code from the EXISTING issuer
@@ -51,64 +64,106 @@ export interface AuthClient {
* issuer — this rides the SAME `iam_session_id` cookie as silent SSO
* (`credentials:'include'`, no credentials in the body). It POSTs
* `/v1/iam/login` with `type:'device'` + the `userCode` the device shows,
* plus the tenant's `application`/`organization`; IAM resolves the user from
* plus the org's `application`/`organization`; IAM resolves the user from
* the session, flips the device code's `UserSignIn=true`, and the CLI's token
* poll then succeeds. Returns `{required:true}` when the app needs consent
* first (rare for first-party apps), or `{error}` with the IAM message.
*/
approveDevice(userCode: string): Promise<DeviceApprovalResult>
/**
* Name the application a pending device code belongs to, so the approval page
* can say WHICH app it is authorizing — `GET
* /v1/iam/oauth/device/<user_code>`, riding the same `iam_session_id` cookie
* as {@link approveDevice}.
*
* Read this and render it; never `org.appName`, which is this portal's own
* branding and names a different application than the one that minted the
* code. IAM answers from the code's own application row.
*
* Session-gated and deliberately terse: an expired session comes back as
* `loginRequired`, and unknown / expired / already-approved all come back as
* ONE indistinguishable refusal, because a user_code is 40 bits and an
* endpoint that told them apart would be an oracle for hunting live codes.
*/
deviceInfo(userCode: string): Promise<DeviceInfoResult>
signup(req: SignupRequest): Promise<LoginResponse>
forgot(req: ForgotRequest): Promise<{ ok: boolean; error?: string }>
authorize(req: OAuthAuthorizeRequest): string
exchange(code: string, codeVerifier?: string): Promise<TokenResponse>
logout(idTokenHint?: string, postLogoutRedirectUri?: string): string
/**
* Sign out COMPLETELY: drop this browser's own tokens, then hand back the
* IdP's RP-initiated logout URL for the caller to navigate to.
*
* `logout()` alone is only half of it, and the missing half is the half a
* person notices. It builds the IdP URL, which ends the session at the
* server — measured, the access token really is revoked there — but it
* touches nothing this browser stored, so `hanzo_iam_access_token` and its
* siblings survive a sign-out. Anything that treats the presence of that key
* as "signed in" then still believes you are (hanzoai/playground's AuthGuard
* reads exactly that key), and a token string that outlives its session is a
* thing to delete on principle even where nothing reads it.
*
* So sign-out is ONE call, not a URL plus a cleanup every caller has to
* remember. Local first, then the redirect: a navigation ends this
* document, and anything left after `location.href` is a coin flip.
*/
signOut(postLogoutRedirectUri?: string): string
/**
* Read the live enabled-auth-methods view for an application from
* `/v1/iam/get-app-login` — the canonical source of truth for which
* sign-in buttons (password / GitHub / Google / Web3) to render.
* Resolves to null when the endpoint is unreachable so callers can fall
* back to the tenant's declared default method set.
* back to the org's declared default method set.
*
* `redirectUri` is validated by IAM against the app's registered list. For a
* cross-app SSO read (e.g. console → hanzo.id, `clientId=hanzo-cloud`) pass the
* DOWNSTREAM app's own OIDC `redirect_uri` — the portal's `/callback` is NOT in
* that app's list, so hardcoding it makes IAM answer `status:error`
* ("Redirect URI … doesn't exist in the allowed list") and drops the whole
* response. Omit it for a bare/own-app read (defaults to the portal callback).
*/
getAppLogin(clientId?: string): Promise<AppLogin | null>
getAppLogin(clientId?: string, redirectUri?: string): Promise<AppLogin | null>
/**
* Complete a social provider login when the provider redirects back to
* `/callback` with a `code` + base64 `state` (see `social.ts`). Exchanges the
* provider code at the IAM backend (the Casdoor `AuthBackend.login` contract)
* and resolves the URL to redirect to — the original OIDC `redirect_uri` with
* an authorization code, which the portal's normal PKCE callback then
* completes. NOTE: pending live verification — runs only once real OAuth
* provider creds are seeded (the buttons are hidden until then).
* Resolve the signed-in user's `{owner, name}` from the IAM session
* (`/v1/iam/get-account`). After a `RequiredMfa` login the IAM session cookie
* already authenticates the user (IAM calls `SetSessionUsername` before
* answering `RequiredMfa`), so this is how the portal learns the identity to
* key the forced-enrollment calls on. Resolves null when unauthenticated.
*/
providerLogin(req: ProviderExchangeRequest): Promise<{ redirectUrl?: string; error?: string }>
}
/** Inputs to {@link AuthClient.providerLogin}, recovered from the /callback return. */
export interface ProviderExchangeRequest {
/** IAM application name (from the decoded state). */
readonly application: string
/** IAM provider record name, e.g. `provider-github`. */
readonly provider: string
/** The provider's authorization code (the `?code=` on the /callback return). */
readonly code: string
/** The ORIGINAL OIDC authorize query string (decoded from the base64 state). */
readonly oidcQuery: string
/** "signin" | "signup". */
readonly method: string
getAccount(): Promise<MfaIdentity | null>
/**
* Begin TOTP enrollment: `POST /v1/iam/mfa/setup/initiate`. Returns the secret
* + `otpauth://` URI + recovery codes. Does NOT persist anything — only
* {@link mfaEnable} does.
*/
mfaInitiate(id: MfaIdentity): Promise<MfaSetup>
/** Verify a TOTP code against a pending secret: `POST /v1/iam/mfa/setup/verify`. */
mfaVerify(req: MfaIdentity & { secret: string; passcode: string }): Promise<{ ok: boolean; error?: string }>
/** Persist a verified TOTP enrollment: `POST /v1/iam/mfa/setup/enable`. */
mfaEnable(req: MfaIdentity & { secret: string; recoveryCode: string }): Promise<{ ok: boolean; error?: string }>
/**
* Answer a `NextMfa` challenge: `POST /v1/iam/login` with `{mfaType, passcode}`
* and NO username, riding the MFA session cookie IAM set with `NextMfa`.
* Returns the same shape as {@link login} (a redirect with an auth code for the
* code flow, or a bare-session signal for portal sign-in).
*/
mfaChallenge(req: MfaChallengeRequest): Promise<LoginResponse>
}
export interface AuthClientOptions {
readonly tenant: TenantConfig
readonly org: OrgConfig
/** Override fetch impl (testing). Defaults to global fetch. */
readonly fetchImpl?: typeof fetch
}
export function createAuthClient(opts: AuthClientOptions): AuthClient {
const tenant = opts.tenant
const org = opts.org
const f = opts.fetchImpl ?? fetch
async function login(req: LoginRequest): Promise<LoginResponse> {
const type = req.redirectUri ? 'code' : 'login'
const url = new URL('/v1/iam/login', tenant.iamUrl)
const url = new URL('/v1/iam/login', org.iamUrl)
url.searchParams.set('clientId', req.clientId)
url.searchParams.set('responseType', 'code')
if (req.redirectUri) url.searchParams.set('redirectUri', req.redirectUri)
@@ -146,8 +201,45 @@ export function createAuthClient(opts: AuthClientOptions): AuthClient {
return parseLoginResponse(res, req)
}
// Resolve the org of the user in the ambient IAM session (the `iam_session_id`
// cookie), or null when there is no live session. Reads `/v1/iam/get-account`;
// the org is the `owner` field (IAM returns the User at the top level or
// under `data`). Used to keep silent SSO from reusing a session that belongs
// to a DIFFERENT org than the app being signed into.
async function sessionOwner(): Promise<string | null> {
try {
const res = await f(new URL('/v1/iam/get-account', org.iamUrl).toString(), {
credentials: 'include',
headers: { Accept: 'application/json' },
})
if (!res.ok) return null
const body = (await res.json()) as Record<string, unknown>
if (body.status === 'error') return null
const nested = (typeof body.data === 'object' && body.data ? body.data : {}) as Record<string, unknown>
const owner = typeof body.owner === 'string' ? body.owner : nested.owner
return typeof owner === 'string' && owner ? owner : null
} catch {
return null
}
}
async function silentLogin(req: SilentLoginRequest): Promise<LoginResponse> {
const url = new URL('/v1/iam/login', tenant.iamUrl)
// Silent SSO may reuse the ambient IAM session ONLY when that session's user
// belongs to the SAME org as the app being signed into. A cross-org app —
// e.g. the admin-guard (client_id=hanzo-admin-guard, org=admin) reached from
// a browser that already holds a hanzo/* session — must NOT mint a code from
// the wrong-org session: that confers owner=hanzo and silently shadows the
// org-scoped credential form (which resolves the admin/* identity). Resolve
// the app's org and the session owner; on no session or an org mismatch,
// return an empty response so Login.tsx falls back to the interactive form,
// which authenticates in the app's own org. Same-org SSO (the common case)
// still mints silently, so seamless sign-in is preserved.
const [app, owner] = await Promise.all([getAppLogin(req.clientId), sessionOwner()])
if (!owner) return {}
const appOrg = app?.organization
if (appOrg && owner !== appOrg) return {}
const url = new URL('/v1/iam/login', org.iamUrl)
url.searchParams.set('clientId', req.clientId)
url.searchParams.set('responseType', 'code')
url.searchParams.set('redirectUri', req.redirectUri)
@@ -177,7 +269,7 @@ export function createAuthClient(opts: AuthClientOptions): AuthClient {
async function approveDevice(userCode: string): Promise<DeviceApprovalResult> {
const code = normalizeUserCode(userCode)
if (!code) return { ok: false, error: 'Enter the code shown on your device.' }
const url = new URL('/v1/iam/login', tenant.iamUrl)
const url = new URL('/v1/iam/login', org.iamUrl)
// IAM's device branch keys the cache off the `userCode` in the BODY; the
// `type` echo on the query mirrors the other login legs. NO credentials —
// the user is already signed in, so this rides the session cookie
@@ -186,12 +278,12 @@ export function createAuthClient(opts: AuthClientOptions): AuthClient {
const body: Record<string, unknown> = {
type: 'device',
userCode: code,
application: tenant.appName,
application: org.appName,
}
// `organization` scopes the application lookup (FindApplicationByName); it
// does NOT resolve the user (that comes from the session), so pinning the
// tenant org here is safe — unlike password login, which omits it.
if (tenant.orgId) body.organization = tenant.orgId
// org org here is safe — unlike password login, which omits it.
if (org.orgId) body.organization = org.orgId
let res: Response
try {
res = await f(url.toString(), {
@@ -221,8 +313,54 @@ export function createAuthClient(opts: AuthClientOptions): AuthClient {
return { ok: true }
}
async function deviceInfo(userCode: string): Promise<DeviceInfoResult> {
const code = normalizeUserCode(userCode)
if (!code) return { ok: false, error: 'Enter the code shown on your device.' }
// POST, and the code rides the BODY — like `approveDevice` beside it, and for
// the reason IAM's own introspection endpoint is POST: the user_code is the one
// secret in this flow, and a request line is copied into ingress and proxy
// access logs where a body is not. This page ships `scrubUrl()` to keep the
// code out of the address bar; putting it into every request line would undo
// that server-side. Same session cookie as the approval: whatever you may look
// at is exactly what you may approve.
const url = new URL('/v1/iam/oauth/device/info', org.iamUrl)
let res: Response
try {
res = await f(url.toString(), {
method: 'POST',
headers: { 'Content-Type': 'application/json', Accept: 'application/json' },
credentials: 'include',
body: JSON.stringify({ userCode: code }),
})
} catch (e) {
return { ok: false, error: String(e) }
}
let parsed: Record<string, unknown> = {}
try {
parsed = (await res.json()) as Record<string, unknown>
} catch {
return { ok: false, error: `HTTP ${res.status} non-JSON response` }
}
if (!res.ok || parsed.status === 'error') {
const error = typeof parsed.msg === 'string' && parsed.msg ? parsed.msg : `HTTP ${res.status}`
// IAM `CodeLoginRequired` (internal/oidc/oidc.go): the session lapsed between
// the page's get-account check and this read. Not a dead end — sign in again.
if (parsed.code === 'login_required') return { ok: false, error, loginRequired: true }
return { ok: false, error }
}
// A name is only worth rendering if the server sent it. An answer with no
// clientId names nothing, so it fails rather than letting the page fall back
// to a guess — showing the WRONG application is the defect this endpoint exists
// to fix. `displayName` falls back to the clientId, which IAM did confirm.
const data = parsed.data as Record<string, unknown> | undefined
const clientId = typeof data?.clientId === 'string' ? data.clientId : ''
const displayName = typeof data?.displayName === 'string' ? data.displayName : ''
if (!clientId) return { ok: false, error: 'IAM did not name the application for this code.' }
return { ok: true, clientId, displayName: displayName || clientId }
}
async function signup(req: SignupRequest): Promise<LoginResponse> {
const url = new URL('/v1/iam/signup', tenant.iamUrl)
const url = new URL('/v1/iam/signup', org.iamUrl)
url.searchParams.set('clientId', req.clientId)
const username = req.email.split('@')[0]
const res = await f(url.toString(), {
@@ -237,22 +375,48 @@ export function createAuthClient(opts: AuthClientOptions): AuthClient {
email: req.email,
password: req.password,
confirm: req.password,
autoSignin: true,
...(req.inviteCode ? { invitationCode: req.inviteCode } : {}),
}),
})
return parseLoginResponse(res)
// Registration is CREATE-ONLY at IAM. `/v1/iam/signup` persists the user and
// answers with the created row — it sets no session cookie and mints no
// authorization code, and its form (`internal/oidc/signup.go`) has no
// `autoSignin`, `redirectUri` or `code_challenge` field to make it do so.
// The `autoSignin: true` this used to post was silently dropped by the Go
// decoder, so "signed up" and "signed in" were never the same event.
//
// Left there, the response fell through `parseLoginResponse`'s no-redirect
// arm to `{ redirectUrl: '/onboarding' }` — every new customer was sent to
// the portal's own onboarding, unauthenticated, while the app that sent them
// waited on a code that was never minted. So finish the job here: a signup
// that leaves you logged out is not a signup.
const created = await parseCreated(res)
if (created.error) return created
return login({
identifier: req.email,
password: req.password,
clientId: req.clientId,
application: req.application,
organization: req.organization,
redirectUri: req.redirectUri,
state: req.state,
codeChallenge: req.codeChallenge,
codeChallengeMethod: req.codeChallengeMethod,
nonce: req.nonce,
})
}
async function forgot(req: ForgotRequest): Promise<{ ok: boolean; error?: string }> {
const url = new URL('/v1/iam/send-verification-code', tenant.iamUrl)
const url = new URL('/v1/iam/send-verification-code', org.iamUrl)
url.searchParams.set('clientId', req.clientId)
url.searchParams.set('organization', req.organization)
const res = await f(url.toString(), {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({
applicationId: `admin/${tenant.appName}`,
applicationId: `admin/${org.appName}`,
organization: req.organization,
dest: req.identifier,
type: req.identifier.includes('@') ? 'email' : 'phone',
@@ -267,7 +431,7 @@ export function createAuthClient(opts: AuthClientOptions): AuthClient {
}
function authorize(req: OAuthAuthorizeRequest): string {
const url = new URL('/v1/iam/oauth/authorize', tenant.iamUrl)
const url = new URL('/v1/iam/oauth/authorize', org.iamUrl)
url.searchParams.set('client_id', req.clientId)
url.searchParams.set('redirect_uri', req.redirectUri)
url.searchParams.set('response_type', req.responseType ?? 'code')
@@ -278,16 +442,20 @@ export function createAuthClient(opts: AuthClientOptions): AuthClient {
url.searchParams.set('code_challenge', req.codeChallenge)
url.searchParams.set('code_challenge_method', req.codeChallengeMethod ?? 'S256')
}
// Naming a provider federates the request to that external IdP instead of
// the hosted credential login. The type has always declared this field;
// never emitting it is why social sign-in had no server side at all.
if (req.provider) url.searchParams.set('provider', req.provider)
return url.toString()
}
async function exchange(code: string, codeVerifier?: string): Promise<TokenResponse> {
const url = new URL('/v1/iam/oauth/token', tenant.iamUrl)
const url = new URL('/v1/iam/oauth/token', org.iamUrl)
const body = new URLSearchParams({
grant_type: 'authorization_code',
code,
client_id: tenant.clientId,
redirect_uri: `${tenant.publicOrigin}/callback`,
client_id: org.clientId,
redirect_uri: `${org.publicOrigin}/callback`,
})
if (codeVerifier) body.set('code_verifier', codeVerifier)
const res = await f(url.toString(), {
@@ -308,21 +476,52 @@ export function createAuthClient(opts: AuthClientOptions): AuthClient {
}
function logout(idTokenHint?: string, postLogoutRedirectUri?: string): string {
const url = new URL('/v1/iam/oauth/logout', tenant.iamUrl)
const url = new URL('/v1/iam/oauth/logout', org.iamUrl)
if (idTokenHint) url.searchParams.set('id_token_hint', idTokenHint)
url.searchParams.set(
'post_logout_redirect_uri',
postLogoutRedirectUri ?? `${tenant.publicOrigin}/login`,
postLogoutRedirectUri ?? `${org.publicOrigin}/login`,
)
return url.toString()
}
async function getAppLogin(clientId?: string): Promise<AppLogin | null> {
const id = clientId ?? tenant.clientId
const url = new URL('/v1/iam/get-app-login', tenant.iamUrl)
function signOut(postLogoutRedirectUri?: string): string {
// Every key the SDK owns is namespaced `hanzo_iam_*` — access_token,
// expires_at, state, code_verifier, current_org, current_project,
// post_login_redirect. Sweep the PREFIX rather than naming them: a list of
// literals here is a second copy of the SDK's key set, and the copy is what
// goes stale when the SDK adds one. The prefix is the contract.
//
// Both storages. The token lives in sessionStorage, but the PKCE verifier
// is deliberately in localStorage (it has to survive the full-page redirect
// to the IdP), and apps cache the token there too.
for (const store of [
typeof sessionStorage !== 'undefined' ? sessionStorage : null,
typeof localStorage !== 'undefined' ? localStorage : null,
]) {
if (!store) continue
// Collect first, then delete: removing while iterating by index
// re-indexes the store and skips every other key.
const doomed: string[] = []
for (let i = 0; i < store.length; i++) {
const k = store.key(i)
if (k && k.startsWith('hanzo_iam_')) doomed.push(k)
}
for (const k of doomed) store.removeItem(k)
}
return logout(undefined, postLogoutRedirectUri)
}
async function getAppLogin(clientId?: string, redirectUri?: string): Promise<AppLogin | null> {
const id = clientId ?? org.clientId
const url = new URL('/v1/iam/get-app-login', org.iamUrl)
url.searchParams.set('clientId', id)
url.searchParams.set('responseType', 'code')
url.searchParams.set('redirectUri', `${tenant.publicOrigin}/callback`)
// Validate against the downstream app's OWN redirect_uri when the caller has
// one (the SSO authorize flow carries it); the portal's own /callback is not
// registered for another app, so IAM would reject the read and we'd surface
// no social buttons. Fall back to the portal callback for a bare/own read.
url.searchParams.set('redirectUri', redirectUri || `${org.publicOrigin}/callback`)
url.searchParams.set('scope', 'openid profile email')
url.searchParams.set('state', 'app-login')
let body: Record<string, unknown>
@@ -334,110 +533,157 @@ export function createAuthClient(opts: AuthClientOptions): AuthClient {
return null
}
if (body.status !== 'ok' || typeof body.data !== 'object' || body.data === null) return null
return parseAppLogin(body.data as Record<string, unknown>, tenant.appName, tenant.orgId)
return parseAppLogin(body.data as Record<string, unknown>, org.appName, org.orgId)
}
async function providerLogin(
req: ProviderExchangeRequest,
): Promise<{ redirectUrl?: string; error?: string }> {
// POST the provider code to the IAM backend together with the app's ORIGINAL
// OIDC authorize params (recovered from the round-tripped `state`). IAM
// exchanges the provider code, signs the user in, and mints an authorization
// code BOUND TO THE APP'S request — the app's client_id and, crucially, its
// PKCE `code_challenge` (C1) — which we then hand back to the originating app
// so its OWN callback exchanges the code with ITS verifier (V1). The minted
// code is bound to C1, so that exchange matches; this is the keystone of the
// social-login PKCE round-trip.
const oidc = new URLSearchParams(req.oidcQuery.replace(/^\?/, ''))
const appRedirectUri = oidc.get('redirect_uri') ?? ''
const appState = oidc.get('state') ?? ''
const url = new URL('/v1/iam/login', tenant.iamUrl)
// OIDC params ride the QUERY — IAM's HandleLoggedIn reads them there first
// when minting the code. Forward exactly the app's request so the code
// carries its client_id, scope, nonce and — load-bearing — its
// `code_challenge`. (`redirect_uri` snake-case is NOT forwarded: IAM reads
// the code's redirect binding from camelCase `redirectUri`, set below.)
for (const [k, v] of oidc) {
if (['client_id', 'response_type', 'scope', 'state', 'nonce', 'code_challenge', 'code_challenge_method'].includes(k)) {
url.searchParams.set(k, v)
}
async function getAccount(): Promise<MfaIdentity | null> {
const url = new URL('/v1/iam/get-account', org.iamUrl)
let body: Record<string, unknown>
try {
const res = await f(url.toString(), { headers: { Accept: 'application/json' }, credentials: 'include' })
if (!res.ok) return null
body = (await res.json()) as Record<string, unknown>
} catch {
return null
}
// Bind the minted code to the APP's redirect_uri (camelCase `redirectUri` —
// the param HandleLoggedIn reads), so the code targets the app, NOT the
// provider-callback host carried in the body below.
if (appRedirectUri) url.searchParams.set('redirectUri', appRedirectUri)
const d = (typeof body.data === 'object' && body.data ? body.data : {}) as Record<string, unknown>
if (typeof d.owner !== 'string' || typeof d.name !== 'string' || !d.owner || !d.name) return null
return { owner: d.owner, name: d.name }
}
// The redirect_uri IAM forwards to the PROVIDER's token endpoint MUST be
// byte-identical to the hop's redirect_uri — the provider's REGISTERED
// callback host (`oauthCallbackOrigin`, e.g. `iam.hanzo.ai`, shared across
// brand portals) — or the provider rejects the exchange `invalid_grant`.
// This is a DIFFERENT redirect_uri from the app's above: one drives the
// provider exchange (body), one binds the minted app code (query).
const callbackOrigin = tenant.oauthCallbackOrigin ?? tenant.publicOrigin
/**
* Build a `/v1/iam/mfa/setup/*` POST URL with EVERY param on the query string
* and send an EMPTY body. This is the one wire shape IAM's authz filter and
* the MFA controller both accept: the controller reads `owner`/`name`/… from
* the merged form (query + body), while the authz filter only extracts the
* `{owner,name}` object from the query when the body is empty (a non-empty
* body is JSON-unmarshalled, and a urlencoded body fails that parse → empty
* object → the self-access match `sub==obj` fails → "Unauthorized operation").
* `owner`/`name` ride the query on EVERY call — including `verify`, which
* otherwise carries no identity — purely so that self-access check passes.
*/
async function mfaSetupPost(path: string, params: Record<string, string>): Promise<Record<string, unknown>> {
const url = new URL(`/v1/iam/mfa/setup/${path}`, org.iamUrl)
for (const [k, v] of Object.entries(params)) url.searchParams.set(k, v)
const res = await f(url.toString(), { method: 'POST', credentials: 'include' })
const body = (await res.json().catch(() => ({}))) as Record<string, unknown>
if (typeof body.status === 'string' && body.status === 'error') {
throw new Error(typeof body.msg === 'string' && body.msg ? body.msg : `HTTP ${res.status}`)
}
if (!res.ok) throw new Error(`HTTP ${res.status}`)
return body
}
async function mfaInitiate(id: MfaIdentity): Promise<MfaSetup> {
const body = await mfaSetupPost('initiate', { owner: id.owner, name: id.name, mfaType: MFA_TOTP })
const d = (typeof body.data === 'object' && body.data ? body.data : {}) as Record<string, unknown>
const secret = typeof d.secret === 'string' ? d.secret : ''
const url = typeof d.url === 'string' ? d.url : ''
if (!secret || !url) throw new Error('IAM returned no TOTP secret')
return {
mfaType: MFA_TOTP,
secret,
url,
recoveryCodes: Array.isArray(d.recoveryCodes) ? d.recoveryCodes.filter((c): c is string => typeof c === 'string') : [],
}
}
async function mfaVerify(req: MfaIdentity & { secret: string; passcode: string }): Promise<{ ok: boolean; error?: string }> {
try {
await mfaSetupPost('verify', { owner: req.owner, name: req.name, mfaType: MFA_TOTP, secret: req.secret, passcode: req.passcode })
return { ok: true }
} catch (e) {
return { ok: false, error: e instanceof Error ? e.message : String(e) }
}
}
async function mfaEnable(req: MfaIdentity & { secret: string; recoveryCode: string }): Promise<{ ok: boolean; error?: string }> {
try {
await mfaSetupPost('enable', {
owner: req.owner,
name: req.name,
mfaType: MFA_TOTP,
secret: req.secret,
recoveryCodes: req.recoveryCode,
})
return { ok: true }
} catch (e) {
return { ok: false, error: e instanceof Error ? e.message : String(e) }
}
}
async function mfaChallenge(req: MfaChallengeRequest): Promise<LoginResponse> {
const type = req.redirectUri ? 'code' : 'login'
const url = new URL('/v1/iam/login', org.iamUrl)
url.searchParams.set('clientId', req.clientId)
url.searchParams.set('responseType', 'code')
if (req.redirectUri) url.searchParams.set('redirectUri', req.redirectUri)
url.searchParams.set('scope', 'openid profile email')
if (req.state) url.searchParams.set('state', req.state)
if (req.codeChallenge) {
url.searchParams.set('code_challenge', req.codeChallenge)
url.searchParams.set('code_challenge_method', req.codeChallengeMethod ?? 'S256')
}
url.searchParams.set('type', type)
const res = await f(url.toString(), {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
credentials: 'include',
body: JSON.stringify({
type: 'code',
type,
// No username: IAM resolves the user from the MFA session cookie it set
// when it answered NextMfa.
mfaType: req.mfaType,
passcode: req.passcode,
application: req.application,
provider: req.provider,
code: req.code,
// IAM's social state guard accepts state == application name.
state: req.application,
redirectUri: `${callbackOrigin}/callback`,
// "signup" = find-or-create-LOGIN (see social.ts); never the link branch.
method: req.method,
organization: req.organization,
enableMfaRemember: req.rememberDevice ?? false,
}),
})
let body: Record<string, unknown> = {}
try {
body = (await res.json()) as Record<string, unknown>
} catch {
return { error: `HTTP ${res.status} non-JSON response` }
}
if (!res.ok || body.status === 'error') {
return { error: typeof body.msg === 'string' ? body.msg : `HTTP ${res.status}` }
}
// IAM returns the freshly-minted authorization CODE in `data` (codeToResponse
// → the bare code string, NOT a URL). Build the redirect back to the app's
// own redirect_uri + original state so the app runs the standard OIDC code
// exchange. (Treating the bare code AS the redirect URL — the prior bug —
// dead-ended the flow on the issuer host and never returned to the app.)
const code = typeof body.data === 'string' ? body.data : ''
if (!code) return { error: 'provider login returned no authorization code' }
if (!appRedirectUri) return { error: 'provider login: missing redirect_uri in social state' }
const sep = appRedirectUri.includes('?') ? '&' : '?'
return {
redirectUrl: `${appRedirectUri}${sep}code=${encodeURIComponent(code)}&state=${encodeURIComponent(appState)}`,
}
return parseLoginResponse(res, req)
}
return { tenant, login, silentLogin, approveDevice, signup, forgot, authorize, exchange, logout, getAppLogin, providerLogin }
return {
org,
login,
silentLogin,
approveDevice,
deviceInfo,
signup,
forgot,
authorize,
exchange,
logout,
signOut,
getAppLogin,
getAccount,
mfaInitiate,
mfaVerify,
mfaEnable,
mfaChallenge,
}
}
/**
* Canonicalize a user-entered device code to the form IAM generated. IAM mints
* codes from `[0-9a-z]{6}` (`util.GetRandomName`), so we lowercase (making entry
* case-insensitive), trim surrounding whitespace, and drop spaces/dashes a user
* might add while transcribing. The DeviceAuthMap key is the exact string, so we
* normalize TO that lowercase alphabet — never uppercase.
* user_codes from an UPPERCASE unambiguous alphabet ([A-HJ-NP-Z2-9], no
* I/L/O/0/1) and keys its DeviceAuthMap on the exact string. A human may
* transcribe it lower-cased or with stray spaces/dashes, so normalize TO
* uppercase and strip separators — case-insensitive entry, an exact-match send.
*/
function normalizeUserCode(raw: string): string {
// IAM mints user_codes from an UPPERCASE unambiguous alphabet
// ([A-HJ-NP-Z2-9], no I/L/O/0/1) and keys its DeviceAuthMap on the exact
// string. A human may transcribe it lower-cased or with stray spaces/dashes,
// so normalize TO uppercase and strip separators — case-insensitive entry,
// an exact-match send.
return raw.trim().toUpperCase().replace(/[\s-]+/g, '')
}
/**
* Map an IAM provider record to its canonical authorize-endpoint `provider`
* key. IAM names providers `provider-<key>` (e.g. `provider-github`); the
* `/v1/iam/oauth/authorize?provider=<key>` param wants the bare key. The
* Web3Onboard wallet provider maps to `web3`.
* The provider's DISPLAY key — `provider-github` → `github` — used to pick an
* icon and a label (`PROVIDER_META`) and to match a `provider_hint`.
*
* It is NOT what the authorize endpoint wants. `federationProvider` matches the
* record name exactly, so `?provider=` must carry the full `provider-github`;
* live, `?provider=github` is refused "unknown or unavailable provider". This
* comment used to assert the opposite — a bare key — which was never true of the
* federation broker.
*/
function providerKey(name: string): string {
return name.replace(/^provider-/, '')
@@ -469,7 +715,7 @@ function parseAppLogin(
// (`rec.provider.name`, e.g. `provider-github`) — that is what the IAM
// backend's social-login lookup (`GetProvider(admin/<name>)`) resolves.
// The OUTER link object's `name` is the app's provider-LINK label, which
// some Casdoor seeds set to a per-app default (e.g. `<org>-iam`); reading
// some IAM seeds set to a per-app default (e.g. `<org>-iam`); reading
// it as the provider name made the hop POST `provider=<org>-iam`, which
// the backend rejects ("The provider: <org>-iam does not exist"). Prefer
// the inner record name; fall back to the outer label only when there is
@@ -505,6 +751,28 @@ function parseAppLogin(
}
}
/**
* Read a create-only IAM response: `{status, msg, data}` where `data` is the
* created row. Success carries nothing the caller can navigate to, so this
* reports only whether it worked — never a redirect. Kept separate from
* `parseLoginResponse` precisely because that one INVENTS a destination when no
* `redirectUri` was requested, which is wrong for a row that is not a session.
*/
async function parseCreated(res: Response): Promise<{ error?: string }> {
let body: Record<string, unknown> = {}
try {
body = (await res.json()) as Record<string, unknown>
} catch {
return { error: `HTTP ${res.status} non-JSON response` }
}
// IAM answers a REFUSAL with HTTP 200 + status:"error" (see the org-less login
// note in this repo's LLM.md), so the status code alone proves nothing.
if (!res.ok || body.status === 'error') {
return { error: typeof body.msg === 'string' ? body.msg : `HTTP ${res.status}` }
}
return {}
}
async function parseLoginResponse(
res: Response,
req?: { redirectUri?: string; state?: string },
@@ -520,6 +788,25 @@ async function parseLoginResponse(
}
const data = body.data
// Multi-factor signal — IAM answers a successful credential check with a
// STRING in `data` (NOT a `mfa_required` boolean): `"RequiredMfa"` when org
// policy forces MFA the user has not enrolled, `"NextMfa"` when the user has
// MFA and must answer a challenge. Branch BEFORE any session/redirect return:
// the password session is not yet usable, so the portal must render the
// enrollment/challenge step rather than navigate on.
if (data === 'RequiredMfa') {
return { mfaRequired: true, mfaStage: 'enroll' }
}
if (data === 'NextMfa') {
// Challenge allow-list: IAM's named `mfa` field first, falling back to
// the legacy untyped `data2` slot until IAM stops emitting it.
const allow = Array.isArray(body.mfa) ? body.mfa : Array.isArray(body.data2) ? body.data2 : []
const mfaTypes = allow
.map((p) => (typeof p === 'object' && p !== null ? (p as Record<string, unknown>).mfaType : undefined))
.filter((t): t is string => typeof t === 'string' && t.length > 0)
return { mfaRequired: true, mfaStage: 'challenge', mfaTypes }
}
// Authorization-code flow: a client redirectUri is present and `data` is the
// freshly minted code — hand the SPA a fully-formed redirect back to the app.
if (req?.redirectUri && typeof data === 'string' && data.length > 0) {
@@ -544,7 +831,5 @@ async function parseLoginResponse(
refreshToken: typeof d.refresh_token === 'string' ? d.refresh_token : undefined,
idToken: typeof d.id_token === 'string' ? d.id_token : undefined,
expiresAt: typeof d.expires_at === 'number' ? d.expires_at : undefined,
mfaRequired: d.mfa_required === true,
mfaChannel: typeof d.mfa_channel === 'string' ? (d.mfa_channel as LoginResponse['mfaChannel']) : undefined,
}
}
+6 -6
View File
@@ -1,8 +1,8 @@
import type { TenantConfig } from '@hanzo/id-shared'
import type { OrgConfig } from '@hanzo/id-shared'
import { IAM } from '@hanzo/iam/browser'
/**
* One IAM browser-SDK instance per tenant, wired to the portal's own
* One IAM browser-SDK instance per org, wired to the portal's own
* `/callback` route. This is the single place that constructs the PKCE
* client — social/web3 sign-in (here) and the callback handler
* (`Callback.tsx`) share it so the PKCE verifier/state the SDK stores on
@@ -11,11 +11,11 @@ import { IAM } from '@hanzo/iam/browser'
* The portal is its own OIDC client (`clientId` = the brand `-id` app), so
* every flow it initiates lands back at `${publicOrigin}/callback`.
*/
export function createIam(tenant: TenantConfig, clientId?: string): IAM {
export function createIam(org: OrgConfig): IAM {
return new IAM({
serverUrl: tenant.iamUrl,
clientId: clientId ?? tenant.clientId,
redirectUri: `${tenant.publicOrigin}/callback`,
serverUrl: org.iamUrl,
clientId: org.clientId,
redirectUri: `${org.publicOrigin}/callback`,
scope: 'openid profile email',
})
}
+15 -9
View File
@@ -1,22 +1,27 @@
export { createAuthClient, type AuthClient, type AuthClientOptions } from './client'
export { createIam } from './iam'
export {
startProviderLogin,
buildProviderAuthUrl,
isHoppableProvider,
encodeState,
decodeState,
type ProviderLoginParams,
} from './social'
createAuthClient,
mfaChannelOf,
MFA_TOTP,
type AuthClient,
type AuthClientOptions,
} from './client'
export { createIam } from './iam'
export { authorizeRequest, matchProviderHint } from './social'
export {
loginWithWalletChain,
detectWalletChains,
ENABLED_WALLET_CHAINS,
WALLET_CHAIN_LABELS,
type WalletLoginContext,
type WalletWindow,
} from './web3'
export type {
LoginRequest,
LoginResponse,
MfaChannel,
MfaChallengeRequest,
MfaIdentity,
MfaSetup,
SignupRequest,
ForgotRequest,
OAuthAuthorizeRequest,
@@ -24,5 +29,6 @@ export type {
AppLogin,
AppProvider,
DeviceApprovalResult,
DeviceInfoResult,
} from './types'
export * from './ui'
+87 -96
View File
@@ -1,115 +1,106 @@
/**
* Provider-hop URL builder tests — pure, no network. Run with:
* Federated sign-in — pure unit tests, no network. Run with:
* pnpm --filter @hanzo/id-auth test
*
* Verifies the URL + base64 state match the Hanzo-IAM (Casdoor) `getAuthUrl`
* contract so the backend `/callback` exchange accepts the return. The
* end-to-end OAuth round-trip still needs live verification once real provider
* creds are seeded — but the URL/state construction is locked down here.
* The browser's whole job in a federated sign-in is to name the provider on
* IAM's authorize endpoint and, when an app sent the user here, to hand that
* app's own request back unchanged so IAM mints the code against it. Those two
* are what these tests pin; the IdP leg belongs to IAM and is not modelled here.
*/
import { test } from 'node:test'
import { test } from 'vitest'
import assert from 'node:assert/strict'
import { buildProviderAuthUrl, isHoppableProvider } from './social.ts'
import { authorizeRequest, matchProviderHint, PROVIDER_ORDER } from './social.ts'
const ORIGIN = 'https://hanzo.id'
// The original OIDC authorize query the portal was bounced here with.
const SEARCH = '?client_id=hanzo-id&redirect_uri=https%3A%2F%2Fhanzo.id%2Fcallback&response_type=code&scope=openid&state=rp123'
const PORTAL = 'hanzo-console'
test('GitHub hop builds the correct endpoint, client_id, redirect_uri, and scope', () => {
const url = buildProviderAuthUrl(
{ application: 'hanzo-id', providerName: 'provider-github', type: 'GitHub', clientId: 'gh_real_123' },
ORIGIN,
SEARCH,
test('Google is offered above GitHub, and the wallet last', () => {
// Order is a deliberate product decision, not an accident of how the buttons
// were typed out. It lived as an unexported constant inside the component and
// had already drifted once with nothing to catch it, which is the whole reason
// it is a value in this module now.
const at = (k: string) => PROVIDER_ORDER.indexOf(k as (typeof PROVIDER_ORDER)[number])
assert.ok(at('google') < at('github'), 'Google leads')
assert.ok(at('github') < at('gitlab'))
assert.equal(PROVIDER_ORDER[PROVIDER_ORDER.length - 1], 'web3', 'the wallet trails')
})
test('every ordered provider is one the hint matcher can also resolve', () => {
// The two provider policies in this module must agree: a key the strip renders
// must be a key a `provider_hint` can name, or the console's one-click hand-off
// silently falls back to the form for a provider that is plainly on screen.
for (const key of PROVIDER_ORDER) {
const found = matchProviderHint([{ name: `provider-${key}`, key }], `provider-${key}`)
assert.equal(found?.key, key)
}
})
test('an app-initiated request is recovered whole, so IAM binds the code to that app', () => {
// What IAM forwards to the hosted login (authorizeForwardQuery) when an app
// sends a user here for a code.
const req = authorizeRequest(
'?client_id=hanzo-app&redirect_uri=https%3A%2F%2Fhanzo.app%2Fcallback&response_type=code' +
'&scope=openid+profile&state=rp123&nonce=n1&code_challenge=C1&code_challenge_method=S256',
PORTAL,
)!
assert.ok(url.startsWith('https://github.com/login/oauth/authorize?'))
assert.ok(url.includes('client_id=gh_real_123'))
// No callbackOrigin → defaults to the browser origin.
assert.ok(url.includes('redirect_uri=https://hanzo.id/callback'))
assert.ok(url.includes('scope=user:email+read:user')) // GitHub default
assert.ok(url.includes('response_type=code'))
assert.equal(req.clientId, 'hanzo-app')
assert.equal(req.redirectUri, 'https://hanzo.app/callback')
assert.equal(req.state, 'rp123')
assert.equal(req.scope, 'openid profile')
assert.equal(req.nonce, 'n1')
// Load-bearing: the code IAM mints is bound to the APP's challenge, so the
// app's own callback completes the exchange with the verifier it kept.
assert.equal(req.codeChallenge, 'C1')
assert.equal(req.codeChallengeMethod, 'S256')
})
test('the registered callback origin overrides the browser origin in redirect_uri', () => {
// The shared OAuth client is registered against iam.hanzo.ai/callback, so the
// hop must return there even though the SPA runs on hanzo.id — otherwise the
// provider rejects the redirect_uri (verified live: Google accepts ONLY
// https://iam.hanzo.ai/callback for this client).
const url = buildProviderAuthUrl(
{ application: 'hanzo-id', providerName: 'provider-google', type: 'Google', clientId: 'goog_1' },
ORIGIN,
SEARCH,
'https://iam.hanzo.ai',
)!
assert.ok(url.includes('redirect_uri=https://iam.hanzo.ai/callback'))
assert.ok(!url.includes('redirect_uri=https://hanzo.id/callback'))
test('a bare portal sign-in has no app to return to', () => {
// No redirect_uri → nothing to return a code to, so the portal starts its own
// PKCE flow instead (the SDK owns the verifier; Callback reads it back).
assert.equal(authorizeRequest('', PORTAL), null)
assert.equal(authorizeRequest('?provider_hint=provider-github', PORTAL), null)
})
test('state base64-encodes the original OIDC query + application/provider/method (round-trips)', () => {
const url = buildProviderAuthUrl(
{ application: 'hanzo-id', providerName: 'provider-github', type: 'GitHub', clientId: 'gh_real_123', method: 'signup' },
ORIGIN,
SEARCH,
)!
const state = new URL(url).searchParams.get('state')!
const decoded = Buffer.from(state, 'base64').toString('utf8')
// The RP's original request survives so the backend can complete it.
assert.ok(decoded.includes('client_id=hanzo-id'))
assert.ok(decoded.includes('state=rp123'))
assert.ok(decoded.includes('application=hanzo-id'))
assert.ok(decoded.includes('provider=provider-github'))
assert.ok(decoded.includes('method=signup'))
test('the portal client id is the fallback, never an override', () => {
const own = authorizeRequest('?redirect_uri=https%3A%2F%2Fhanzo.id%2Fcallback', PORTAL)!
assert.equal(own.clientId, PORTAL, 'no client_id on the query → the portal is the client')
const app = authorizeRequest('?client_id=hanzo-app&redirect_uri=https%3A%2F%2Fhanzo.app%2Fcallback', PORTAL)!
assert.equal(app.clientId, 'hanzo-app', "the app's own client_id wins — the code is minted for IT")
})
test('a pre-existing provider= in the upstream query is stripped — state carries exactly ONE provider', () => {
// The console→hanzo.id SSO SDK appends `provider=hanzo-iam` (its per-org IDP
// hint) to the upstream authorize query. The hop appends the REAL social
// provider; the upstream one MUST be stripped, because `Callback` recovers the
// provider with `URLSearchParams.get` (the FIRST match) — two `provider=`
// params would make it post `hanzo-iam`, which the IAM backend rejects.
const searchWithHint =
'?client_id=hanzo-console&redirect_uri=https%3A%2F%2Fiam.hanzo.ai%2Fcallback&response_type=code&scope=openid&state=rp123&provider=hanzo-iam'
const url = buildProviderAuthUrl(
{ application: 'hanzo-console', providerName: 'provider-google', type: 'Google', clientId: 'goog_1' },
ORIGIN,
searchWithHint,
'https://iam.hanzo.ai',
)!
const state = new URL(url).searchParams.get('state')!
const decoded = Buffer.from(state, 'base64').toString('utf8')
const params = new URLSearchParams(decoded.replace(/^\?/, ''))
// Exactly one provider, and it is the real social one (not the upstream hint).
assert.deepEqual(params.getAll('provider'), ['provider-google'])
assert.equal(params.get('provider'), 'provider-google') // FIRST match = the social provider
assert.ok(!decoded.includes('hanzo-iam')) // the upstream hint is gone entirely
// The rest of the upstream OIDC request is preserved so the backend completes it.
assert.ok(decoded.includes('client_id=hanzo-console'))
assert.ok(decoded.includes('state=rp123'))
test('a leading ? is optional and absent params stay absent', () => {
const withMark = authorizeRequest('?redirect_uri=https%3A%2F%2Fhanzo.id%2Fcallback', PORTAL)!
const without = authorizeRequest('redirect_uri=https%3A%2F%2Fhanzo.id%2Fcallback', PORTAL)!
assert.deepEqual(withMark, without)
// Undefined, not '' — client.authorize omits a param it was not given, and an
// empty code_challenge is not the same request as no code_challenge.
assert.equal(withMark.codeChallenge, undefined)
assert.equal(withMark.nonce, undefined)
assert.equal(withMark.scope, undefined)
assert.equal(withMark.state, '', 'state is always sent, empty when the app sent none')
})
test('Google uses its own endpoint + scope; a custom provider scope overrides', () => {
const g = buildProviderAuthUrl(
{ application: 'hanzo-id', providerName: 'provider-google', type: 'Google', clientId: 'goog_1' },
ORIGIN,
SEARCH,
)!
assert.ok(g.startsWith('https://accounts.google.com/o/oauth2/v2/auth?'))
assert.ok(g.includes('scope=profile+email'))
const custom = buildProviderAuthUrl(
{ application: 'hanzo-id', providerName: 'provider-github', type: 'GitHub', clientId: 'gh_1', scopes: 'repo+user' },
ORIGIN,
SEARCH,
)!
assert.ok(custom.includes('scope=repo+user'))
test('only the two PKCE methods RFC 7636 defines are carried through', () => {
const base = 'redirect_uri=https%3A%2F%2Fhanzo.id%2Fcallback&code_challenge=C1&code_challenge_method='
assert.equal(authorizeRequest(base + 'S256', PORTAL)!.codeChallengeMethod, 'S256')
assert.equal(authorizeRequest(base + 'plain', PORTAL)!.codeChallengeMethod, 'plain')
// Anything else is dropped rather than forwarded, so client.authorize applies
// its S256 default instead of asking IAM to honor a method it does not define.
assert.equal(authorizeRequest(base + 'md5', PORTAL)!.codeChallengeMethod, undefined)
})
test('an unconfigured (empty clientId) or unknown provider type yields no URL', () => {
assert.equal(buildProviderAuthUrl({ application: 'a', providerName: 'p', type: 'GitHub', clientId: '' }, ORIGIN, SEARCH), null)
assert.equal(buildProviderAuthUrl({ application: 'a', providerName: 'p', type: 'Mystery', clientId: 'x' }, ORIGIN, SEARCH), null)
})
test('isHoppableProvider knows the OAuth set, not wallet', () => {
assert.equal(isHoppableProvider('GitHub'), true)
assert.equal(isHoppableProvider('Google'), true)
assert.equal(isHoppableProvider('Web3Onboard'), false)
test('matchProviderHint resolves the console hint, the bare key, and case, else undefined', () => {
const providers = [
{ name: 'provider-github', key: 'github' },
{ name: 'provider-google', key: 'google' },
]
// The console sends the IAM record name verbatim (`provider-github`).
assert.equal(matchProviderHint(providers, 'provider-github')?.key, 'github')
assert.equal(matchProviderHint(providers, 'provider-google')?.key, 'google')
// The bare key and any case also resolve, so the two sides need no shared constant.
assert.equal(matchProviderHint(providers, 'github')?.key, 'github')
assert.equal(matchProviderHint(providers, 'GitHub')?.key, 'github')
// A hint for a provider this app doesn't offer, or an empty hint, matches nothing.
assert.equal(matchProviderHint(providers, 'provider-apple'), undefined)
assert.equal(matchProviderHint(providers, ''), undefined)
})
+90 -129
View File
@@ -1,144 +1,105 @@
import type { OAuthAuthorizeRequest } from './types'
/**
* Social provider redirect — the "hop" that sends the browser to GitHub /
* Google / … to start an OAuth login, replicating the Hanzo-IAM (Casdoor)
* front-end `Provider.getAuthUrl` contract so the IAM backend's `/callback`
* exchange accepts the return.
* Federated sign-in — the portal's half of IAM identity federation.
*
* Why this exists: the IAM backend's OIDC authorize endpoint, for an
* unauthenticated request, 302s to its OWN login-page route (`/login/oauth/
* authorize`) and expects the FRONT END to read `?provider=` and bounce to the
* provider. We replaced that front end with this portal, so the portal must do
* the bounce. `iam.signinRedirect({provider})` does NOT — it just re-enters the
* authorize endpoint and loops.
* IAM is the relying party; this SPA is not. `/v1/iam/oauth/authorize?provider=…`
* IS the entry point: having already validated the client_id, the EXACT
* redirect_uri and the PKCE policy, IAM stashes the app-leg request server-side,
* sets a single-use browser-binding cookie and sends the browser to the IdP
* (`internal/oidc/federation.go::beginFederation`). The IdP returns to IAM's own
* fixed callback — `/v1/iam/oauth/callback`, never a route in this SPA — where
* IAM, which holds the client SECRET a browser cannot, exchanges the code, links
* or provisions the user, and mints an IAM authorization code bound to the
* original PKCE challenge, redirect_uri and nonce. The ordinary code→token
* exchange then completes unchanged.
*
* Contract (from `web/src/auth/Provider.tsx::getAuthUrl` + `Util.tsx::
* getStateFromQueryParams` in the IAM fork):
* url = `${endpoint}?client_id=${clientId}&redirect_uri=${origin}/callback`
* `&scope=${scope}&response_type=code&state=${state}`
* state = encodeState(`${window.location.search}&application=${app}&provider=`
* `${providerName}&method=${method}`) // URL-SAFE base64 of the
* // ORIGINAL OIDC query + app/provider/method, so the backend recovers
* // the original request when the provider returns to /callback.
* This file used to build the IdP URL here in the browser, replicating a contract
* from an IAM fork whose front end no longer exists. Nothing could ever finish it:
* the SPA has no client secret and IAM has no endpoint that exchanges a raw
* provider code, so GitHub returned to `/callback` with a code nobody could spend
* and the flow died there. The browser's only job is to NAME the provider.
*
* Only the standard OAuth2 set is wired here (the providers a `-id` app
* actually enables: github, google, +web3 handled elsewhere). Apple uses the
* backend callback and is added when needed.
*
* NOTE: live-verify this end-to-end once real OAuth credentials are seeded —
* it cannot be exercised while every provider carries placeholder creds (the
* buttons are hidden until then; see SocialButtons + AppProvider.configured).
* The name is the IAM provider RECORD name (`provider-github`), never the bare
* key: `federationProvider` matches `ProviderItem.Name` exactly, and
* `EnrichProviders` resolves that same name to the record, so the two are one
* string by construction. Verified live — `provider=github` is refused with
* "unknown or unavailable provider"; `provider=provider-github` redirects to
* GitHub.
*/
/** Provider `type` → OAuth2 authorize endpoint + default scope (IAM `authInfo`). */
const AUTH_INFO: Record<string, { endpoint: string; scope: string }> = {
GitHub: { endpoint: 'https://github.com/login/oauth/authorize', scope: 'user:email+read:user' },
// Canonical Google OAuth2 authorize endpoint. (Google aliases the legacy
// `/signin/oauth` path, but `/o/oauth2/v2/auth` is the documented, stable one.)
Google: { endpoint: 'https://accounts.google.com/o/oauth2/v2/auth', scope: 'profile+email' },
}
export interface ProviderLoginParams {
/** IAM application name the portal authenticates as (e.g. `hanzo-id`). */
readonly application: string
/** IAM provider record name, e.g. `provider-github`. */
readonly providerName: string
/** IAM provider `type`, e.g. `GitHub` / `Google` (selects the endpoint). */
readonly type: string
/** The provider's real OAuth client id (from `get-app-login`). */
readonly clientId: string
/** Override scope; falls back to the type default. */
readonly scopes?: string
/**
* IAM social-auth method. Defaults to "signup" — the find-or-create-LOGIN
* branch (Casdoor canonical). "signin" is the account-LINK branch and needs
* an existing session, so it is NOT used for interactive provider sign-in.
*/
readonly method?: 'signin' | 'signup'
/**
* The authorize request this portal is standing in for, read from its own URL.
*
* When an app sends a user here for a code, IAM forwards that app's validated
* request on the query (`authorizeForwardQuery`). Re-entering authorize with it —
* plus `provider` — is what makes IAM mint the code against THAT app: its
* client_id, its redirect_uri, its PKCE challenge. The browser is returned
* straight to the app, so this portal's own `/callback` never runs and no token
* is ever handed across on a URL.
*
* Null when there is no app to return to (a bare portal sign-in), which is the
* signal to start the portal's own PKCE flow instead. Keyed on `redirect_uri`
* because that is the one parameter that makes a request returnable — the same
* condition the password path branches on (`Login.completeAfterAuth`).
*
* The result is an {@link OAuthAuthorizeRequest} because that is exactly what
* `client.authorize` consumes: one type, read and written in one shape.
*/
export function authorizeRequest(search: string, clientId: string): OAuthAuthorizeRequest | null {
const q = new URLSearchParams(search.startsWith('?') ? search.slice(1) : search)
const redirectUri = q.get('redirect_uri')
if (!redirectUri) return null
const method = q.get('code_challenge_method')
return {
clientId: q.get('client_id') || clientId,
redirectUri,
state: q.get('state') ?? '',
scope: q.get('scope') ?? undefined,
nonce: q.get('nonce') ?? undefined,
codeChallenge: q.get('code_challenge') ?? undefined,
codeChallengeMethod: method === 'plain' || method === 'S256' ? method : undefined,
}
}
/**
* URL-safe base64 (RFC 4648 §5) for the round-tripped `state`.
* The order federated providers are offered in, most-used first.
*
* The provider reflects `state` back on a URL query string. Standard base64
* (`btoa`) emits `+` `/` `=`, and on the return `URLSearchParams` turns `+`
* into a space (then `atob` strips it) — silently CORRUPTING the encoded OIDC
* request, including the `code_challenge`. A corrupted, non-empty challenge is
* exactly what makes the app's later token exchange fail with
* `invalid_grant: code_verifier does not match code_challenge`. Encoding the
* state with the URL-safe alphabet (and decoding it the same way in `Callback`)
* makes the round-trip byte-exact. The payload is ASCII (an OAuth query), so a
* plain `btoa`/`atob` core is sufficient.
* Order is PROVIDER POLICY, so it lives here beside the other provider policy
* rather than in the component that happens to paint the buttons — and being a
* value in a pure module, it is the only part of the button strip a test can
* actually hold still. It had drifted before: an unexported constant inside the
* component, covered by nothing.
*
* Google leads because it is the account most people arrive already signed into,
* and the wallet trails because it is the specialist entry. A provider absent
* from this list does not render at all, which is deliberate: a name here is the
* statement that the portal knows how to draw and finish that provider's flow.
*/
export function encodeState(raw: string): string {
return btoa(raw).replace(/\+/g, '-').replace(/\//g, '_').replace(/=+$/, '')
}
/** Inverse of {@link encodeState}; tolerant of standard or URL-safe input. */
export function decodeState(state: string): string {
let b64 = state.replace(/-/g, '+').replace(/_/g, '/')
while (b64.length % 4) b64 += '='
return atob(b64)
}
export const PROVIDER_ORDER = ['google', 'github', 'gitlab', 'web3'] as const
/**
* Build the provider authorize URL (pure; testable without navigating).
*
* `callbackOrigin` is the origin of the `/callback` that MUST be registered as
* the provider's authorized redirect URI. The OAuth client (one per provider)
* is registered against a SINGLE callback host — the IAM backend host
* (`iam.hanzo.ai`) — shared across every brand portal, so the provider only
* accepts that exact `redirect_uri`. Sending the browser's own origin (e.g.
* `hanzo.id`) yields `redirect_uri_mismatch`. Callers pass the registered
* origin; it defaults to `origin` for the single-host / local-dev case.
*
* `iam.hanzo.ai/callback` serves the SAME `@hanzo/id` SPA (the headless
* `Callback` page — no login UI), which decodes the base64 `state` to recover
* the original app's `redirect_uri`, exchanges the provider `code` at the IAM
* backend, and forwards the browser back to the originating app.
* Resolve a `provider_hint` from the authorize query to one of the app's
* configured providers. A client that already knows which provider the user
* chose (the console passes `?provider_hint=provider-github` when a user clicks
* "Continue with GitHub" over there) sends the hint so this portal launches that
* provider straight away — no second button press, no bounce through a login
* page. Accepts the IAM record name (`provider-github`), the normalized key
* (`github`), or the record name with the `provider-` prefix stripped, so the
* two sides agree without a shared constant. Returns undefined when nothing
* matches (the caller falls back to the interactive form).
*/
export function buildProviderAuthUrl(
p: ProviderLoginParams,
origin: string,
search: string,
callbackOrigin: string = origin,
): string | null {
const info = AUTH_INFO[p.type]
if (!info || !p.clientId) return null
const scope = p.scopes && p.scopes.trim() !== '' ? p.scopes : info.scope
const redirectUri = `${callbackOrigin}/callback`
// IAM's social branch does FIND-OR-CREATE-LOGIN only under method `signup`
// (the canonical Casdoor default — web `Util.tsx` getEvent → getAuthUrl(...,
// "signup")). Any other value (incl. `signin`) takes the account-LINK branch,
// which requires an EXISTING session and 400s a fresh "Continue with Google".
const method = p.method ?? 'signup'
// The console SSO SDK appends a unified `provider=<org>-iam` hint to the
// upstream authorize query (`search`). We append the REAL social provider
// below, so strip any pre-existing `provider=` first — otherwise the state
// carries TWO `provider=` params and the /callback exchange resolves the
// wrong one (`<org>-iam`, which IAM rejects). One provider, one source of
// truth — don't rely on "backend reads the last param".
const baseQ = new URLSearchParams(search.startsWith('?') ? search.slice(1) : search)
baseQ.delete('provider')
const baseSearch = `?${baseQ.toString()}`
// Base64 of the original OIDC query + routing — the backend decodes this on
// the /callback return to complete the original authorize request.
const state = encodeState(`${baseSearch}&application=${encodeURIComponent(p.application)}&provider=${encodeURIComponent(p.providerName)}&method=${method}`)
return `${info.endpoint}?client_id=${p.clientId}&redirect_uri=${redirectUri}&scope=${scope}&response_type=code&state=${state}`
}
/** True when this portal knows how to start an OAuth hop for the given type. */
export function isHoppableProvider(type: string): boolean {
return type in AUTH_INFO
}
/**
* Redirect the browser to the provider to begin login. No-op return on bad input.
*
* `callbackOrigin` (the provider's registered redirect host, e.g.
* `https://iam.hanzo.ai`) defaults to the current origin when omitted.
*/
export function startProviderLogin(p: ProviderLoginParams, callbackOrigin?: string): void {
if (typeof window === 'undefined') return
const url = buildProviderAuthUrl(p, window.location.origin, window.location.search, callbackOrigin ?? window.location.origin)
if (url) window.location.assign(url)
export function matchProviderHint<P extends { name: string; key: string }>(
providers: Iterable<P>,
hint: string,
): P | undefined {
const h = hint.trim().toLowerCase()
if (h === '') return undefined
const bare = h.replace(/^provider-/, '')
for (const p of providers) {
const name = p.name.toLowerCase()
const key = p.key.toLowerCase()
if (name === h || key === h || key === bare) return p
}
return undefined
}
+98 -4
View File
@@ -17,7 +17,7 @@ export interface LoginRequest {
* session) while a brand-only identity resolves to its own brand org. This is
* why the portal does NOT pin the brand org here — pinning `hanzo` would
* resolve a colliding `hanzo/<name>` row and truncate a global admin to one
* org. Set it only to FORCE a specific tenant (e.g. a brand that deliberately
* org. Set it only to FORCE a specific org (e.g. a brand that deliberately
* scopes its portal to a single org). Signup, by contrast, MUST carry a
* concrete org (you cannot create a user in "no org").
*/
@@ -62,14 +62,33 @@ export interface SilentLoginRequest {
readonly nonce?: string
}
/** A multi-factor channel the portal can render a code entry for. */
export type MfaChannel = 'totp' | 'sms' | 'email'
export interface LoginResponse {
readonly accessToken?: string
readonly refreshToken?: string
readonly idToken?: string
readonly expiresAt?: number
readonly redirectUrl?: string
/**
* Set when IAM answered the login with a multi-factor signal instead of a
* session/code. `mfaStage` discriminates the two IAM states:
* - `'enroll'` — IAM returned `data:"RequiredMfa"`: org policy forces MFA
* and the user has none yet → render forced TOTP enrollment.
* - `'challenge'` — IAM returned `data:"NextMfa"`: the user has MFA enabled
* → render a code challenge for one of `mfaTypes`.
* The password session is NOT established until the enrollment/challenge
* completes, so the portal must not navigate past this signal.
*/
readonly mfaRequired?: boolean
readonly mfaChannel?: 'totp' | 'sms' | 'email'
readonly mfaStage?: 'enroll' | 'challenge'
/**
* The IAM MFA types available for a `'challenge'` (from the login response's
* named `mfa` field, legacy `data2`), in IAM's own vocabulary: `app` (TOTP),
* `sms`, `email`. Empty for enrollment.
*/
readonly mfaTypes?: readonly string[]
readonly error?: string
}
@@ -89,13 +108,88 @@ export interface DeviceApprovalResult {
readonly error?: string
}
/**
* WHICH application a pending device code belongs to
* ({@link AuthClient.deviceInfo}) — the one thing the approval page exists to
* tell a human, and the one thing it cannot know on its own.
*
* Both fields come off the device code's own application row, so a page that
* renders them names the party it is actually authorizing. They are the ONLY
* honest source: `org.appName` is this portal's static branding and names the
* wrong app for every code minted by anything else.
*
* Discriminated on `ok` so a caller cannot read `displayName` without having
* proved the server confirmed one. `loginRequired` singles out the expired
* session (IAM `code:"login_required"`) — the page's cue to sign the human in
* and come back, not an error to show. Every other failure is IAM's single
* opaque refusal, surfaced verbatim.
*/
export type DeviceInfoResult =
| { readonly ok: true; readonly clientId: string; readonly displayName: string }
| { readonly ok: false; readonly error: string; readonly loginRequired?: boolean }
/**
* The TOTP enrollment material minted by `/v1/iam/mfa/setup/initiate`. The
* secret + `url` (an `otpauth://` URI) are rendered locally as a QR code — the
* secret never leaves the browser to a third party. `recoveryCodes[0]` must be
* echoed back to `/v1/iam/mfa/setup/enable`.
*/
export interface MfaSetup {
/** IAM MFA type — `app` for TOTP. */
readonly mfaType: string
/** Base32 TOTP secret. */
readonly secret: string
/** `otpauth://totp/...` provisioning URI for the authenticator app. */
readonly url: string
/** One-time recovery codes issued alongside the secret. */
readonly recoveryCodes: readonly string[]
}
/** The signed-in user's identity, resolved from the IAM session for MFA setup. */
export interface MfaIdentity {
readonly owner: string
readonly name: string
}
/** A TOTP challenge submission for a user who already enrolled (`NextMfa`). */
export interface MfaChallengeRequest {
/** IAM MFA type, e.g. `app` (TOTP), `sms`, `email`. */
readonly mfaType: string
readonly passcode: string
readonly clientId: string
readonly application: string
readonly organization: string
readonly redirectUri?: string
readonly state?: string
readonly codeChallenge?: string
readonly codeChallengeMethod?: 'S256' | 'plain'
/** Honor the org's "remember this device" window after a successful code. */
readonly rememberDevice?: boolean
}
export interface SignupRequest {
readonly email: string
readonly password: string
readonly clientId: string
readonly application: string
/**
* The org to create the user in. REQUIRED — unlike login's optional
* lookup hint, you cannot create a user in "no org", and IAM gates this
* against the application's own org.
*/
readonly organization: string
readonly inviteCode?: string
/**
* The downstream OIDC request, when an app sent the user here to register.
* Registration completes by signing the new user in, so these are forwarded
* to that sign-in: without them the minted code carries no PKCE binding and
* there is nowhere to return the user to.
*/
readonly redirectUri?: string
readonly state?: string
readonly codeChallenge?: string
readonly codeChallengeMethod?: 'S256' | 'plain'
readonly nonce?: string
}
export interface ForgotRequest {
@@ -121,9 +215,9 @@ export interface OAuthAuthorizeRequest {
export interface ProviderInfo {
readonly name: string
readonly displayName?: string
/** Casdoor provider type, e.g. GitHub, Google, Apple, Web3Onboard. */
/** IAM provider type, e.g. GitHub, Google, Apple, Web3Onboard. */
readonly type?: string
/** Casdoor category, e.g. OAuth, Web3, SAML. */
/** IAM category, e.g. OAuth, Web3, SAML. */
readonly category?: string
readonly canSignIn?: boolean
readonly canSignUp?: boolean
+6 -6
View File
@@ -20,8 +20,8 @@ export function ForgotForm(props: ForgotFormProps) {
try {
const res = await client.forgot({
identifier,
clientId: client.tenant.clientId,
organization: client.tenant.orgId,
clientId: client.org.clientId,
organization: client.org.orgId,
})
if (!res.ok) setError(res.error ?? 'send failed')
else {
@@ -40,13 +40,13 @@ export function ForgotForm(props: ForgotFormProps) {
}
return (
<form onSubmit={onSubmit} className="hanzo-id-forgot-form" aria-busy={busy}>
<label>
<form onSubmit={onSubmit} className="hanzo-id-form" aria-busy={busy}>
<label className="hanzo-id-field">
<span>Email</span>
<input type="email" autoComplete="email" value={identifier} onChange={(e) => setIdentifier(e.target.value)} required />
<input className="hanzo-id-input" type="email" autoComplete="email" value={identifier} onChange={(e) => setIdentifier(e.target.value)} required />
</label>
{error ? <p role="alert" className="hanzo-id-error">{error}</p> : null}
<button type="submit" disabled={busy}>{busy ? 'Sending…' : 'Send reset link'}</button>
<button type="submit" className="hanzo-id-btn" disabled={busy}>{busy ? 'Sending…' : 'Send reset link'}</button>
</form>
)
}
+47 -21
View File
@@ -1,6 +1,7 @@
import { useState, type FormEvent } from 'react'
import type { AuthClient } from '../client'
import type { LoginResponse } from '../types'
import { PasswordField } from './PasswordField'
export interface LoginFormProps {
readonly client: AuthClient
@@ -33,17 +34,45 @@ export function LoginForm(props: LoginFormProps) {
setBusy(true)
setError(null)
try {
// Authenticate against the ORG OF THE APP being logged into, not the
// portal's own brand. When a downstream app initiates the login it passes
// its own `client_id` (props.clientIdOverride); that app may live in a
// different org than this brand portal — e.g. the admin-guard
// (client_id=hanzo-admin-guard) is in the `admin` org, so its operators
// must resolve to the admin/* identity (owner=admin), NOT this brand's
// hanzo/* row. get-app-login is the canonical clientId -> {application,
// organization} map; resolve through it and post BOTH so IAM scopes the
// credential check to the app's org.
//
// BOTH entry points resolve the same way — the downstream-app login
// (clientIdOverride) and the brand portal's own bare sign-in. They used to
// differ: the bare portal deliberately posted NO `organization` so IAM's
// cross-org fallback landed a colliding identity (z@hanzo.ai exists in both
// `admin` and `hanzo`) on admin/* and returned the full multi-org session.
//
// That is gone, on purpose, at the server. iam2 scopes every credential
// lookup to one org and treats the collision it relied on as a defect —
// "the F-2 bug where z@hanzo.ai collided across admin and hanzo" — because
// cross-org resolution coupled lockout counters across rows and gave a
// brute-force oracle on the superadmin. So it now REFUSES an org-less login
// with "organization, username and password are required". It answers HTTP
// **200**, which the form then renders as if the user's own password were
// wrong, and which every status-code monitor reads as green — the apex form
// was dead on hanzo.id, lux.id, iam.hanzo.ai and pars.id simultaneously.
//
// Posting the app's own org is the established answer (it is what the
// override path already does, and what reaches admin/* for admin-org apps).
// A global admin is no longer resolved by omission; they reach the admin
// identity by signing into an admin-org app, which is the explicit path.
const app = await client.getAppLogin(props.clientIdOverride ?? client.org.clientId)
const application = app?.application ?? client.org.appName
const organization = app?.organization ?? client.org.loginOrg
const res = await client.login({
identifier,
password,
clientId: props.clientIdOverride ?? client.tenant.clientId,
application: client.tenant.appName,
// Org-agnostic by default: `loginOrg` is unset, so no `organization` is
// posted and IAM resolves the user cross-org by credentials. A global
// admin (identity in the `admin` org) lands in the global multi-org
// session; a brand-only user lands in their own org. Pinning the brand
// org here would truncate a global admin to one org (the live bug).
organization: client.tenant.loginOrg,
clientId: props.clientIdOverride ?? client.org.clientId,
application,
organization,
redirectUri: props.redirectUri,
state: props.state,
codeChallenge: props.codeChallenge,
@@ -71,10 +100,11 @@ export function LoginForm(props: LoginFormProps) {
}
return (
<form onSubmit={onSubmit} className="hanzo-id-login-form" aria-busy={busy}>
<label>
<form onSubmit={onSubmit} className="hanzo-id-form" aria-busy={busy}>
<label className="hanzo-id-field">
<span>Email or username</span>
<input
className="hanzo-id-input"
type="text"
autoComplete="username"
value={identifier}
@@ -82,18 +112,14 @@ export function LoginForm(props: LoginFormProps) {
required
/>
</label>
<label>
<span>Password</span>
<input
type="password"
autoComplete="current-password"
value={password}
onChange={(e) => setPassword(e.target.value)}
required
/>
</label>
<PasswordField
label="Password"
value={password}
onChange={setPassword}
autoComplete="current-password"
/>
{error ? <p role="alert" className="hanzo-id-error">{error}</p> : null}
<button type="submit" disabled={busy}>{busy ? 'Signing in…' : 'Sign in'}</button>
<button type="submit" className="hanzo-id-btn" disabled={busy}>{busy ? 'Signing in…' : 'Sign in'}</button>
</form>
)
}
+132
View File
@@ -0,0 +1,132 @@
import { useEffect, useMemo, useState } from 'react'
import encodeQR from '@paulmillr/qr'
import type { AuthClient } from '../client'
import type { MfaIdentity, MfaSetup } from '../types'
import { OTPForm } from './OTPForm'
export interface MfaEnrollFormProps {
readonly client: AuthClient
/**
* Called once the user has verified a TOTP code AND the enrollment is
* persisted. The caller continues the session (onboarding or the OIDC
* code redirect).
*/
readonly onComplete: () => void
}
/**
* Forced TOTP enrollment, shown when IAM answers a login with `RequiredMfa`
* (org policy requires MFA and the user has none). There is intentionally NO
* skip / dismiss control — the only way past this screen is to enroll an
* authenticator. The QR is rendered locally from the `otpauth://` URI, so the
* TOTP secret never leaves the browser.
*
* Flow: `getAccount` (resolve identity from the session IAM set with
* `RequiredMfa`) → `mfaInitiate` (secret + QR) → user scans → `mfaVerify`
* (prove the code) → `mfaEnable` (persist) → `onComplete`.
*/
export function MfaEnrollForm({ client, onComplete }: MfaEnrollFormProps) {
const [identity, setIdentity] = useState<MfaIdentity | null>(null)
const [setup, setSetup] = useState<MfaSetup | null>(null)
const [fatal, setFatal] = useState<string | null>(null)
const [error, setError] = useState<string | null>(null)
const [busy, setBusy] = useState(false)
useEffect(() => {
let cancelled = false
async function begin() {
try {
const id = await client.getAccount()
if (!id) throw new Error('Your session could not be resolved. Please sign in again.')
const s = await client.mfaInitiate(id)
if (cancelled) return
setIdentity(id)
setSetup(s)
} catch (e) {
if (!cancelled) setFatal(e instanceof Error ? e.message : String(e))
}
}
void begin()
return () => {
cancelled = true
}
}, [client])
const qrSvg = useMemo(() => (setup ? encodeQR(setup.url, 'svg') : ''), [setup])
async function onCode(code: string) {
if (!identity || !setup || busy) return
setBusy(true)
setError(null)
try {
const verified = await client.mfaVerify({ owner: identity.owner, name: identity.name, secret: setup.secret, passcode: code })
if (!verified.ok) {
setError(verified.error ?? 'That code did not match. Try the current code from your app.')
return
}
const enabled = await client.mfaEnable({
owner: identity.owner,
name: identity.name,
secret: setup.secret,
recoveryCode: setup.recoveryCodes[0] ?? '',
})
if (!enabled.ok) {
setError(enabled.error ?? 'Could not enable two-factor authentication.')
return
}
onComplete()
} finally {
setBusy(false)
}
}
if (fatal) {
return (
<div className="hanzo-id-mfa-enroll">
<h2>Two-factor setup</h2>
<p role="alert" className="hanzo-id-error">{fatal}</p>
</div>
)
}
if (!setup) {
return (
<div className="hanzo-id-mfa-enroll">
<h2>Two-factor setup</h2>
<p className="lede">Preparing your authenticator</p>
</div>
)
}
const recoveryCode = setup.recoveryCodes[0]
return (
<div className="hanzo-id-mfa-enroll">
<h2>Set up two-factor authentication</h2>
<p className="lede">
Your organization requires two-factor authentication. Scan this QR code with an
authenticator app (Google Authenticator, 1Password, Authy), then enter the 6-digit code it
shows.
</p>
<div
className="hanzo-id-mfa-qr"
role="img"
aria-label="TOTP enrollment QR code"
// Local SVG from @paulmillr/qr — the otpauth secret never leaves the browser.
dangerouslySetInnerHTML={{ __html: qrSvg }}
/>
<details className="hanzo-id-mfa-manual">
<summary>Can't scan? Enter this key manually</summary>
<code className="hanzo-id-mfa-secret">{setup.secret}</code>
</details>
{error ? <p role="alert" className="hanzo-id-error">{error}</p> : null}
<OTPForm channel="totp" onSubmit={onCode} />
{recoveryCode ? (
<p className="hanzo-id-mfa-recovery">
Save this recovery code somewhere safe it lets you sign in if you lose your device:
<br />
<code>{recoveryCode}</code>
</p>
) : null}
</div>
)
}
+4 -3
View File
@@ -26,10 +26,11 @@ export function OTPForm(props: OTPFormProps) {
const label = channel === 'sms' ? 'SMS code' : channel === 'email' ? 'Email code' : 'Authenticator code'
return (
<form onSubmit={onSubmit} className="hanzo-id-otp-form" aria-busy={busy}>
<label>
<form onSubmit={onSubmit} className="hanzo-id-form" aria-busy={busy}>
<label className="hanzo-id-field">
<span>{label}</span>
<input
className="hanzo-id-input"
type="text"
inputMode="numeric"
pattern={`\\d{${length}}`}
@@ -41,7 +42,7 @@ export function OTPForm(props: OTPFormProps) {
/>
</label>
{channel === 'sms' ? <SmsConsentNotice /> : null}
<button type="submit" disabled={busy || code.length !== length}>{busy ? 'Verifying…' : 'Verify'}</button>
<button type="submit" className="hanzo-id-btn" disabled={busy || code.length !== length}>{busy ? 'Verifying…' : 'Verify'}</button>
</form>
)
}
+73
View File
@@ -0,0 +1,73 @@
import { useId, useState } from 'react'
import { EyeIcon, EyeOffIcon } from './icons'
/**
* The ONE password field. Every credential this portal collects is typed into
* this component, so "can I see what I typed" is answered in one place rather
* than per form.
*
* A masked field with no way to unmask is the single most common cause of a
* failed sign-in that looks like a wrong password, and it is worst exactly where
* typing is least reliable: a phone keyboard, one glyph at a time, with
* autocorrect and a shifted layout in the way. On signup it is worse still — the
* account is CREATED with whatever was actually typed, so a typo behind the dots
* locks the user out of an account they believe they just made.
*
* The reveal is the user's call, not ours. It defaults to masked, so nothing
* about the resting page changes; unmasking is a deliberate act with the eye
* open as its own state indicator.
*/
export interface PasswordFieldProps {
readonly label: string
readonly value: string
readonly onChange: (value: string) => void
/**
* `current-password` on sign-in, `new-password` on registration. It is what
* lets a password manager fill the right thing, and getting it wrong is how a
* manager offers to "update" a saved credential during a fresh signup.
*/
readonly autoComplete: 'current-password' | 'new-password'
readonly minLength?: number
}
export function PasswordField(props: PasswordFieldProps) {
const [shown, setShown] = useState(false)
const id = useId()
return (
<div className="hanzo-id-field">
{/* A <label> wrapping the control is the idiom everywhere else here, but a
<button> inside a <label> is activated twice — once as the button, once
as the label forwarding to its control — so the toggle would fight
itself. `htmlFor` keeps the same click-the-text-to-focus behaviour with
the button safely outside. */}
<label htmlFor={id}>{props.label}</label>
<div className="hanzo-id-reveal">
<input
id={id}
className="hanzo-id-input"
type={shown ? 'text' : 'password'}
autoComplete={props.autoComplete}
value={props.value}
onChange={(e) => props.onChange(e.target.value)}
minLength={props.minLength}
required
/>
<button
// NOT a submit. A bare <button> in a <form> defaults to type=submit,
// so without this, revealing the password would submit the form.
type="button"
className="hanzo-id-revealbtn"
// The control is a toggle, so it says what it DOES and reports its
// state separately; a label that flips between "Show"/"Hide" makes
// screen readers announce a change of control rather than of state.
aria-label="Show password"
aria-pressed={shown}
aria-controls={id}
onClick={() => setShown((s) => !s)}
>
{shown ? <EyeOffIcon width={20} height={20} /> : <EyeIcon width={20} height={20} />}
</button>
</div>
</div>
)
}
-80
View File
@@ -1,80 +0,0 @@
import type { AuthClient } from '../client'
import type { ProviderInfo } from '../types'
interface Meta {
readonly label: string
/** stable brand token used for the CSS class + data attribute (for icon styling) */
readonly brand: string
}
// Keyed by a normalized provider token (type or name, lowercased, alnum-only,
// "provider" prefix stripped). Falls back to a generic label for anything new.
const META: Record<string, Meta> = {
google: { label: 'Continue with Google', brand: 'google' },
github: { label: 'Continue with GitHub', brand: 'github' },
apple: { label: 'Continue with Apple', brand: 'apple' },
facebook: { label: 'Continue with Facebook', brand: 'facebook' },
web3: { label: 'Connect wallet', brand: 'web3' },
web3onboard: { label: 'Connect wallet', brand: 'web3' },
metamask: { label: 'Connect wallet', brand: 'web3' },
}
function metaFor(p: ProviderInfo): Meta {
const key = (p.type || p.name || '')
.toLowerCase()
.replace(/[^a-z0-9]/g, '')
.replace(/^provider/, '')
return META[key] ?? { label: `Continue with ${p.displayName || p.name}`, brand: 'generic' }
}
export interface ProviderButtonsProps {
readonly client: AuthClient
readonly providers: ProviderInfo[]
readonly mode: 'login' | 'signup'
readonly redirectUri?: string
readonly state?: string
readonly clientIdOverride?: string
}
/**
* Renders one button per social/wallet provider attached to the application.
* Each links to the IAM authorize endpoint with `provider=<name>`, which
* initiates that provider's OAuth and returns to `${publicOrigin}/callback`.
* The set is driven by the live app config (AuthClient.appLogin) — no
* hardcoded provider list — so enabling a provider in IAM surfaces it here.
*/
export function ProviderButtons(props: ProviderButtonsProps) {
const { client, providers, mode } = props
const usable = providers.filter((p) =>
mode === 'signup' ? p.canSignUp !== false : p.canSignIn !== false,
)
if (usable.length === 0) return null
const redirectUri = props.redirectUri ?? `${client.tenant.publicOrigin}/callback`
return (
<div className="hanzo-id-providers">
{usable.map((p) => {
const m = metaFor(p)
const href = client.authorize({
clientId: props.clientIdOverride ?? client.tenant.clientId,
redirectUri,
state: props.state ?? mode,
provider: p.name,
})
return (
<a
key={p.name}
className={`hanzo-id-provider-btn hanzo-id-provider-${m.brand}`}
href={href}
data-provider={m.brand}
>
{m.label}
</a>
)
})}
<div className="hanzo-id-or">
<span>or</span>
</div>
</div>
)
}
+58 -23
View File
@@ -1,10 +1,21 @@
import { useState, type FormEvent } from 'react'
import type { AuthClient } from '../client'
import { PasswordField } from './PasswordField'
export interface SignupFormProps {
readonly client: AuthClient
readonly inviteCode?: string
readonly onSuccess?: () => void
/**
* The downstream OIDC request the user arrived with, when an app sent them
* here to register. Forwarded to the sign-in that follows account creation so
* the flow ends where it started — back at the app, holding a code.
*/
readonly redirectUri?: string
readonly state?: string
readonly clientIdOverride?: string
readonly codeChallenge?: string
readonly codeChallengeMethod?: 'S256' | 'plain'
readonly nonce?: string
}
export function SignupForm(props: SignupFormProps) {
@@ -19,17 +30,45 @@ export function SignupForm(props: SignupFormProps) {
setBusy(true)
setError(null)
try {
const res = await client.signup({
// Register against the app the user CAME FROM, not this portal. IAM's
// signup resolves the application by clientId and then gates the org
// against that app's own org, so a downstream `client_id` must reach
// it or the account is created under the portal's app instead.
const clientId = props.clientIdOverride ?? client.org.clientId
const app = await client.getAppLogin(clientId, props.redirectUri)
const application = app?.application ?? client.org.appName
const organization = app?.organization ?? client.org.orgId
const session = await client.signup({
email,
password,
clientId: client.tenant.clientId,
application: client.tenant.appName,
organization: client.tenant.orgId,
clientId,
application,
organization,
inviteCode: props.inviteCode,
redirectUri: props.redirectUri,
state: props.state,
codeChallenge: props.codeChallenge,
codeChallengeMethod: props.codeChallengeMethod,
nonce: props.nonce,
})
if (res.error) setError(res.error)
else if (res.redirectUrl) window.location.href = res.redirectUrl
else props.onSuccess?.()
if (session.error) {
setError(session.error)
return
}
if (session.redirectUrl) {
window.location.href = session.redirectUrl
return
}
// The account exists but the session did not complete here — an org that
// forces MFA answers the login with an enrollment step. Hand the user to
// the sign-in page, carrying the same OIDC request, rather than leaving
// them on a form that has nothing left to do.
if (session.mfaRequired) {
window.location.href = `/login${window.location.search}`
return
}
setError('Your account was created, but sign-in did not complete. Please sign in.')
} catch (err) {
setError(String(err))
} finally {
@@ -38,24 +77,20 @@ export function SignupForm(props: SignupFormProps) {
}
return (
<form onSubmit={onSubmit} className="hanzo-id-signup-form" aria-busy={busy}>
<label>
<form onSubmit={onSubmit} className="hanzo-id-form" aria-busy={busy}>
<label className="hanzo-id-field">
<span>Email</span>
<input type="email" autoComplete="email" value={email} onChange={(e) => setEmail(e.target.value)} required />
</label>
<label>
<span>Password</span>
<input
type="password"
autoComplete="new-password"
value={password}
onChange={(e) => setPassword(e.target.value)}
minLength={12}
required
/>
<input className="hanzo-id-input" type="email" autoComplete="email" value={email} onChange={(e) => setEmail(e.target.value)} required />
</label>
<PasswordField
label="Password"
value={password}
onChange={setPassword}
autoComplete="new-password"
minLength={12}
/>
{error ? <p role="alert" className="hanzo-id-error">{error}</p> : null}
<button type="submit" disabled={busy}>{busy ? 'Creating account…' : 'Create account'}</button>
<button type="submit" className="hanzo-id-btn" disabled={busy}>{busy ? 'Creating account…' : 'Create account'}</button>
</form>
)
}
+173 -55
View File
@@ -1,11 +1,17 @@
import { useEffect, useState } from 'react'
import { Fragment, useEffect, useRef, useState } from 'react'
import type { ComponentType, SVGProps } from 'react'
import type { Chain } from '@hanzo/id-connect'
import type { AuthClient } from '../client'
import type { AppProvider } from '../types'
import { startProviderLogin, isHoppableProvider } from '../social'
import { loginWithWalletChain, ENABLED_WALLET_CHAINS, WALLET_CHAIN_LABELS } from '../web3'
import { GitHubIcon, GoogleIcon, WalletIcon } from './icons'
import { authorizeRequest, matchProviderHint, PROVIDER_ORDER } from '../social'
import { createIam } from '../iam'
import {
loginWithWalletChain,
detectWalletChains,
ENABLED_WALLET_CHAINS,
WALLET_CHAIN_LABELS,
} from '../web3'
import { GitHubIcon, GitLabIcon, GoogleIcon, WalletIcon } from './icons'
import { Divider } from './Divider'
/**
@@ -19,13 +25,17 @@ import { Divider } from './Divider'
* real creds land. When the config is unreadable we render none.
*
* Two sign-in shapes, decomplected:
* - OAuth (github/google) → the provider "hop" (`startProviderLogin`): redirect
* straight to the provider; the IAM backend `/callback` exchange completes it.
* - OAuth (github/google/gitlab) → FEDERATION: name the provider on IAM's own
* authorize endpoint (`?provider=provider-github`) and let IAM run the entire
* IdP leg server-side, where the client secret lives. See `social.ts`. This
* browser never builds an IdP URL and never sees a provider code.
* - Web3/wallet → native Sign-In-With-X (`loginWithWalletChain`): connect a
* wallet with `@hanzo/id-connect` (no WalletConnect, no projectId), sign the
* IAM-minted challenge, POST `/v1/iam/web3/verify`, then follow the SAME
* redirect the password flow returns. The wallet provider expands into one
* button per ENABLED chain.
* redirect the password flow returns. The wallet provider renders ONE
* chain-agnostic "Connect Wallet" button: it auto-detects the injected
* chain (`detectWalletChains`) and connects straight when exactly one is
* present, else reveals a chooser so either EVM or Solana stays reachable.
*/
export interface SocialButtonsProps {
readonly client: AuthClient
@@ -41,6 +51,22 @@ export interface SocialButtonsProps {
* a bare portal sign-in that lands on onboarding.
*/
readonly postLoginRedirect?: string
/**
* A `provider_hint` from the authorize query — the console passes
* `?provider_hint=provider-github` when a user clicks "Continue with GitHub"
* over there. When set, once the app config resolves this component launches
* the matching provider's hop straight away (the SAME hop the button runs) and
* renders NOTHING: it is headless, a pure side-effect, so the caller shows its
* own "signing you in" state. If the hint matches no configured provider,
* `onAutoStartResolved(false)` fires so the caller can fall back to the form.
*/
readonly autoStart?: string
/**
* Called once, in `autoStart` mode, after the app config resolves: `true` when
* the hinted provider launched, `false` when the hint matched nothing (so the
* caller can drop to the interactive form instead of a blank redirect state).
*/
readonly onAutoStartResolved?: (started: boolean) => void
}
interface ProviderMeta {
@@ -52,16 +78,12 @@ interface ProviderMeta {
/** Display metadata for the providers the portal knows how to render. */
const PROVIDER_META: Record<string, ProviderMeta> = {
github: { key: 'github', label: 'GitHub', Icon: GitHubIcon },
gitlab: { key: 'gitlab', label: 'GitLab', Icon: GitLabIcon },
google: { key: 'google', label: 'Google', Icon: GoogleIcon },
web3: { key: 'web3', label: 'Wallet', Icon: WalletIcon },
}
/** Canonical render order. */
const ORDER = ['github', 'google', 'web3']
interface Resolved {
/** IAM application name (for the provider-hop state). */
readonly application: string
/** Configured + renderable providers, keyed by their normalized key. */
readonly providers: Record<string, AppProvider>
}
@@ -71,21 +93,76 @@ export function SocialButtons({
clientIdOverride,
intent = 'signin',
postLoginRedirect,
autoStart,
onAutoStartResolved,
}: SocialButtonsProps) {
const [resolved, setResolved] = useState<Resolved | null>(null)
const [error, setError] = useState<string | null>(null)
const [busyChain, setBusyChain] = useState<Chain | null>(null)
// The chain-agnostic wallet entry reveals a chooser only when it can't decide
// for the user (zero or multiple injected wallets); a single injected wallet
// connects straight without ever showing it.
const [walletMenu, setWalletMenu] = useState(false)
const autoStarted = useRef(false)
// Start federation: hand the provider's NAME to IAM's authorize endpoint and
// let IAM run the whole IdP leg. Shared by the button click and the `autoStart`
// auto-launch so both take the identical path.
//
// Two arms, and they are the same two the password path already branches on
// (`Login.completeAfterAuth`) — the question is only who owns the PKCE verifier:
//
// an app sent the user here → re-enter authorize with THAT app's request, so
// IAM mints the code against its client_id, redirect_uri and challenge and
// returns the browser straight to it. The app holds the verifier; this portal
// is never in the return path and never touches a token.
//
// a bare portal sign-in → the portal is its own client, so the IAM SDK mints
// and stores the verifier that `Callback` reads back. `post_login_redirect`
// carries a non-OIDC "come back here" target (device approval), which is why
// it belongs to this arm alone: it is only ever read by the portal's own
// callback, and only this arm runs it.
function hop(provider: AppProvider) {
const app = authorizeRequest(window.location.search, clientIdOverride ?? client.org.clientId)
if (app) {
sessionStorage.removeItem('post_login_redirect')
window.location.assign(client.authorize({ ...app, provider: provider.name }))
return
}
if (postLoginRedirect) sessionStorage.setItem('post_login_redirect', postLoginRedirect)
else sessionStorage.removeItem('post_login_redirect')
// Bare arm = the portal signs in as ITSELF, always. Threading a downstream
// app's client_id in here paired it with the portal's own `/callback` — a
// hybrid no app registers (that pairing was the older model), so IAM
// answered "invalid redirect_uri", and `Callback` (portal client) could
// never have redeemed the code anyway. An app that wants a code arrives as
// a full authorize request and takes the arm above.
createIam(client.org)
.signinRedirect({ additionalParams: { provider: provider.name } })
.catch((e) => setError(String(e)))
}
useEffect(() => {
let cancelled = false
// Read the app config against the DOWNSTREAM app's own redirect_uri (carried
// on the authorize query in the SSO flow), not the portal's /callback — IAM
// validates it against the app's registered list, and a cross-app clientId
// (e.g. console's `hanzo-cloud` viewed from hanzo.id) does NOT register the
// portal callback, so hardcoding it drops the whole response and no social
// resolves. Absent (bare portal / device flow) → getAppLogin defaults it.
const oidcRedirectUri =
typeof window !== 'undefined'
? new URLSearchParams(window.location.search).get('redirect_uri') ?? undefined
: undefined
client
.getAppLogin(clientIdOverride)
.getAppLogin(clientIdOverride, oidcRedirectUri)
.then((app) => {
if (cancelled) return
if (!app) {
// Can't read the app config → render no social rather than risk a
// dead-end button. Password / email-code still render.
setResolved({ application: '', providers: {} })
setResolved({ providers: {} })
onAutoStartResolved?.(false)
return
}
const want = intent === 'signup' ? (p: AppProvider) => p.canSignUp : (p: AppProvider) => p.canSignIn
@@ -99,43 +176,49 @@ export function SocialButtons({
const enabled = p.key === 'web3' ? want(p) : want(p) && p.configured
if (enabled && p.key in PROVIDER_META) providers[p.key] = p
}
setResolved({ application: app.application, providers })
setResolved({ providers })
// A client that already knows the provider (console `?provider_hint=…`)
// launches it straight away — the SAME hop the button runs, so a click
// over there lands directly in the provider flow, no second press and no
// bounce through this login page.
if (autoStart && !autoStarted.current) {
autoStarted.current = true
const target = matchProviderHint(Object.values(providers), autoStart)
if (target) {
onAutoStartResolved?.(true)
hop(target)
} else {
// Hint names a provider this app doesn't offer → let the caller show
// the form rather than dead-end on a blank "signing you in".
onAutoStartResolved?.(false)
}
}
})
.catch(() => {
if (!cancelled) setResolved({ application: '', providers: {} })
if (cancelled) return
setResolved({ providers: {} })
onAutoStartResolved?.(false)
})
return () => {
cancelled = true
}
// Run once on mount: getAppLogin is a one-shot and autoStart is fixed for
// the life of the page; the ref guards the hop against a double-fire.
// eslint-disable-next-line react-hooks/exhaustive-deps
}, [client, clientIdOverride, intent])
// In autoStart mode the component is headless — it exists only to run the hop
// above; the caller renders its own "signing you in" state. Render nothing.
if (autoStart) return null
if (resolved === null) return null // resolving — render nothing rather than flicker
const ordered = ORDER.filter((k) => k in resolved.providers)
const ordered = PROVIDER_ORDER.filter((k) => k in resolved.providers)
if (ordered.length === 0) return null
const verb = intent === 'signup' ? 'Sign up' : 'Continue'
function startOAuth(provider: AppProvider) {
setError(null)
// Persist the downstream target across the IAM round-trip; `Callback`
// reads it back and forwards tokens there (else lands on onboarding).
if (postLoginRedirect) sessionStorage.setItem('post_login_redirect', postLoginRedirect)
else sessionStorage.removeItem('post_login_redirect')
const method = intent === 'signup' ? 'signup' : 'signin'
startProviderLogin(
{
application: resolved!.application,
providerName: provider.name,
type: provider.type,
clientId: provider.clientId,
scopes: provider.scopes,
method,
},
// The shared OAuth client is registered against the IAM backend's
// /callback (not this brand host), so the hop must return there or the
// provider rejects the redirect_uri. Catalog-driven; defaults to host.
client.tenant.oauthCallbackOrigin,
)
hop(provider)
}
async function startWallet(chain: Chain) {
@@ -164,6 +247,16 @@ export function SocialButtons({
}
}
// The chain-agnostic entry: auto-detect the injected wallet and connect
// straight when exactly one chain is available; otherwise reveal the chooser
// so the user picks EVM or Solana. Both underlying flows stay reachable.
function onConnectWallet() {
setError(null)
const detected = detectWalletChains()
if (detected.length === 1) startWallet(detected[0]!)
else setWalletMenu(true)
}
return (
<>
<div className="hanzo-id-social">
@@ -172,31 +265,56 @@ export function SocialButtons({
// Web3 expands into one connect button per ENABLED chain; OAuth
// providers render a single hop button.
if (k === 'web3') {
return ENABLED_WALLET_CHAINS.map((chain) => (
<button
key={`web3-${chain}`}
type="button"
className="hanzo-id-social-btn"
data-provider="web3"
data-chain={chain}
disabled={busyChain !== null}
onClick={() => startWallet(chain)}
>
<WalletIcon />
<span>
{busyChain === chain ? 'Connecting…' : `${verb} with ${WALLET_CHAIN_LABELS[chain]}`}
</span>
</button>
))
// ONE chain-agnostic entry. It connects straight when a single
// wallet is detected, else expands into the chooser below — so the
// page always shows exactly one "Connect Wallet" button, with both
// EVM and Solana reachable from it.
return (
<Fragment key="web3">
<button
type="button"
className="hanzo-id-btn ghost"
data-provider="web3"
data-wallet-connect="true"
aria-expanded={walletMenu}
disabled={busyChain !== null}
onClick={onConnectWallet}
>
<WalletIcon />
<span>{busyChain !== null && !walletMenu ? 'Connecting…' : 'Connect Wallet'}</span>
</button>
{walletMenu ? (
<div
className="hanzo-id-wallet-chains"
role="group"
aria-label="Choose a wallet network"
>
{ENABLED_WALLET_CHAINS.map((chain) => (
<button
key={`web3-${chain}`}
type="button"
className="hanzo-id-btn ghost"
data-provider="web3"
data-chain={chain}
disabled={busyChain !== null}
onClick={() => startWallet(chain)}
>
<WalletIcon />
<span>{busyChain === chain ? 'Connecting…' : WALLET_CHAIN_LABELS[chain]}</span>
</button>
))}
</div>
) : null}
</Fragment>
)
}
if (!isHoppableProvider(provider.type)) return null
const meta = PROVIDER_META[k]!
const { Icon } = meta
return (
<button
key={k}
type="button"
className="hanzo-id-social-btn"
className="hanzo-id-btn ghost"
data-provider={k}
onClick={() => startOAuth(provider)}
>
+34
View File
@@ -32,6 +32,14 @@ export function GoogleIcon(props: SVGProps<SVGSVGElement>) {
)
}
export function GitLabIcon(props: SVGProps<SVGSVGElement>) {
return (
<svg {...base(props)} fill="currentColor">
<path d="M23.955 13.587l-1.342-4.135-2.664-8.189a.455.455 0 0 0-.867 0L16.418 9.45H7.582L4.919 1.263a.455.455 0 0 0-.867 0L1.388 9.452-.001 13.587a.924.924 0 0 0 .331 1.023L12 23.054l11.625-8.443a.92.92 0 0 0 .33-1.024" />
</svg>
)
}
export function WalletIcon(props: SVGProps<SVGSVGElement>) {
return (
<svg {...base(props)} fill="none" stroke="currentColor" strokeWidth={2} strokeLinecap="round" strokeLinejoin="round">
@@ -41,3 +49,29 @@ export function WalletIcon(props: SVGProps<SVGSVGElement>) {
</svg>
)
}
/**
* Password reveal, the only pair here that is a STATE rather than a brand: eye
* = currently masked (tap to reveal), struck-through eye = currently visible
* (tap to hide). Stroked rather than filled, because at 20px a filled eye reads
* as a blob; `base` supplies the 18px default, and PasswordField overrides to 20
* so the glyph holds its own inside a 44px target.
*/
export function EyeIcon(props: SVGProps<SVGSVGElement>) {
return (
<svg {...base(props)} fill="none" stroke="currentColor" strokeWidth="1.75" strokeLinecap="round" strokeLinejoin="round">
<path d="M2 12s3.5-6.5 10-6.5S22 12 22 12s-3.5 6.5-10 6.5S2 12 2 12Z" />
<circle cx="12" cy="12" r="2.75" />
</svg>
)
}
export function EyeOffIcon(props: SVGProps<SVGSVGElement>) {
return (
<svg {...base(props)} fill="none" stroke="currentColor" strokeWidth="1.75" strokeLinecap="round" strokeLinejoin="round">
<path d="M10.6 6.1A9.9 9.9 0 0 1 12 5.5c6.5 0 10 6.5 10 6.5a17 17 0 0 1-3.1 3.9M6.4 7.9A17 17 0 0 0 2 12s3.5 6.5 10 6.5a9.9 9.9 0 0 0 3.6-.65" />
<path d="M10.1 10.1a2.75 2.75 0 0 0 3.8 3.8" />
<path d="m3 3 18 18" />
</svg>
)
}
+2
View File
@@ -2,6 +2,8 @@ export { LoginForm } from './LoginForm'
export { SignupForm } from './SignupForm'
export { ForgotForm } from './ForgotForm'
export { OTPForm } from './OTPForm'
export { MfaEnrollForm, type MfaEnrollFormProps } from './MfaEnrollForm'
export { SmsConsentNotice, SMS_CONSENT_TEXT } from './SmsConsent'
export { SocialButtons, type SocialButtonsProps } from './SocialButtons'
export { Divider } from './Divider'
export { PasswordField, type PasswordFieldProps } from './PasswordField'
+34 -8
View File
@@ -7,19 +7,20 @@
* fake signer (the injectable `WalletSigner` seam — the real one lazy-loads the
* wallet libs, which this test never touches).
*/
import { test } from 'node:test'
import { test } from 'vitest'
import assert from 'node:assert/strict'
import { createAuthClient } from './client.ts'
import {
loginWithWalletChain,
detectWalletChains,
ENABLED_WALLET_CHAINS,
WALLET_CHAIN_LABELS,
type WalletSigner,
} from './web3.ts'
import type { Chain, LoginChallenge, SignedProof } from '@hanzo/id-connect'
import type { TenantConfig } from '@hanzo/id-shared'
import type { OrgConfig } from '@hanzo/id-shared'
function tenant(overrides: Partial<TenantConfig> = {}): TenantConfig {
function org(overrides: Partial<OrgConfig> = {}): OrgConfig {
return {
orgId: 'hanzo',
iamUrl: 'https://hanzo.id',
@@ -90,7 +91,7 @@ function fakeSigner() {
test('fetches the nonce for the chosen chain, signs the returned challenge, POSTs the proof', async () => {
const { calls, fetchImpl } = capturingFetch()
const { seen, sign } = fakeSigner()
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
const res = await loginWithWalletChain(client, 'evm', {}, fetchImpl, sign)
@@ -132,7 +133,7 @@ test('fetches the nonce for the chosen chain, signs the returned challenge, POST
test('SSO flow (downstream redirectUri) sends type=code and returns the app redirect with the minted code', async () => {
const { calls, fetchImpl } = capturingFetch('CODE_XYZ')
const { sign } = fakeSigner()
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
const res = await loginWithWalletChain(
client,
@@ -170,7 +171,7 @@ test('disabled chains are not offered and fail closed without any network or sig
signed = true
throw new Error('signer must not run for a disabled chain')
}
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
const res = await loginWithWalletChain(client, 'ton', {}, fetchImpl, sign)
assert.match(res.error ?? '', /not enabled/)
assert.equal(calls.length, 0)
@@ -182,7 +183,7 @@ test('a wallet rejection surfaces as { error } (not a throw), and verify is neve
const sign: WalletSigner = async () => {
throw new Error('User rejected the request')
}
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
const res = await loginWithWalletChain(client, 'evm', {}, fetchImpl, sign)
assert.equal(res.error, 'User rejected the request')
// nonce was fetched (1 call) but verify was NOT (no 2nd call).
@@ -190,6 +191,31 @@ test('a wallet rejection surfaces as { error } (not a throw), and verify is neve
assert.match(calls[0]!.url, /\/v1\/iam\/web3\/nonce/)
})
test('detectWalletChains: a single injected wallet resolves to exactly its chain', () => {
// EVM only → [evm]; the UI connects straight, no chooser.
assert.deepEqual(detectWalletChains({ ethereum: {} }), ['evm'])
// Solana only, via any of the recognized injected providers → [solana].
assert.deepEqual(detectWalletChains({ solana: {} }), ['solana'])
assert.deepEqual(detectWalletChains({ solflare: {} }), ['solana'])
assert.deepEqual(detectWalletChains({ backpack: {} }), ['solana'])
})
test('detectWalletChains: both injected → both, in enabled order (chooser)', () => {
assert.deepEqual(detectWalletChains({ ethereum: {}, solana: {} }), ['evm', 'solana'])
})
test('detectWalletChains: nothing injected → [] (chooser, both still reachable)', () => {
// No window (server / node) and an empty window both resolve to none — the UI
// then reveals the chooser so EVM and Solana stay selectable regardless.
assert.deepEqual(detectWalletChains({}), [])
assert.deepEqual(detectWalletChains(undefined), [])
assert.deepEqual(detectWalletChains(), []) // node has no global window
// Every detectable chain is one the wallet flow actually enables.
for (const c of detectWalletChains({ ethereum: {}, solana: {} })) {
assert.ok(ENABLED_WALLET_CHAINS.includes(c))
}
})
test('an IAM verify error is returned as { error }', async () => {
const calls: string[] = []
const fetchImpl: typeof fetch = async (input) => {
@@ -201,7 +227,7 @@ test('an IAM verify error is returned as { error }', async () => {
return new Response(JSON.stringify({ status: 'error', msg: 'web3: bad signature' }), { status: 200 })
}
const { sign } = fakeSigner()
const client = createAuthClient({ tenant: tenant(), fetchImpl })
const client = createAuthClient({ org: org(), fetchImpl })
const res = await loginWithWalletChain(client, 'evm', {}, fetchImpl, sign)
assert.equal(res.error, 'web3: bad signature')
assert.equal(res.redirectUrl, undefined)
+50 -11
View File
@@ -17,7 +17,7 @@
* POST {iamUrl}/v1/iam/web3/verify body = SignedProof + routing fields
* → same success shape as /v1/iam/login (auth code | session cookie).
*/
import type { TenantConfig } from '@hanzo/id-shared'
import type { OrgConfig } from '@hanzo/id-shared'
import type { Chain, LoginChallenge, SignedProof } from '@hanzo/id-connect'
import type { AuthClient } from './client'
import type { LoginResponse } from './types'
@@ -55,9 +55,48 @@ export const WALLET_CHAIN_LABELS: Record<Chain, string> = {
xrp: 'XRP',
}
/** The `window` fields the injected-wallet sniff reads — kept local so the
* wallet libs stay out of this module (detection is a pure property read). */
export interface WalletWindow {
readonly ethereum?: unknown
readonly solana?: unknown
readonly solflare?: unknown
readonly backpack?: unknown
}
/** Is an injected wallet for `chain` present on `w`? Mirrors the connectors'
* own discovery: EVM = `window.ethereum` (EIP-1193 / EIP-6963 legacy handle),
* Solana = Phantom/Solflare/Backpack injected providers. */
function chainInjected(chain: Chain, w: WalletWindow): boolean {
switch (chain) {
case 'evm':
return Boolean(w.ethereum)
case 'solana':
return Boolean(w.solana || w.solflare || w.backpack)
default:
// A chain with no sniff is never auto-detected; the chooser still offers
// it. Only the ENABLED set is ever consulted, so this stays unreachable.
return false
}
}
/**
* The ENABLED wallet chains that currently have an injected provider. A pure
* `window` sniff — no connect, no I/O — that powers the chain-agnostic "Connect
* Wallet" entry: exactly one match → connect straight; zero or many → let the
* user pick. Derived from {@link ENABLED_WALLET_CHAINS} so there is ONE source
* of truth for what wallet login offers.
*/
export function detectWalletChains(
w: WalletWindow | undefined = typeof window === 'undefined' ? undefined : (window as WalletWindow),
): Chain[] {
if (!w) return []
return ENABLED_WALLET_CHAINS.filter((c) => chainInjected(c, w))
}
/** Routing context for the verify POST — exactly what the password flow carries. */
export interface WalletLoginContext {
/** Override the OAuth client_id (downstream app); defaults to tenant.clientId. */
/** Override the OAuth client_id (downstream app); defaults to org.clientId. */
readonly clientId?: string
/** Downstream app `redirect_uri`; presence flips the flow to the auth-code (SSO) path. */
readonly redirectUri?: string
@@ -75,7 +114,7 @@ export interface WalletLoginContext {
* the UI can't act on; expected failures (user rejects, bad signature) come back
* as `{ error }`.
*
* `client.tenant.iamUrl` is the fetch base (HIP-0111 host-relative — the brand's
* `client.org.iamUrl` is the fetch base (HIP-0111 host-relative — the brand's
* own `*.id` host), matching every other AuthClient call.
*/
export async function loginWithWalletChain(
@@ -85,7 +124,7 @@ export async function loginWithWalletChain(
fetchImpl: typeof fetch = fetch,
sign: WalletSigner = defaultSigner,
): Promise<LoginResponse> {
const tenant = client.tenant
const org = client.org
if (!ENABLED_WALLET_CHAINS.includes(chain)) {
return { error: `wallet login not enabled for ${chain}` }
}
@@ -96,7 +135,7 @@ export async function loginWithWalletChain(
// from the SIGNED message, so there is no second round-trip to scope it.
let proof: SignedProof
try {
const challenge = await fetchNonce(tenant, chain, fetchImpl)
const challenge = await fetchNonce(org, chain, fetchImpl)
proof = await sign(chain, challenge)
} catch (err) {
return { error: errMessage(err) }
@@ -104,17 +143,17 @@ export async function loginWithWalletChain(
// 2. Verify the proof + routing at IAM. Type defaults to "login" (session
// cookie) server-side; a downstream redirectUri makes it the code flow.
const url = new URL('/v1/iam/web3/verify', tenant.iamUrl)
const url = new URL('/v1/iam/web3/verify', org.iamUrl)
const res = await fetchImpl(url.toString(), {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
credentials: 'include',
body: JSON.stringify({
// routing
organization: tenant.loginOrg ?? '',
application: tenant.appName,
organization: org.loginOrg ?? '',
application: org.appName,
method: 'login',
clientId: ctx.clientId ?? tenant.clientId,
clientId: ctx.clientId ?? org.clientId,
redirectUri: ctx.redirectUri ?? '',
state: ctx.state ?? '',
scope: 'openid profile email',
@@ -138,11 +177,11 @@ export async function loginWithWalletChain(
/** GET the CAIP-122 challenge for (chain) from IAM; throws on a non-ok payload. */
async function fetchNonce(
tenant: TenantConfig,
org: OrgConfig,
chain: Chain,
fetchImpl: typeof fetch,
): Promise<LoginChallenge> {
const url = new URL('/v1/iam/web3/nonce', tenant.iamUrl)
const url = new URL('/v1/iam/web3/nonce', org.iamUrl)
url.searchParams.set('chain', chain)
const res = await fetchImpl(url.toString(), { headers: { Accept: 'application/json' } })
let body: Record<string, unknown> = {}
+9
View File
@@ -0,0 +1,9 @@
{
"extends": "../../tsconfig.base.json",
"compilerOptions": {
"outDir": "dist",
"noEmit": true
},
"include": ["src"],
"exclude": ["src/**/*.test.ts"]
}
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@hanzo/id-idv",
"version": "0.1.0",
"version": "0.1.1",
"description": "Pluggable identity verification (KYC / KYB / liveness). Provider-agnostic — wires Persona, Onfido, Veriff, Sumsub, or any custom backend behind a single React surface.",
"license": "BSD-3-Clause",
"type": "module",
+1 -1
View File
@@ -30,7 +30,7 @@ export type IDVStatus =
export interface IDVSubject {
/** Stable subject identifier (typically the IAM user id). */
readonly subjectId: string
/** Tenant org slug (for multi-tenant providers). */
/** Org org slug (for multi-org providers). */
readonly orgId: string
/** Email + display name carried through for vendor pre-fill. */
readonly email?: string
+3 -4
View File
@@ -1,6 +1,6 @@
{
"name": "@hanzo/id-onboarding",
"version": "0.1.0",
"version": "0.1.5",
"description": "Post-login onboarding for the Hanzo ID portal: choose/create org → optional project → optional wallet link. White-labeled by host. Domain / service / UI split.",
"license": "BSD-3-Clause",
"type": "module",
@@ -15,12 +15,11 @@
"files": ["src", "!src/**/*.test.ts"],
"scripts": {
"build": "tsc --noEmit",
"tc": "tsc --noEmit",
"test": "node --test --experimental-strip-types 'src/**/*.test.ts'"
"tc": "tsc --noEmit"
},
"dependencies": {
"@hanzo/id-shared": "workspace:*",
"@hanzo/iam": "^0.13.1"
"@hanzo/iam": "^0.21.1"
},
"peerDependencies": {
"react": ">=19",
+55 -2
View File
@@ -16,7 +16,7 @@
*/
/** Identifier for each step in the onboarding flow. */
export type StepId = 'org' | 'project' | 'wallet' | 'done'
export type StepId = 'org' | 'project' | 'wallet' | 'consent' | 'plan' | 'done'
/** A step's place in the linear flow. */
export interface StepDesc {
@@ -53,11 +53,29 @@ export const STEPS: readonly StepDesc[] = [
byline: 'Connect a Web3 wallet to sign and pay onchain. Optional.',
skippable: true,
},
{
id: 'consent',
title: 'Data sharing',
byline: 'Choose whether to share usage data to improve the products.',
// Not skippable: the agreement needs an explicit ANSWER (yes or no, both
// valid), recorded once on the user so it is never re-asked. Skipping is
// how this page kept going missing.
skippable: false,
},
{
id: 'plan',
title: 'Choose how you pay',
byline: 'Pick a plan, or pay as you go with a prepaid balance.',
// The LAST page, and a required choice: the platform is prepay-only, so an
// account is not usable until a plan or a balance exists. "Pay as you go"
// IS a choice — there is nothing to skip to.
skippable: false,
},
] as const
/** A minimal org reference the UI lists in the "choose org" step. */
export interface OrgRef {
/** Casdoor org slug (the `<org>` in `<org>-<app>`). */
/** IAM org slug (the `<org>` in `<org>-<app>`). */
readonly name: string
/** Human-facing name; falls back to `name` when unset. */
readonly displayName: string
@@ -85,8 +103,43 @@ export interface OnboardingState {
readonly projectName?: string
/** Wallet address linked in step 3, if any. */
readonly walletAddress?: string
/** The data-sharing answer given in step 4 (true = opted in). */
readonly dataSharingConsent?: boolean
/**
* The payment choice made on the final step: a plan slug from the billing
* catalog, or the literal 'payg' for a prepaid pay-as-you-go balance.
*/
readonly planChoice?: string
}
/**
* One purchasable plan as the billing catalog serves it (GET /v1/billing/plans
* on the pay origin). Prices are the CATALOG's — this pkg never states one.
*/
export interface PlanInfo {
readonly slug: string
readonly name: string
readonly description?: string
/** Monthly price in CENTS (the catalog's `price` — 900 = $9/mo). */
readonly priceCents: number
/**
* Monthly-equivalent price in CENTS when billed annually (the catalog's
* `priceAnnual` — 825 = $8.25/mo ≈ $99/yr). Absent when the plan has no
* annual rate.
*/
readonly priceAnnualCents?: number
readonly popular?: boolean
}
/**
* User-record keys the onboarding persists under `Properties`. ONE writer
* (saveOnboarding) and one reader (readOnboarding); the names are part of the
* user record's public shape, so change them never.
*/
export const PROP_COMPLETED = 'onboarding.completedAt'
export const PROP_CONSENT = 'onboarding.dataSharingConsent'
export const PROP_PLAN = 'onboarding.plan'
/** Resolve a step descriptor by id. */
export function stepById(id: StepId): StepDesc | undefined {
return STEPS.find((s) => s.id === id)
+6 -4
View File
@@ -1,9 +1,10 @@
// @hanzo/id-onboarding — post-login onboarding for the Hanzo ID portal.
//
// Three-step flow: choose/create org → optional project → optional wallet
// link. White-labeled by the host's brand name. Domain (serializable types +
// step machine) / service (IAM-backed writes) / UI (self-contained flow)
// split. Auth lives in @hanzo/id-auth — import login/signup from there.
// Five-step flow: choose/create org → optional project → optional wallet
// link → data-sharing consent → plan or pay-as-you-go. White-labeled by the
// host's brand name. Domain (serializable types + step machine) / service
// (IAM-backed writes) / UI (self-contained flow) split. Auth lives in
// @hanzo/id-auth — import login/signup from there.
// ── Domain ──────────────────────────────────────────────────────
export {
@@ -16,6 +17,7 @@ export {
type OrgRef,
type ProjectRef,
type OnboardingState,
type PlanInfo,
} from './domain/types'
// ── Service ─────────────────────────────────────────────────────
+126 -17
View File
@@ -7,18 +7,20 @@
* Covers the React-free surface: the domain step machine and the service's
* request shaping + IAM response translation (with an injected fake fetch).
*/
import { test } from 'node:test'
import { test } from 'vitest'
import assert from 'node:assert/strict'
import { STEPS, stepById, nextStep, prevStep } from './domain/types.ts'
import { createOnboardingService } from './service/onboarding.ts'
// ── Domain: step machine ────────────────────────────────────────────
test('step machine walks org → project → wallet → done', () => {
test('step machine walks org → project → wallet → consent → plan → done', () => {
assert.equal(STEPS[0]!.id, 'org')
assert.equal(nextStep('org'), 'project')
assert.equal(nextStep('project'), 'wallet')
assert.equal(nextStep('wallet'), 'done')
assert.equal(nextStep('wallet'), 'consent')
assert.equal(nextStep('consent'), 'plan')
assert.equal(nextStep('plan'), 'done')
assert.equal(nextStep('done'), 'done') // terminal is a fixpoint
})
@@ -26,12 +28,20 @@ test('prevStep is the inverse within the flow, undefined at the head', () => {
assert.equal(prevStep('org'), undefined)
assert.equal(prevStep('project'), 'org')
assert.equal(prevStep('wallet'), 'project')
assert.equal(prevStep('consent'), 'wallet')
assert.equal(prevStep('plan'), 'consent')
})
test('only org is required; project and wallet are skippable', () => {
test('project and wallet are skippable; org, consent and plan are not', () => {
assert.equal(stepById('org')!.skippable, false)
assert.equal(stepById('project')!.skippable, true)
assert.equal(stepById('wallet')!.skippable, true)
// Consent needs an ANSWER (either answer) and plan is the prepay gate —
// neither may be walked past. plan is LAST so the choice hands straight
// off to the pay surface.
assert.equal(stepById('consent')!.skippable, false)
assert.equal(stepById('plan')!.skippable, false)
assert.equal(STEPS[STEPS.length - 1]!.id, 'plan')
})
// ── Service: fake-fetch harness ─────────────────────────────────────
@@ -86,24 +96,51 @@ test('listOrgs hits get-organizations with the bearer token and maps rows', asyn
])
})
test('listOrgs decodes rows from the legacy data2 slot until IAM stops emitting it', async () => {
const { service } = harness(() => ({ json: { status: 'ok', data2: [{ name: 'acme' }] } }))
assert.deepEqual(await service.listOrgs(), [{ name: 'acme', displayName: 'acme' }])
})
test('listOrgs returns [] (not throw) on a server error', async () => {
const { service } = harness(() => ({ status: 500, json: { status: 'error', msg: 'boom' } }))
assert.deepEqual(await service.listOrgs(), [])
})
test('createOrg posts the org and reports IAM error messages', async () => {
const ok = harness(() => ({ json: { status: 'ok' } }))
// Founding an org goes through the SELF-SERVICE front door, never the
// add-organization admin verb — that one is bearer-only entity CRUD filed under
// owner "admin", so a person founding their first org gets 401/403 there. This is
// the regression guard for the hanzo.id/onboarding "HTTP 401".
test('createOrg founds the org through /v1/iam/onboard, never the admin verb', async () => {
const ok = harness(() => ({ json: { org: 'acme', accessKey: 'pk-live-x' } }))
const res = await ok.service.createOrg({ name: 'acme', displayName: 'Acme Inc' })
assert.equal(ok.calls[0]!.url, 'https://hanzo.id/v1/iam/add-organization')
assert.equal(ok.calls[0]!.url, 'https://hanzo.id/v1/iam/onboard')
assert.equal(ok.calls[0]!.method, 'POST')
const sent = JSON.parse(ok.calls[0]!.body!)
assert.equal(sent.name, 'acme')
assert.equal(sent.displayName, 'Acme Inc')
assert.ok(!ok.calls.some((c) => c.url.includes('add-organization')))
// The DISPLAY name is what travels: the server owns the slug policy.
assert.deepEqual(JSON.parse(ok.calls[0]!.body!), { name: 'Acme Inc' })
// …and the slug it answers with is authoritative, not the client's guess.
assert.deepEqual(res, { ok: true, value: { name: 'acme', displayName: 'Acme Inc' } })
})
const denied = harness(() => ({ status: 403, json: { status: 'error', msg: 'permission denied' } }))
const fail = await denied.service.createOrg({ name: 'x', displayName: 'X' })
assert.deepEqual(fail, { ok: false, error: 'HTTP 403' })
test('createOrg carries BOTH credentials — the portal session mints no bearer', async () => {
const { service, calls } = harness(() => ({ json: { org: 'acme' } }))
await service.createOrg({ name: 'acme', displayName: 'Acme Inc' })
assert.equal(calls[0]!.headers.Authorization, 'Bearer tok-123')
assert.equal(calls[0]!.headers['Content-Type'], 'application/json')
})
test('createOrg surfaces the front doors own error text, not a bare HTTP code', async () => {
const taken = harness(() => ({ status: 409, json: { error: 'the organization "acme" already exists' } }))
assert.deepEqual(await taken.service.createOrg({ name: 'acme', displayName: 'Acme' }), {
ok: false,
error: 'the organization "acme" already exists',
})
const anon = harness(() => ({ status: 401, json: { error: 'please sign in first' } }))
assert.deepEqual(await anon.service.createOrg({ name: 'x', displayName: 'X' }), {
ok: false,
error: 'please sign in first',
})
})
test('linkWallet rejects a malformed address before any network call', async () => {
@@ -113,24 +150,96 @@ test('linkWallet rejects a malformed address before any network call', async ()
assert.equal(calls.length, 0)
})
test('linkWallet resolves the user via get-account then writes web3onboard', async () => {
// IAM's update-user is a FULL-ROW write that ignores `columns=` — a minimal
// body silently blanks every field it omits. So the contract under test is
// read-merge-write: the row that comes back from get-account goes back OUT
// with only the mutation applied. This is the regression guard for the wallet
// step wiping displayName/email on every link.
test('linkWallet reads the full row and writes it back with web3onboard merged in', async () => {
const addr = '0x' + 'a'.repeat(40)
const { service, calls } = harness((rec) => {
if (rec.url.includes('get-account')) return { json: { status: 'ok', data: { owner: 'hanzo', name: 'alice' } } }
if (rec.url.includes('get-account'))
return {
json: {
status: 'ok',
data: { owner: 'hanzo', name: 'alice', displayName: 'Alice', email: 'alice@hanzo.ai' },
},
}
return { json: { status: 'ok' } }
})
const res = await service.linkWallet(addr)
assert.deepEqual(res, { ok: true, value: addr })
// 1) get-account, 2) update-user keyed by owner/name, column-scoped
// 1) get-account, 2) update-user keyed by owner/name with the WHOLE row
assert.match(calls[0]!.url, /get-account$/)
const upd = calls[1]!
assert.ok(upd.url.includes('/v1/iam/update-user'))
assert.ok(upd.url.includes('id=hanzo%2Falice') || upd.url.includes('id=hanzo/alice'))
assert.ok(upd.url.includes('columns=web3onboard'))
const sent = JSON.parse(upd.body!)
assert.equal(sent.web3onboard, addr)
assert.equal(sent.owner, 'hanzo')
assert.equal(sent.name, 'alice')
// The fields the mutation did not touch MUST survive the round trip.
assert.equal(sent.displayName, 'Alice')
assert.equal(sent.email, 'alice@hanzo.ai')
})
test('saveOnboarding merges properties without dropping existing ones; readOnboarding decodes them', async () => {
const { service, calls } = harness((rec) => {
if (rec.url.includes('get-account'))
return {
json: {
status: 'ok',
data: {
owner: 'hanzo',
name: 'alice',
properties: { 'onboarding.dataSharingConsent': 'true', unrelated: 'kept' },
},
},
}
return { json: { status: 'ok' } }
})
const res = await service.saveOnboarding({ plan: 'pro', completedAt: '2026-08-04T00:00:00Z' })
assert.deepEqual(res, { ok: true, value: true })
const sent = JSON.parse(calls[1]!.body!)
assert.deepEqual(sent.properties, {
'onboarding.dataSharingConsent': 'true',
unrelated: 'kept',
'onboarding.plan': 'pro',
'onboarding.completedAt': '2026-08-04T00:00:00Z',
})
})
test('readOnboarding reports null completedAt/consent/plan for a fresh user', async () => {
const { service } = harness(() => ({ json: { status: 'ok', data: { owner: 'hanzo', name: 'bob' } } }))
assert.deepEqual(await service.readOnboarding(), { completedAt: null, consent: null, plan: null })
})
// The live catalog prices in CENTS (go=900 means $9/mo, priceAnnual=825 means
// $8.25/mo billed annually) and carries other product lines (dns-*) in the
// same list. This test pins both facts with production-shaped rows.
test('listPlans keeps cents unscaled, keeps personal+team only; [] on failure', async () => {
const { service, calls } = harness(() => ({
json: [
{ slug: 'pro', name: 'Pro', category: 'personal', price: 4900, priceAnnual: 4150, popular: true },
{ slug: 'go', name: 'Go', category: 'personal', price: 900, priceAnnual: 825 },
{ slug: 'team', name: 'Team', category: 'team', price: 2500, priceAnnual: 2000 },
{ slug: 'dns-pro', name: 'DNS Pro', category: 'dns', price: 500 }, // other product line → dropped
{ slug: 'enterprise', name: 'Enterprise', category: 'enterprise', price: 0 }, // not self-serve → dropped
{ slug: '', name: 'broken', category: 'personal', price: 500 }, // no slug → dropped
],
}))
const plans = await service.listPlans('https://pay.hanzo.ai/')
assert.equal(calls[0]!.url, 'https://pay.hanzo.ai/v1/billing/plans')
assert.deepEqual(
plans.map((p) => p.slug),
['pro', 'go', 'team'],
)
assert.equal(plans[0]!.priceCents, 4900)
assert.equal(plans[0]!.priceAnnualCents, 4150)
assert.equal(plans[0]!.popular, true)
const down = harness(() => ({ status: 503, json: { error: 'nope' } }))
assert.deepEqual(await down.service.listPlans('https://pay.hanzo.ai'), [])
})
test('linkWallet fails closed when there is no signed-in user', async () => {
+165 -37
View File
@@ -3,20 +3,39 @@
* wallet flow.
*
* One way: every write goes through the canonical IAM REST surface under
* `/v1/iam/*` (the same Casdoor-compat paths the auth client uses), carrying
* the user's bearer token. There is no separate onboarding backend the org
* and project records live in IAM, which is the identity registry.
* `/v1/iam/*` (the same IAM paths the auth client uses). There is no separate
* onboarding backend the org and project records live in IAM, which is the
* identity registry.
*
* listOrgs() GET /v1/iam/get-organizations (user-scoped server-side)
* createOrg() POST /v1/iam/add-organization
* createOrg() POST /v1/iam/onboard (the self-service front door)
* createProject POST /v1/iam/add-project
* linkWallet() client-side wallet connect IAM update-user (host-driven)
*
* Token is supplied by the host through `getAccessToken` (the portal already
* holds the session after login). The service never stores it.
* Founding an org goes through `onboard`, NOT the `add-organization` admin verb.
* They are different doors: add-organization is entity CRUD behind IAM's
* authenticated Guard, filed under owner "admin", and a human may only write an
* org row named after the org they are already in so a person founding their
* FIRST org is refused there by construction (403), and with no bearer at all the
* Guard refuses before that (401). `onboard` is the door built for this: it
* resolves the caller from their own session or bearer and provisions the whole
* org org stamped with them as Founder, them moved in as its owner, one
* metered API key under their own authority as its founder.
*
* Both credentials are offered on every call: `credentials: 'include'` for the
* portal session cookie (a bare portal sign-in mints NO bearer, which is why the
* bearer-only door 401'd), and `Authorization` when the host does hold a token.
* IAM resolves session first, then bearer.
*/
import type { Organization, Project } from '@hanzo/iam'
import type { OrgRef, ProjectRef } from '../domain/types'
import type { Project } from '@hanzo/iam'
import {
PROP_COMPLETED,
PROP_CONSENT,
PROP_PLAN,
type OrgRef,
type PlanInfo,
type ProjectRef,
} from '../domain/types'
/** Result of a write that can fail gracefully (no throw on expected errors). */
export type Result<T> = { readonly ok: true; readonly value: T } | { readonly ok: false; readonly error: string }
@@ -39,10 +58,30 @@ export interface OnboardingService {
* resulting address.
*/
linkWallet(address: string): Promise<Result<string>>
/**
* Read the persisted onboarding record from the signed-in user's
* `properties`. All-null when the user has never completed onboarding
* which is the ONLY case the host should mount the flow for.
*/
readOnboarding(): Promise<{ completedAt: string | null; consent: boolean | null; plan: string | null }>
/**
* Persist onboarding fields onto the user record, read-merge-write. THIS is
* what stops the flow repeating: completion lives on the USER, not in any
* browser storage, so a new device, a cleared cache and a re-login all see
* it done.
*/
saveOnboarding(patch: { completedAt?: string; consent?: boolean; plan?: string }): Promise<Result<true>>
/**
* List purchasable plans from the billing catalog on the PAY origin. The
* catalog is the only price authority this pkg renders what it serves and
* states no price of its own. Returns [] on any failure; the plan step then
* offers the two choices without a price grid.
*/
listPlans(payUrl: string): Promise<PlanInfo[]>
}
export interface OnboardingServiceOptions {
/** IAM origin, no trailing slash (the tenant's `iamUrl`, i.e. hanzo.id). */
/** IAM origin, no trailing slash (the org's `iamUrl`, i.e. hanzo.id). */
readonly iamUrl: string
/** Owning org slug used as the default `owner` for new records. */
readonly orgId: string
@@ -80,16 +119,36 @@ export function createOnboardingService(opts: OnboardingServiceOptions): Onboard
return rows.map(toOrgRef).filter((o): o is OrgRef => o !== null)
}
/**
* Found the caller's own organization through the self-service front door.
*
* The server owns the slug: it derives it from the display name under the ONE
* policy every surface shares, so the returned `org` is authoritative and the
* client's slug preview is only a preview. It answers `{org}` on success and
* `{error}` with a 4xx/5xx on failure not the casibase `{status,msg}`
* envelope the entity CRUD returns so read it directly.
*/
async function createOrg(input: { name: string; displayName: string }): Promise<Result<OrgRef>> {
const url = new URL('/v1/iam/add-organization', base)
const org: Partial<Organization> = {
owner: 'admin',
name: input.name,
displayName: input.displayName,
isPersonal: false,
balanceCurrency: 'USD',
const url = new URL('/v1/iam/onboard', base)
const displayName = input.displayName || input.name
try {
const res = await f(url.toString(), {
method: 'POST',
headers: await authHeaders(),
credentials: 'include',
body: JSON.stringify({ name: displayName }),
})
const body = (await res.json().catch(() => ({}))) as Record<string, unknown>
if (!res.ok) {
const msg = typeof body.error === 'string' && body.error ? body.error : `HTTP ${res.status}`
return { ok: false, error: msg }
}
const org = typeof body.org === 'string' ? body.org : ''
if (!org) return { ok: false, error: 'request failed' }
return { ok: true, value: { name: org, displayName } }
} catch (e) {
return { ok: false, error: String(e) }
}
return writeRecord(url, org, () => ({ name: input.name, displayName: input.displayName }))
}
async function createProject(input: {
@@ -116,45 +175,114 @@ export function createOnboardingService(opts: OnboardingServiceOptions): Onboard
async function linkWallet(address: string): Promise<Result<string>> {
const trimmed = address.trim()
if (!isHexAddress(trimmed)) return { ok: false, error: 'invalid wallet address' }
// Resolve the signed-in user (owner/name) from the session — IAM's
// update-user is keyed by `id=<owner>/<name>`, not a "self" alias.
const account = await getAccount()
if (!account) return { ok: false, error: 'not signed in' }
const res = await updateSelf((row) => {
row.web3onboard = trimmed
})
return res.ok ? { ok: true, value: trimmed } : res
}
async function readOnboarding(): Promise<{
completedAt: string | null
consent: boolean | null
plan: string | null
}> {
const row = await getAccount()
const props = (row?.properties ?? {}) as Record<string, unknown>
const str = (k: string): string | null => (typeof props[k] === 'string' && props[k] ? (props[k] as string) : null)
const consentRaw = str(PROP_CONSENT)
return {
completedAt: str(PROP_COMPLETED),
consent: consentRaw === null ? null : consentRaw === 'true',
plan: str(PROP_PLAN),
}
}
async function saveOnboarding(patch: {
completedAt?: string
consent?: boolean
plan?: string
}): Promise<Result<true>> {
const res = await updateSelf((row) => {
const props = { ...((row.properties as Record<string, string> | undefined) ?? {}) }
if (patch.completedAt !== undefined) props[PROP_COMPLETED] = patch.completedAt
if (patch.consent !== undefined) props[PROP_CONSENT] = String(patch.consent)
if (patch.plan !== undefined) props[PROP_PLAN] = patch.plan
row.properties = props
})
return res.ok ? { ok: true, value: true } : res
}
async function listPlans(payUrl: string): Promise<PlanInfo[]> {
try {
const res = await f(trimSlash(payUrl) + '/v1/billing/plans', { headers: { Accept: 'application/json' } })
if (!res.ok) return []
const body = (await res.json()) as unknown
const rows = Array.isArray(body) ? body : []
return rows
.filter((r): r is Record<string, unknown> => typeof r === 'object' && r !== null)
// Onboarding offers the account plans; other product lines in the same
// catalog (dns-*, enterprise) have their own surfaces.
.filter((r) => r.category === 'personal' || r.category === 'team')
.map((r) => ({
slug: typeof r.slug === 'string' ? r.slug : '',
name: typeof r.name === 'string' ? r.name : '',
description: typeof r.description === 'string' ? r.description : undefined,
// Catalog prices are CENTS (900 = $9/mo); passed through unscaled.
priceCents: typeof r.price === 'number' ? r.price : NaN,
priceAnnualCents: typeof r.priceAnnual === 'number' && r.priceAnnual > 0 ? r.priceAnnual : undefined,
popular: r.popular === true,
}))
.filter((p) => p.slug && p.name && Number.isFinite(p.priceCents) && p.priceCents > 0)
} catch {
return []
}
}
/**
* Read-merge-write the signed-in user's FULL row. IAM's update-user is a
* FULL-ROW write (internal/users Update: "this is a full-row write") and it
* ignores the v1 `columns=` scoping param so a minimal body silently
* blanks every field it omits. The wallet step used to do exactly that,
* wiping displayName/email on every link. Every self-write goes through
* here now: fetch the row, mutate, post the whole thing back.
*/
async function updateSelf(mutate: (row: Record<string, unknown>) => void): Promise<Result<true>> {
const row = await getAccount()
if (!row) return { ok: false, error: 'not signed in' }
const owner = typeof row.owner === 'string' ? row.owner : ''
const name = typeof row.name === 'string' ? row.name : ''
if (!owner || !name) return { ok: false, error: 'not signed in' }
mutate(row)
row.owner = owner
row.name = name
const url = new URL('/v1/iam/update-user', base)
url.searchParams.set('id', `${account.owner}/${account.name}`)
// Scope the write to the single `web3onboard` column so the rest of the
// user row is untouched (Casdoor replaces unscoped writes wholesale).
url.searchParams.set('columns', 'web3onboard')
url.searchParams.set('id', `${owner}/${name}`)
try {
const res = await f(url.toString(), {
method: 'POST',
headers: await authHeaders(),
credentials: 'include',
// Casdoor's User JSON tag is lowercase `web3onboard`; send the full
// owner/name so the row identity is unambiguous on the server.
body: JSON.stringify({ owner: account.owner, name: account.name, web3onboard: trimmed }),
body: JSON.stringify(row),
})
if (!res.ok) return { ok: false, error: `HTTP ${res.status}` }
const body = (await res.json().catch(() => ({}))) as Record<string, unknown>
if (body.status === 'error') return { ok: false, error: msgOf(body) }
return { ok: true, value: trimmed }
return { ok: true, value: true }
} catch (e) {
return { ok: false, error: String(e) }
}
}
/** Read the signed-in user's `{owner, name}` from `/v1/iam/get-account`. */
async function getAccount(): Promise<{ owner: string; name: string } | null> {
/** Read the signed-in user's FULL row from `/v1/iam/get-account`. */
async function getAccount(): Promise<Record<string, unknown> | null> {
const url = new URL('/v1/iam/get-account', base)
try {
const res = await f(url.toString(), { headers: await authHeaders(false), credentials: 'include' })
if (!res.ok) return null
const body = (await res.json()) as Record<string, unknown>
const data = (body.data ?? body) as Record<string, unknown>
const owner = typeof data.owner === 'string' ? data.owner : ''
const name = typeof data.name === 'string' ? data.name : ''
if (!owner || !name) return null
return { owner, name }
if (typeof data !== 'object' || data === null) return null
return data
} catch {
return null
}
@@ -181,10 +309,10 @@ export function createOnboardingService(opts: OnboardingServiceOptions): Onboard
}
}
return { listOrgs, createOrg, createProject, linkWallet }
return { listOrgs, createOrg, createProject, linkWallet, readOnboarding, saveOnboarding, listPlans }
}
/** Pull the array payload out of an IAM list response (`data` or `data2`). */
/** Rows of an IAM list response: the named `data` slot, falling back to the legacy `data2` slot until IAM stops emitting it. */
function extractRows(body: Record<string, unknown>): Record<string, unknown>[] {
const candidate = Array.isArray(body.data) ? body.data : Array.isArray(body.data2) ? body.data2 : []
return candidate.filter((r): r is Record<string, unknown> => typeof r === 'object' && r !== null)
+209 -12
View File
@@ -5,6 +5,7 @@ import {
prevStep,
stepById,
type OnboardingState,
type PlanInfo,
type StepId,
} from '../domain/types'
import type { OnboardingService } from '../service/onboarding'
@@ -24,7 +25,7 @@ import type { OnboardingService } from '../service/onboarding'
*/
export interface OnboardingFlowProps {
readonly service: OnboardingService
/** Brand display name for headings (e.g. the resolved tenant brand). */
/** Brand display name for headings (e.g. the resolved org brand). */
readonly brandName: string
/**
* Host-supplied wallet connector. Returns the connected address (0x) or
@@ -36,6 +37,11 @@ export interface OnboardingFlowProps {
readonly connectWallet?: () => Promise<string | null>
/** Called once the flow reaches `done`, with the final accumulated state. */
readonly onComplete: (state: OnboardingState) => void
/**
* Pay origin serving the billing catalog (GET /v1/billing/plans). The plan
* step renders the catalog's own prices no price is stated here.
*/
readonly payUrl: string
}
interface FlowState {
@@ -58,7 +64,7 @@ function reducer(state: FlowState, action: FlowAction): FlowState {
}
}
export function OnboardingFlow({ service, brandName, connectWallet, onComplete }: OnboardingFlowProps) {
export function OnboardingFlow({ service, brandName, connectWallet, onComplete, payUrl }: OnboardingFlowProps) {
const [state, dispatch] = useReducer(reducer, { step: 'org', data: {} })
// Terminal step: hand the accumulated state back to the host exactly once.
@@ -106,6 +112,12 @@ export function OnboardingFlow({ service, brandName, connectWallet, onComplete }
onNext={advance}
/>
) : null}
{state.step === 'consent' ? (
<ConsentStep service={service} showBack={showBack} onBack={back} onNext={advance} />
) : null}
{state.step === 'plan' ? (
<PlanStep service={service} payUrl={payUrl} showBack={showBack} onBack={back} onNext={advance} />
) : null}
{state.step === 'done' ? <DoneStep brandName={brandName} data={state.data} /> : null}
</div>
)
@@ -153,16 +165,17 @@ function OrgStep({
onNext({ orgName: res.value.name, orgCreated: true })
}
// Onboarding never lists other tenants' organizations — a brand-new user only
// Onboarding never lists other orgs' organizations — a brand-new user only
// ever creates their own org or skips. Listing the org directory would leak
// every tenant's name to anyone who signs up. Joining an existing org happens
// every org's name to anyone who signs up. Joining an existing org happens
// by invitation, handled outside this flow.
return (
<div className="hanzo-id-onboarding-body">
<form onSubmit={create} aria-busy={busy}>
<label>
<form onSubmit={create} className="hanzo-id-form" aria-busy={busy}>
<label className="hanzo-id-field">
<span>Organization name</span>
<input
className="hanzo-id-input"
type="text"
value={displayName}
onChange={(e) => setDisplayName(e.target.value)}
@@ -177,7 +190,7 @@ function OrgStep({
<button type="button" className="hanzo-id-btn ghost" onClick={() => onNext({})} disabled={busy}>
Skip for now
</button>
<button type="submit" className="hanzo-id-btn primary" disabled={busy}>
<button type="submit" className="hanzo-id-btn" disabled={busy}>
{busy ? 'Creating…' : 'Create organization'}
</button>
</div>
@@ -236,7 +249,7 @@ function ProjectStep({
Back
</button>
) : null}
<button type="button" className="hanzo-id-btn primary" onClick={() => onNext({})}>
<button type="button" className="hanzo-id-btn" onClick={() => onNext({})}>
Continue
</button>
</div>
@@ -246,10 +259,11 @@ function ProjectStep({
return (
<div className="hanzo-id-onboarding-body">
<form onSubmit={create} aria-busy={busy}>
<label>
<form onSubmit={create} className="hanzo-id-form" aria-busy={busy}>
<label className="hanzo-id-field">
<span>Project name</span>
<input
className="hanzo-id-input"
type="text"
value={displayName}
onChange={(e) => setDisplayName(e.target.value)}
@@ -268,7 +282,7 @@ function ProjectStep({
<button type="button" className="hanzo-id-btn ghost" onClick={() => onNext({})} disabled={busy}>
Skip
</button>
<button type="submit" className="hanzo-id-btn primary" disabled={busy}>
<button type="submit" className="hanzo-id-btn" disabled={busy}>
{busy ? 'Creating…' : 'Create project'}
</button>
</div>
@@ -334,7 +348,7 @@ function WalletStep({
Skip
</button>
{connectWallet ? (
<button type="button" className="hanzo-id-btn primary" onClick={link} disabled={busy}>
<button type="button" className="hanzo-id-btn" onClick={link} disabled={busy}>
{busy ? 'Connecting…' : 'Connect wallet'}
</button>
) : null}
@@ -343,6 +357,189 @@ function WalletStep({
)
}
// ── Step 4: data-sharing consent ────────────────────────────────────
function ConsentStep({
service,
showBack,
onBack,
onNext,
}: {
service: OnboardingService
showBack: boolean
onBack: () => void
onNext: (patch: Partial<OnboardingState>) => void
}) {
const [agreed, setAgreed] = useState(false)
const [busy, setBusy] = useState(false)
const [error, setError] = useState<string | null>(null)
// Either answer continues; the answer itself is what must exist. It is
// persisted on the USER (not browser storage) before the flow advances, so
// this page is asked exactly once per account, ever.
async function answer() {
setBusy(true)
setError(null)
const res = await service.saveOnboarding({ consent: agreed })
setBusy(false)
if (!res.ok) {
setError(res.error)
return
}
onNext({ dataSharingConsent: agreed })
}
return (
<div className="hanzo-id-onboarding-body">
<div className="hanzo-id-consent">
<p>
Sharing usage data helps improve the models and products you use. It
covers product usage patterns and diagnostics never the content of
your conversations, code, or files. You can change this any time in
account settings.
</p>
<label className="hanzo-id-consent-check">
<input
type="checkbox"
checked={agreed}
onChange={(e) => setAgreed(e.target.checked)}
/>
<span>I agree to share usage data to improve products and models.</span>
</label>
</div>
{error ? <p role="alert" className="hanzo-id-error">{error}</p> : null}
<div className="hanzo-id-onboarding-actions">
{showBack ? (
<button type="button" className="hanzo-id-btn ghost" onClick={onBack} disabled={busy}>
Back
</button>
) : null}
<button type="button" className="hanzo-id-btn" onClick={answer} disabled={busy}>
{busy ? 'Saving…' : 'Continue'}
</button>
</div>
</div>
)
}
// ── Step 5 (last): plan or pay-as-you-go ────────────────────────────
/** Format catalog CENTS as dollars — "$9" or "$8.25", never "$9.00". */
function usd(cents: number): string {
const dollars = cents / 100
return Number.isInteger(dollars) ? `$${dollars}` : `$${dollars.toFixed(2)}`
}
function PlanStep({
service,
payUrl,
showBack,
onBack,
onNext,
}: {
service: OnboardingService
payUrl: string
showBack: boolean
onBack: () => void
onNext: (patch: Partial<OnboardingState>) => void
}) {
const [plans, setPlans] = useState<PlanInfo[] | null>(null)
const [busy, setBusy] = useState<string | null>(null)
const [error, setError] = useState<string | null>(null)
useEffect(() => {
let alive = true
service.listPlans(payUrl).then((p) => {
if (alive) setPlans(p)
})
return () => {
alive = false
}
// payUrl is fixed for the page's life.
// eslint-disable-next-line react-hooks/exhaustive-deps
}, [])
// The choice is persisted (with completion) BEFORE the flow advances, so a
// user who bounces off the payment page still never re-enters onboarding —
// they land on the portal, where the top-up surface remains one click away.
async function choose(choice: string) {
setBusy(choice)
setError(null)
const res = await service.saveOnboarding({
plan: choice,
completedAt: new Date().toISOString(),
})
setBusy(null)
if (!res.ok) {
setError(res.error)
return
}
onNext({ planChoice: choice })
}
return (
<div className="hanzo-id-onboarding-body">
{plans === null ? (
<p className="lede">Loading plans</p>
) : (
<div className="hanzo-id-plans" role="list">
{plans.length === 0 ? (
// The catalog fetch failed or came back empty. Say so — a plan
// picker showing ONLY pay-as-you-go with no explanation reads as
// "there are no plans", which is false. Pay as you go still works,
// and plans remain choosable later from billing.
<p role="alert" className="hanzo-id-plans-empty">
Plans are unavailable right now you can start with pay as you
go and pick a plan later from Billing.
</p>
) : null}
{plans.map((p) => (
<button
key={p.slug}
type="button"
role="listitem"
className={p.popular ? 'hanzo-id-plan popular' : 'hanzo-id-plan'}
onClick={() => choose(p.slug)}
disabled={busy !== null}
aria-busy={busy === p.slug}
>
{p.popular ? <span className="hanzo-id-plan-badge">Popular</span> : null}
<span className="hanzo-id-plan-name">{p.name}</span>
<span className="hanzo-id-plan-price">
{usd(p.priceCents)}/mo
{p.priceAnnualCents ? <em> · {usd(p.priceAnnualCents * 12)}/yr billed annually</em> : null}
</span>
{p.description ? <span className="hanzo-id-plan-desc">{p.description}</span> : null}
</button>
))}
<button
type="button"
role="listitem"
className="hanzo-id-plan payg"
onClick={() => choose('payg')}
disabled={busy !== null}
aria-busy={busy === 'payg'}
>
<span className="hanzo-id-plan-name">Pay as you go</span>
<span className="hanzo-id-plan-price">Prepaid balance · $5 minimum</span>
<span className="hanzo-id-plan-desc">
No subscription. Top up a balance and pay only for what you use.
</span>
</button>
</div>
)}
{error ? <p role="alert" className="hanzo-id-error">{error}</p> : null}
{showBack ? (
<div className="hanzo-id-onboarding-actions">
<button type="button" className="hanzo-id-btn ghost" onClick={onBack} disabled={busy !== null}>
Back
</button>
</div>
) : null}
</div>
)
}
// ── Terminal: success ───────────────────────────────────────────────
function DoneStep({ brandName, data }: { brandName: string; data: OnboardingState }) {
+10 -6
View File
@@ -1,24 +1,28 @@
{
"name": "@hanzo/id-shared",
"version": "0.1.1",
"description": "Shared types + tenant resolver for the Hanzo ID portal. No UI deps.",
"version": "0.1.4",
"description": "Shared types + org resolver for the Hanzo ID portal. No UI deps.",
"license": "BSD-3-Clause",
"type": "module",
"main": "./src/index.ts",
"types": "./src/index.ts",
"exports": {
".": "./src/index.ts",
"./tenant": "./src/tenant.ts",
"./org": "./src/org.ts",
"./brand": "./src/brand.ts",
"./package.json": "./package.json"
},
"files": ["src"],
"files": [
"src"
],
"scripts": {
"tc": "tsc --noEmit",
"build": "tsc --noEmit",
"test": "node --test --experimental-strip-types 'src/**/*.test.ts'"
"build": "tsc --noEmit"
},
"devDependencies": {
"typescript": "^5.9.3"
},
"dependencies": {
"@hanzo/brand": "^1.4.0"
}
}
+20 -2
View File
@@ -1,7 +1,7 @@
import type { BrandContract } from './types'
/**
* Resolve a BrandContract from a tenant's brand package.
* Resolve a BrandContract from a org's brand package.
*
* Each per-org brand pkg (`@hanzo/brand`, `@luxfi/brand`, `@zooai/brand`,
* `@parsdao/brand`) ships a `brand.json` at the package root. This loader
@@ -50,7 +50,7 @@ export async function loadBrand(brandPackage: string): Promise<BrandContract> {
* Last-resort brand when the asset is unreachable after retries. Keeps the
* login form usable (a generic heading) instead of blanking the page. The
* display name is derived from the pkg scope (`@hanzo/brand` -> "Hanzo"); the
* few tenants whose scope differs from their display name are mapped.
* few orgs whose scope differs from their display name are mapped.
*/
function fallbackBrand(brandPackage: string): BrandContract {
const scope = brandPackage.replace(/^@/, '').split('/')[0] ?? 'hanzo'
@@ -76,6 +76,24 @@ export interface BrandRuntime {
readonly accentColor?: string
}
/**
* Neutral identity-portal label always "<Brand> ID". `BrandContract.name`
* is meant to be the bare org display ("Lux"), but some brand packages ship
* the product name ("Lux Exchange" / "Zoo Exchange"), which leaks a sibling
* surface into the IAM portal heading + tab title. Prefer the org orgId
* ("lux" "Lux"); otherwise strip a trailing product word from the brand
* name. So id.lux.network reads "Lux ID", never "Lux Exchange".
*/
export function idBrandLabel(brand: { name: string }, orgId?: string): string {
const cap = (s: string) => (s ? s.charAt(0).toUpperCase() + s.slice(1) : '')
const short =
cap((orgId ?? '').trim()) ||
(brand.name ?? '').replace(/\s+(Exchange|Network|Labs|Foundation|DAO|Wallet)\b.*$/i, '').trim() ||
brand.name ||
'Account'
return `${short} ID`
}
export function toBrandRuntime(b: BrandContract): BrandRuntime {
return {
name: b.name,
+1 -1
View File
@@ -1,3 +1,3 @@
export * from './tenant'
export * from './org'
export * from './brand'
export * from './types'
+284
View File
@@ -0,0 +1,284 @@
/**
* Org-resolver tests run with the Node built-in runner + native TS strip:
*
* pnpm --filter @hanzo/id-shared test
*
* Focus: a host that exists ONLY in the runtime catalog (no built-in entry)
* must resolve to ITS OWN brand and issuer never inherit Hanzo's. This is the
* osage.id brand-leak regression: the catalog carries `brandUrl`, and the
* resolver must map it to `brandPackage` and derive issuer/origin from the host.
*/
import { test } from 'vitest'
import assert from 'node:assert/strict'
import { resolveOrg, parseCatalog, catalogOf } from './org.ts'
// Mirrors the K8s ConfigMap shape: entries carry `brandUrl`, not `brandPackage`.
const CATALOG = {
'lux.id': {
orgId: 'lux',
clientId: 'lux-cloud',
appName: 'lux-cloud',
brandUrl: 'https://cdn.jsdelivr.net/npm/@luxfi/brand@latest/brand.json',
},
'osage.id': {
orgId: 'osage',
clientId: 'osage-id-portal',
appName: 'osage-id',
brandUrl: 'https://cdn.jsdelivr.net/npm/@osage/brand@latest/brand.json',
},
// No brandUrl on purpose — no bootnode brand package is published. This is
// the shape that must NOT fall back to Hanzo.
'id.bootno.de': {
orgId: 'bootnode',
clientId: 'bootnode-platform',
appName: 'bootnode-platform',
},
}
test('a built-in host resolves to its own brand with no catalog', () => {
const t = resolveOrg('hanzo.id')
assert.equal(t.orgId, 'hanzo')
assert.equal(t.brandPackage, '@hanzo/brand')
assert.equal(t.iamUrl, 'https://hanzo.id')
})
test('a catalog entry overrides clientId/appName but keeps a consistent brand', () => {
const t = resolveOrg('lux.id', { catalog: CATALOG })
assert.equal(t.orgId, 'lux')
assert.equal(t.clientId, 'lux-cloud')
assert.equal(t.brandPackage, '@luxfi/brand')
assert.equal(t.iamUrl, 'https://lux.id')
assert.equal(t.publicOrigin, 'https://lux.id')
})
test('a catalog-ONLY host does NOT leak the Hanzo brand (osage.id regression)', () => {
const t = resolveOrg('osage.id', { catalog: CATALOG })
assert.equal(t.orgId, 'osage')
assert.equal(t.clientId, 'osage-id-portal')
// brandUrl is mapped onto brandPackage, and it is NOT Hanzo's.
assert.equal(t.brandPackage, '@osage/brand')
assert.notEqual(t.brandPackage, '@hanzo/brand')
// issuer + origin are the host itself, never hanzo.id.
assert.equal(t.iamUrl, 'https://osage.id')
assert.equal(t.iamIssuer, 'https://osage.id')
assert.equal(t.publicOrigin, 'https://osage.id')
})
test('a catalog host with NO brandUrl still does not leak Hanzo (id.bootno.de)', () => {
const t = resolveOrg('id.bootno.de', { catalog: CATALOG })
assert.equal(t.orgId, 'bootnode')
assert.equal(t.clientId, 'bootnode-platform')
// The interesting case: no brandUrl at all. It must resolve EMPTY so the
// loader shows a neutral wordmark — never another brand's mark. Before this
// host was in the catalog it fell through to the `hanzo` default and the page
// read "Sign in to Hanzo ID" on a Bootnode surface.
assert.equal(t.brandPackage, '')
assert.notEqual(t.brandPackage, '@hanzo/brand')
// issuer + origin derive from the host itself, never hanzo.id.
assert.equal(t.iamUrl, 'https://id.bootno.de')
assert.equal(t.iamIssuer, 'https://id.bootno.de')
assert.equal(t.publicOrigin, 'https://id.bootno.de')
})
test('an unknown host FAILS CLOSED — it never inherits another brand', () => {
const t = resolveOrg('totally-unregistered.example', { catalog: CATALOG })
// Was: DEFAULT_TENANTS['hanzo.id'] — orgId hanzo, @hanzo/brand, and
// iamUrl https://hanzo.id. Empty clientId means the portal refuses rather
// than authenticating as some other brand's IAM application.
assert.equal(t.orgId, '')
assert.equal(t.clientId, '')
assert.equal(t.brandPackage, '')
assert.equal(t.iamUrl, 'https://totally-unregistered.example')
assert.notEqual(t.iamUrl, 'https://hanzo.id')
})
test('a catalog host with a FAILED catalog fetch does not leak Hanzo', () => {
// The live failure mode: App.tsx tolerates a failed /config.json, so these
// real hosts resolve with NO catalog at all. Every one of them used to come
// back as Hanzo — same brand, same mark, and credentials posted at hanzo.id.
for (const host of [
'zoolabs.id',
'www.zoolabs.id',
'id.zoo.network',
'id.lux.network',
'iam.lux.network',
'id.pars.network',
'id.bootno.de',
]) {
const t = resolveOrg(host) // no catalog — the fetch failed
assert.notEqual(t.orgId, 'hanzo', `${host} leaked orgId hanzo`)
assert.notEqual(t.brandPackage, '@hanzo/brand', `${host} leaked the Hanzo mark`)
assert.notEqual(t.iamUrl, 'https://hanzo.id', `${host} would post credentials at hanzo.id`)
assert.equal(t.iamUrl, `https://${host}`)
assert.equal(t.iamIssuer, `https://${host}`)
}
})
test('zoo.id is gone — it is NXDOMAIN and must not be a built-in', () => {
const t = resolveOrg('zoo.id')
assert.equal(t.orgId, '')
assert.equal(t.clientId, '')
})
test('pars built-in uses the working pars-console portal app (not the missing pars-id)', () => {
const t = resolveOrg('pars.id')
assert.equal(t.clientId, 'pars-console')
assert.equal(t.brandPackage, '@parsdao/brand')
})
test('osage built-in resolves to Osage even with NO catalog (fallback safety)', () => {
const t = resolveOrg('osage.id')
assert.equal(t.orgId, 'osage')
assert.equal(t.brandPackage, '@osage/brand')
assert.notEqual(t.brandPackage, '@hanzo/brand')
assert.equal(t.iamUrl, 'https://osage.id')
})
test('an unknown host keeps its own origin AND does not inherit an org', () => {
const t = resolveOrg('preview.example.com')
// This assertion used to be `orgId === 'hanzo'` — it pinned the cross-brand
// fallback as intended behaviour. Keeping its own origin is right; being
// handed Hanzo's org, mark and issuer is the defect that shipped behind it.
assert.equal(t.orgId, '')
assert.equal(t.publicOrigin, 'https://preview.example.com')
})
test('parseCatalog tolerates junk', () => {
assert.deepEqual(parseCatalog(undefined), {})
assert.deepEqual(parseCatalog(null), {})
assert.deepEqual(parseCatalog('not json'), {})
assert.deepEqual(parseCatalog('{"osage.id":{"orgId":"osage"}}'), {
'osage.id': { orgId: 'osage' },
})
})
/**
* THE SOCIAL-LOGIN REGRESSION. Google and GitHub each accept a fixed list of
* redirect URIs and we hold ONE shared OAuth client per provider, so every
* brand must send the same `redirect_uri` or the provider answers
* `redirect_uri_mismatch`.
*
* `oauthCallbackOrigin` used to default to `publicOrigin` the brand's own
* host and no catalog entry overrode it, so each property sent a different
* URI and social login could work on at most one of them. The failure surfaced
* at Google, not here, which is why it read as a credentials problem for days.
*
* The default is the ORG'S HOSTED ID HOST hanzo.id for hanzo, lux.id for lux.
*
* This test used to assert the default was `iamIssuer`, which was the FIRST
* attempt at the fix and was abandoned in the same change that shipped the real
* one: `hostSkeleton` derives the issuer from the REQUEST HOST, so on an app
* host the issuer IS the brand host and the bug is unchanged. `org.ts` says so
* in place. The assertion was left behind and had been failing on `main` ever
* since, against code that is correct.
*
* It was also asking the question on hosts that carry no org: `resolveOrg` was
* called with NO catalog, so hanzo.app and hanzo.chat fell to the deliberate
* unknown-host skeleton (empty orgId, fail closed, never another brand's
* config). No design can make two ORG-LESS hosts agree on one org's callback
* the old assertion could not have passed under either default.
*
* So ask it the way production does: the catalog (`/config.json`) is what
* supplies the orgId, and the invariant that matters is one registered URI PER
* ORG.
*/
test('the provider callback is the org hosted-ID host, one per org', () => {
const catalog = parseCatalog(
JSON.stringify({
'hanzo.app': { orgId: 'hanzo', clientId: 'hanzo-app' },
'hanzo.chat': { orgId: 'hanzo', clientId: 'hanzo-chat' },
'console.hanzo.ai': { orgId: 'hanzo', clientId: 'hanzo-cloud' },
'id.lux.network': { orgId: 'lux', clientId: 'lux-id' },
}),
)
const app = resolveOrg('hanzo.app', { catalog })
const chat = resolveOrg('hanzo.chat', { catalog })
const consoleHost = resolveOrg('console.hanzo.ai', { catalog })
const portal = resolveOrg('hanzo.id', { catalog }) // built-in; needs no row
// Whatever the property, ONE registered redirect_uri serves them all.
for (const t of [app, chat, consoleHost, portal]) {
assert.equal(t.oauthCallbackOrigin, 'https://hanzo.id')
}
// And it is NOT the brand host — the precise shape of the bug.
assert.notEqual(app.oauthCallbackOrigin, app.publicOrigin)
assert.notEqual(chat.oauthCallbackOrigin, chat.publicOrigin)
assert.notEqual(consoleHost.oauthCallbackOrigin, consoleHost.publicOrigin)
// Per-ORG, not global, and NOT the issuer: lux gets lux.id. This is the
// assertion that pins the abandoned first attempt out of the codebase —
// under `iamIssuer` this host would send its own origin and break again.
const lux = resolveOrg('id.lux.network', { catalog })
assert.equal(lux.oauthCallbackOrigin, 'https://lux.id')
assert.notEqual(lux.oauthCallbackOrigin, lux.iamIssuer)
})
test('an explicit catalog oauthCallbackOrigin still wins over the issuer', () => {
const catalog = parseCatalog(
JSON.stringify({ 'per-host.example': { oauthCallbackOrigin: 'https://its-own-client.example' } }),
)
const org = resolveOrg('per-host.example', { catalog })
assert.equal(org.oauthCallbackOrigin, 'https://its-own-client.example')
})
// The catalog key is the SERVER'S name. Reading the wrong one returns undefined,
// the app falls back to a global the runtime never injects, and every
// catalog-only host silently drops to the bundled defaults — a total catalog
// outage that looks like nothing at all. Pinned here, next to the resolver it
// feeds, because the last time these tests were deleted the function went with
// them and the image stopped building.
test('catalogOf reads the key the runtime actually serves', () => {
const served = { iamTenantConfigJson: '{"hanzo.id":{"clientId":"hanzo-console"}}', v: 1 }
assert.equal(catalogOf(served), '{"hanzo.id":{"clientId":"hanzo-console"}}')
assert.equal(parseCatalog(catalogOf(served))['hanzo.id']!.clientId, 'hanzo-console')
// Anything else is not the catalog: no guessing, no second accepted key.
assert.equal(catalogOf({ iamOrgConfigJson: '{"hanzo.id":{}}' }), undefined)
assert.equal(catalogOf({}), undefined)
assert.equal(catalogOf(null), undefined)
assert.equal(catalogOf(undefined), undefined)
assert.equal(catalogOf({ iamTenantConfigJson: 42 }), undefined)
})
// The regression that broke the build: App.tsx imports catalogOf from the
// package barrel, so exporting it from org.ts alone is not enough.
test('catalogOf is reachable from the package barrel', async () => {
const barrel = await import('./index.ts')
assert.equal(typeof barrel.catalogOf, 'function')
assert.equal(barrel.catalogOf({ iamTenantConfigJson: '{}' }), '{}')
})
// The fallback and the catalog must name the SAME application per host.
//
// resolveOrg spreads the catalog OVER the built-in table, so a host where the
// two disagree authenticates as one app or the other depending on whether
// /config.json won a race. hanzo.id said `hanzo-id` in the table and
// `hanzo-console` in the catalog, and those two differ in enableSignUp — false
// vs true. On a load where the fetch did not arrive, the signup form rendered
// and the POST came back "the application does not allow to sign up new
// account". Intermittent, device-dependent, and reported from a phone.
//
// This asserts the property that makes the race harmless: for a host present in
// BOTH, resolving with and without the catalog yields the same clientId. It is
// checked against the catalog fixture above, which mirrors the ConfigMap in
// universe/infra/k8s/id/configmap.yaml — so a change to one side that is not
// mirrored in the other fails here rather than in a customer's browser.
test('the built-in fallback names the SAME app as the catalog, per host', () => {
const CANONICAL: Record<string, string> = {
'hanzo.id': 'hanzo-console',
'lux.id': 'lux-cloud',
'pars.id': 'pars-console',
}
for (const [host, clientId] of Object.entries(CANONICAL)) {
// No catalog — the failed-/config.json path a real visitor can hit.
assert.equal(resolveOrg(host).clientId, clientId, `${host} fallback`)
// With the catalog — the ordinary path.
assert.equal(
resolveOrg(host, { catalog: { [host]: { clientId, appName: clientId } } }).clientId,
clientId,
`${host} with catalog`,
)
}
})
+286
View File
@@ -0,0 +1,286 @@
import type { OrgConfig } from './types'
/**
* Resolve a OrgConfig by hostname.
*
* Resolution order (first hit wins):
* 1. `IAM_TENANT_CONFIG_JSON` runtime catalog (set in K8s ConfigMap, served
* to the browser via `/config.json` at pod startup).
* 2. Built-in defaults, for the identity hosts that have one.
* 3. A skeleton derived from the REQUESTED HOST never another brand.
*
* There is deliberately no cross-brand default. An unknown host resolves to
* itself with an empty clientId and fails closed, because the alternative is a
* visitor on one brand's host being shown another brand's login and posting
* credentials there.
*
* No hardcoded hostname switches anywhere downstream. Adding a org
* means editing the runtime catalog, never editing source.
*/
const TRIM_TRAILING_SLASH = (s: string): string => s.replace(/\/+$/, '')
/**
* Built-in orgs for the four canonical identity hosts.
*
* `iamUrl` is the per-brand OIDC ISSUER the host that serves
* `/.well-known/openid-configuration` and the `/v1/iam/*` surface. Per
* HIP-0111 this is the brand's own `*.id` host (hanzo.id / lux.id / ),
* NOT `iam.hanzo.ai`: discovery must be host-relative so the SDK never
* resolves to the wrong origin (or the IAM SPA HTML catch-all). The IAM
* backend org-scopes on the `organization` body param; one backend
* serves every brand behind its own issuer host.
*
* `clientId` MUST name the SAME application the runtime catalog names for that
* host. This table is a fallback for a failed `/config.json`, not a second
* opinion `resolveOrg` spreads the catalog OVER it, so any key where the two
* disagree makes the host authenticate as one app or the other depending on
* whether a network fetch won a race.
*
* That is not hypothetical. `hanzo.id` said `hanzo-id` here while the catalog
* says `hanzo-console`, and the two differ in `enableSignUp` false vs true.
* So on a load where `/config.json` did not arrive, hanzo.id authenticated as
* `hanzo-id`, the signup form rendered, and the POST came back "the application
* does not allow to sign up new account". Intermittent, device-dependent, and
* indistinguishable from a bad password to the person hitting it. Reported from
* a phone, where a dropped request is ordinary.
*
* Both now match the catalog (`universe/infra/k8s/id/configmap.yaml`,
* SPA_IAM_TENANT_CONFIG_JSON). `org.test.ts` pins them, so a future edit to one
* side has to confront the other. Fixing the fallback to be SURVIVABLE was the
* wrong shape the fix is that there is only one answer per host.
*/
const DEFAULT_TENANTS: Record<string, OrgConfig> = {
'hanzo.id': {
orgId: 'hanzo',
iamUrl: 'https://hanzo.id',
iamIssuer: 'https://hanzo.id',
clientId: 'hanzo-console',
appName: 'hanzo-console',
publicOrigin: 'https://hanzo.id',
brandPackage: '@hanzo/brand',
},
'lux.id': {
orgId: 'lux',
iamUrl: 'https://lux.id',
iamIssuer: 'https://lux.id',
clientId: 'lux-cloud',
appName: 'lux-cloud',
publicOrigin: 'https://lux.id',
brandPackage: '@luxfi/brand',
},
'pars.id': {
orgId: 'pars',
iamUrl: 'https://pars.id',
iamIssuer: 'https://pars.id',
// The portal app is `pars-console` (it carries the https://pars.id/callback
// redirect); a bare `pars-id` app does not exist in IAM.
clientId: 'pars-console',
appName: 'pars-console',
publicOrigin: 'https://pars.id',
brandPackage: '@parsdao/brand',
},
// Osage is served by this portal too; without a built-in it would fall back
// to the Hanzo default and leak the wrong brand if the runtime catalog ever
// fails to load. (osage-id-portal is pre-launch — no IAM app yet — but the
// brand must read as Osage, never Hanzo.)
'osage.id': {
orgId: 'osage',
iamUrl: 'https://osage.id',
iamIssuer: 'https://osage.id',
clientId: 'osage-id-portal',
appName: 'osage-id',
publicOrigin: 'https://osage.id',
brandPackage: '@osage/brand',
},
'www.osage.id': {
orgId: 'osage',
iamUrl: 'https://www.osage.id',
iamIssuer: 'https://www.osage.id',
clientId: 'osage-id-portal',
appName: 'osage-id',
publicOrigin: 'https://www.osage.id',
brandPackage: '@osage/brand',
},
}
/**
* A runtime catalog entry as it appears in the K8s ConfigMap / `/config.json`.
* It carries the human-authored shape notably `brandUrl` (a CDN URL), which
* this module maps onto the code-facing `brandPackage`. All fields optional;
* whatever is present overrides the host-derived base.
*/
export type CatalogEntry = Partial<OrgConfig> & {
/** CDN URL of the brand package, e.g. `…/npm/@osage/brand@latest/brand.json`. */
readonly brandUrl?: string
}
export interface ResolveOptions {
/** Optional runtime catalog (parsed from IAM_TENANT_CONFIG_JSON or /config.json). */
readonly catalog?: Record<string, CatalogEntry>
}
export function resolveOrg(hostname: string, opts: ResolveOptions = {}): OrgConfig {
const host = stripPort(hostname).toLowerCase()
const catalogEntry = opts.catalog?.[host]
const builtIn = DEFAULT_TENANTS[host]
if (catalogEntry || builtIn) {
// Base = the built-in org if one exists, else a skeleton derived from
// THIS host. Never another brand's config: a catalog-only host (osage.id,
// zoolabs.id) must not inherit Hanzo's issuer or brand package.
const base = builtIn ?? hostSkeleton(host)
const merged: OrgConfig = { ...base, ...fromCatalog(catalogEntry) } as OrgConfig
return normalize(merged)
}
// Unknown host → derive from the host ITSELF. Never another brand's org.
//
// This used to return DEFAULT_TENANTS[`${defaultOrg}.id`], i.e. Hanzo's. Eight
// real hosts have no built-in entry and live only in the runtime catalog —
// zoolabs.id, www.zoolabs.id, id.zoo.network, id.lux.network, iam.lux.network,
// id.pars.network, id.bootno.de, iam.hanzo.ai — and App.tsx deliberately
// tolerates a failed /config.json fetch. So whenever that fetch failed, a Zoo,
// Lux, Pars or Bootnode visitor was handed orgId `hanzo`, `@hanzo/brand` and
// iamUrl `https://hanzo.id`: shown "Sign in to Hanzo ID" under the Hanzo mark
// and POSTING THEIR CREDENTIALS AT hanzo.id. The comment ten lines up already
// promised this could not happen ("Never another brand's config") — it held
// only while the catalog loaded.
//
// The skeleton carries an empty clientId, so the portal fails closed rather
// than silently authenticating as some other brand's IAM application. A login
// page that cannot resolve its org must refuse, not guess.
return normalize(hostSkeleton(host))
}
/**
* A host-derived org skeleton for a catalog-only host (no built-in entry).
* URLs point at the host itself so nothing leaks from another brand; the
* catalog entry spread over this supplies orgId / clientId / appName /
* brandPackage. brandPackage defaults empty the brand loader falls back to a
* neutral wordmark rather than showing the wrong brand.
*/
function hostSkeleton(host: string): OrgConfig {
return {
orgId: '',
iamUrl: `https://${host}`,
iamIssuer: `https://${host}`,
clientId: '',
appName: '',
publicOrigin: `https://${host}`,
brandPackage: '',
}
}
/**
* Project a catalog entry onto a OrgConfig patch, mapping `brandUrl`
* `brandPackage` (the code-facing field) when an explicit `brandPackage` isn't
* given. Only defined string fields are emitted, so the host-derived base shows
* through for anything the entry omits.
*/
function fromCatalog(entry: CatalogEntry | undefined): Partial<OrgConfig> {
if (!entry) return {}
const out: Record<string, string> = {}
for (const k of ['orgId', 'loginOrg', 'iamUrl', 'iamIssuer', 'clientId', 'appName', 'publicOrigin', 'oauthCallbackOrigin', 'brandPackage', 'payUrl'] as const) {
const v = entry[k]
if (typeof v === 'string' && v.length > 0) out[k] = v
}
if (!out.brandPackage && typeof entry.brandUrl === 'string') {
const pkg = brandPackageFromUrl(entry.brandUrl)
if (pkg) out.brandPackage = pkg
}
return out as Partial<OrgConfig>
}
/**
* Extract the npm package name from a CDN brand URL, e.g.
* `https://cdn.jsdelivr.net/npm/@osage/brand@latest/brand.json` `@osage/brand`.
*/
function brandPackageFromUrl(url: string): string {
const m = /\/npm\/(@[^/]+\/[^@/]+|[^@/]+)(?:@|\/)/.exec(url)
return m ? m[1]! : ''
}
function stripPort(h: string): string {
return h.replace(/:\d+$/, '')
}
/**
* The org's hosted-ID origin hanzo.id for hanzo, lux.id for lux, and so on
* derived from DEFAULT_TENANTS so the `.id` hosts are declared exactly once.
* Returns '' for an org with no hosted ID host (local dev, per-host clients).
*/
function idOriginFor(orgId: string): string {
if (!orgId) return ''
for (const [host, t] of Object.entries(DEFAULT_TENANTS)) {
if (t.orgId === orgId && host.endsWith('.id')) return `https://${host}`
}
return ''
}
function normalize(t: OrgConfig): OrgConfig {
const publicOrigin = TRIM_TRAILING_SLASH(t.publicOrigin)
const iamIssuer = TRIM_TRAILING_SLASH(t.iamIssuer || t.iamUrl)
return {
...t,
iamUrl: TRIM_TRAILING_SLASH(t.iamUrl),
iamIssuer,
publicOrigin,
// ONE provider callback for the whole org, and it is the hosted ID host —
// hanzo.id for hanzo, lux.id for lux. A social provider must never learn
// about individual apps: it holds ONE OAuth client with ONE registered
// redirect_uri, so every app's hop has to arrive from the same origin or
// the provider answers `redirect_uri_mismatch`.
//
// This defaulted to `publicOrigin` — the BRAND'S OWN host — and no catalog
// entry overrode it, so hanzo.app sent hanzo.app/callback, hanzo.chat sent
// hanzo.chat/callback, console sent console.hanzo.ai/callback, and social
// login could work on at most ONE property. Defaulting to `iamIssuer` does
// NOT fix it: hostSkeleton derives the issuer from the REQUEST HOST too, so
// it is per-brand for exactly the same reason. It has to be a per-ORG
// constant, which is what idOriginFor reads out of DEFAULT_TENANTS.
//
// The failure surfaced at Google, not here, which is why it read as a
// credentials or KMS problem for days. It never was: the client_id reached
// Google intact every time and Google's own error decoded to
// `redirect_uri_mismatch`.
//
// hanzo.id then completes the exchange and forwards the browser back to the
// originating app, so the app hosts stay entirely invisible to the provider.
oauthCallbackOrigin: TRIM_TRAILING_SLASH(
t.oauthCallbackOrigin || idOriginFor(t.orgId) || iamIssuer || publicOrigin,
),
}
}
/**
* Pull the catalog JSON string out of the `/config.json` payload.
*
* The key is the SERVER'S name, not ours: the runtime serves
* `{"iamTenantConfigJson": "<json>", "v": 1}`. Reading any other key yields
* `undefined`, App.tsx falls back to a global the runtime never injects, and
* EVERY catalog-only host silently drops to the bundled defaults a total
* catalog outage that looks like nothing at all. That is why this is one named
* function pinned by tests next to the resolver it feeds, rather than an inline
* property read at the call site.
*
* Restored after c153004 deleted it together with its tests while App.tsx still
* imported it: the id image failed to build from that commit onward
* (`"catalogOf" is not exported by "../../pkgs/shared/src/index.ts"`), which is
* why 0.2.22 never existed. Deleting a function and its tests in one move
* removes the thing that would have reported the deletion.
*/
export function catalogOf(payload: unknown): string | undefined {
if (!payload || typeof payload !== 'object') return undefined
const raw = (payload as { iamTenantConfigJson?: unknown }).iamTenantConfigJson
return typeof raw === 'string' ? raw : undefined
}
/** Parse the runtime catalog JSON safely; returns {} on any error. */
export function parseCatalog(raw: string | undefined | null): Record<string, Partial<OrgConfig>> {
if (!raw) return {}
try {
const parsed = JSON.parse(raw)
return parsed && typeof parsed === 'object' ? parsed : {}
} catch {
return {}
}
}
-87
View File
@@ -1,87 +0,0 @@
/**
* Tenant-resolver tests run with the Node built-in runner + native TS strip:
*
* pnpm --filter @hanzo/id-shared test
*
* Focus: a host that exists ONLY in the runtime catalog (no built-in entry)
* must resolve to ITS OWN brand and issuer never inherit Hanzo's. This is the
* osage.id brand-leak regression: the catalog carries `brandUrl`, and the
* resolver must map it to `brandPackage` and derive issuer/origin from the host.
*/
import { test } from 'node:test'
import assert from 'node:assert/strict'
import { resolveTenant, parseCatalog } from './tenant.ts'
// Mirrors the K8s ConfigMap shape: entries carry `brandUrl`, not `brandPackage`.
const CATALOG = {
'lux.id': {
orgId: 'lux',
clientId: 'lux-cloud',
appName: 'lux-cloud',
brandUrl: 'https://cdn.jsdelivr.net/npm/@luxfi/brand@latest/brand.json',
},
'osage.id': {
orgId: 'osage',
clientId: 'osage-id-portal',
appName: 'osage-id',
brandUrl: 'https://cdn.jsdelivr.net/npm/@osage/brand@latest/brand.json',
},
}
test('a built-in host resolves to its own brand with no catalog', () => {
const t = resolveTenant('hanzo.id')
assert.equal(t.orgId, 'hanzo')
assert.equal(t.brandPackage, '@hanzo/brand')
assert.equal(t.iamUrl, 'https://hanzo.id')
})
test('a catalog entry overrides clientId/appName but keeps a consistent brand', () => {
const t = resolveTenant('lux.id', { catalog: CATALOG })
assert.equal(t.orgId, 'lux')
assert.equal(t.clientId, 'lux-cloud')
assert.equal(t.brandPackage, '@luxfi/brand')
assert.equal(t.iamUrl, 'https://lux.id')
assert.equal(t.publicOrigin, 'https://lux.id')
})
test('a catalog-ONLY host does NOT leak the Hanzo brand (osage.id regression)', () => {
const t = resolveTenant('osage.id', { catalog: CATALOG })
assert.equal(t.orgId, 'osage')
assert.equal(t.clientId, 'osage-id-portal')
// brandUrl is mapped onto brandPackage, and it is NOT Hanzo's.
assert.equal(t.brandPackage, '@osage/brand')
assert.notEqual(t.brandPackage, '@hanzo/brand')
// issuer + origin are the host itself, never hanzo.id.
assert.equal(t.iamUrl, 'https://osage.id')
assert.equal(t.iamIssuer, 'https://osage.id')
assert.equal(t.publicOrigin, 'https://osage.id')
})
test('pars built-in uses the working pars-console portal app (not the missing pars-id)', () => {
const t = resolveTenant('pars.id')
assert.equal(t.clientId, 'pars-console')
assert.equal(t.brandPackage, '@parsdao/brand')
})
test('osage built-in resolves to Osage even with NO catalog (fallback safety)', () => {
const t = resolveTenant('osage.id')
assert.equal(t.orgId, 'osage')
assert.equal(t.brandPackage, '@osage/brand')
assert.notEqual(t.brandPackage, '@hanzo/brand')
assert.equal(t.iamUrl, 'https://osage.id')
})
test('an unknown host falls back to the default org but keeps its own origin', () => {
const t = resolveTenant('preview.example.com')
assert.equal(t.orgId, 'hanzo')
assert.equal(t.publicOrigin, 'https://preview.example.com')
})
test('parseCatalog tolerates junk', () => {
assert.deepEqual(parseCatalog(undefined), {})
assert.deepEqual(parseCatalog(null), {})
assert.deepEqual(parseCatalog('not json'), {})
assert.deepEqual(parseCatalog('{"osage.id":{"orgId":"osage"}}'), {
'osage.id': { orgId: 'osage' },
})
})
-208
View File
@@ -1,208 +0,0 @@
import type { TenantConfig } from './types'
/**
* Resolve a TenantConfig by hostname.
*
* Resolution order (first hit wins):
* 1. `IAM_TENANT_CONFIG_JSON` runtime catalog (set in K8s ConfigMap, served
* to the browser via `/config.json` at pod startup).
* 2. Built-in defaults for the four canonical Hanzo identity hosts.
* 3. `IAM_DEFAULT_ORG` (or "hanzo") fallback used for unknown hosts
* (preview deploys, local dev, custom domains pre-launch).
*
* No hardcoded hostname switches anywhere downstream. Adding a tenant
* means editing the runtime catalog, never editing source.
*/
const TRIM_TRAILING_SLASH = (s: string): string => s.replace(/\/+$/, '')
/**
* Built-in tenants for the four canonical identity hosts.
*
* `iamUrl` is the per-brand OIDC ISSUER the host that serves
* `/.well-known/openid-configuration` and the `/v1/iam/*` surface. Per
* HIP-0111 this is the brand's own `*.id` host (hanzo.id / lux.id / ),
* NOT `iam.hanzo.ai`: discovery must be host-relative so the SDK never
* resolves to the wrong origin (or the IAM SPA HTML catch-all). The IAM
* backend tenant-scopes on the `organization` body param; one backend
* serves every brand behind its own issuer host.
*
* `clientId` is the brand `-id` app registered in `init_data.json`
* (`hanzo-id`, `lux-id`, ) so the portal authenticates as that app the
* same app whose enabled providers (password + GitHub + Google + Web3)
* `get-app-login` reports.
*/
const DEFAULT_TENANTS: Record<string, TenantConfig> = {
'hanzo.id': {
orgId: 'hanzo',
iamUrl: 'https://hanzo.id',
iamIssuer: 'https://hanzo.id',
clientId: 'hanzo-id',
appName: 'hanzo-id',
publicOrigin: 'https://hanzo.id',
brandPackage: '@hanzo/brand',
},
'lux.id': {
orgId: 'lux',
iamUrl: 'https://lux.id',
iamIssuer: 'https://lux.id',
clientId: 'lux-id',
appName: 'lux-id',
publicOrigin: 'https://lux.id',
brandPackage: '@luxfi/brand',
},
'zoo.id': {
orgId: 'zoo',
iamUrl: 'https://zoo.id',
iamIssuer: 'https://zoo.id',
clientId: 'zoo-id',
appName: 'zoo-id',
publicOrigin: 'https://zoo.id',
brandPackage: '@zooai/brand',
},
'pars.id': {
orgId: 'pars',
iamUrl: 'https://pars.id',
iamIssuer: 'https://pars.id',
// The portal app is `pars-console` (it carries the https://pars.id/callback
// redirect); a bare `pars-id` app does not exist in IAM.
clientId: 'pars-console',
appName: 'pars-console',
publicOrigin: 'https://pars.id',
brandPackage: '@parsdao/brand',
},
// Osage is served by this portal too; without a built-in it would fall back
// to the Hanzo default and leak the wrong brand if the runtime catalog ever
// fails to load. (osage-id-portal is pre-launch — no IAM app yet — but the
// brand must read as Osage, never Hanzo.)
'osage.id': {
orgId: 'osage',
iamUrl: 'https://osage.id',
iamIssuer: 'https://osage.id',
clientId: 'osage-id-portal',
appName: 'osage-id',
publicOrigin: 'https://osage.id',
brandPackage: '@osage/brand',
},
'www.osage.id': {
orgId: 'osage',
iamUrl: 'https://www.osage.id',
iamIssuer: 'https://www.osage.id',
clientId: 'osage-id-portal',
appName: 'osage-id',
publicOrigin: 'https://www.osage.id',
brandPackage: '@osage/brand',
},
}
/**
* A runtime catalog entry as it appears in the K8s ConfigMap / `/config.json`.
* It carries the human-authored shape notably `brandUrl` (a CDN URL), which
* this module maps onto the code-facing `brandPackage`. All fields optional;
* whatever is present overrides the host-derived base.
*/
export type CatalogEntry = Partial<TenantConfig> & {
/** CDN URL of the brand package, e.g. `…/npm/@osage/brand@latest/brand.json`. */
readonly brandUrl?: string
}
export interface ResolveOptions {
/** Optional runtime catalog (parsed from IAM_TENANT_CONFIG_JSON or /config.json). */
readonly catalog?: Record<string, CatalogEntry>
/** Default org slug when host has no entry. */
readonly defaultOrg?: string
}
export function resolveTenant(hostname: string, opts: ResolveOptions = {}): TenantConfig {
const host = stripPort(hostname).toLowerCase()
const catalogEntry = opts.catalog?.[host]
const builtIn = DEFAULT_TENANTS[host]
if (catalogEntry || builtIn) {
// Base = the built-in tenant if one exists, else a skeleton derived from
// THIS host. Never another brand's config: a catalog-only host (osage.id,
// zoolabs.id) must not inherit Hanzo's issuer or brand package.
const base = builtIn ?? hostSkeleton(host)
const merged: TenantConfig = { ...base, ...fromCatalog(catalogEntry) } as TenantConfig
return normalize(merged)
}
const defaultOrg = opts.defaultOrg ?? 'hanzo'
const fallback = DEFAULT_TENANTS[`${defaultOrg}.id`] ?? DEFAULT_TENANTS['hanzo.id']
return normalize({ ...fallback, publicOrigin: `https://${host}` })
}
/**
* A host-derived tenant skeleton for a catalog-only host (no built-in entry).
* URLs point at the host itself so nothing leaks from another brand; the
* catalog entry spread over this supplies orgId / clientId / appName /
* brandPackage. brandPackage defaults empty the brand loader falls back to a
* neutral wordmark rather than showing the wrong brand.
*/
function hostSkeleton(host: string): TenantConfig {
return {
orgId: '',
iamUrl: `https://${host}`,
iamIssuer: `https://${host}`,
clientId: '',
appName: '',
publicOrigin: `https://${host}`,
brandPackage: '',
}
}
/**
* Project a catalog entry onto a TenantConfig patch, mapping `brandUrl`
* `brandPackage` (the code-facing field) when an explicit `brandPackage` isn't
* given. Only defined string fields are emitted, so the host-derived base shows
* through for anything the entry omits.
*/
function fromCatalog(entry: CatalogEntry | undefined): Partial<TenantConfig> {
if (!entry) return {}
const out: Record<string, string> = {}
for (const k of ['orgId', 'loginOrg', 'iamUrl', 'iamIssuer', 'clientId', 'appName', 'publicOrigin', 'oauthCallbackOrigin', 'brandPackage'] as const) {
const v = entry[k]
if (typeof v === 'string' && v.length > 0) out[k] = v
}
if (!out.brandPackage && typeof entry.brandUrl === 'string') {
const pkg = brandPackageFromUrl(entry.brandUrl)
if (pkg) out.brandPackage = pkg
}
return out as Partial<TenantConfig>
}
/**
* Extract the npm package name from a CDN brand URL, e.g.
* `https://cdn.jsdelivr.net/npm/@osage/brand@latest/brand.json` `@osage/brand`.
*/
function brandPackageFromUrl(url: string): string {
const m = /\/npm\/(@[^/]+\/[^@/]+|[^@/]+)(?:@|\/)/.exec(url)
return m ? m[1]! : ''
}
function stripPort(h: string): string {
return h.replace(/:\d+$/, '')
}
function normalize(t: TenantConfig): TenantConfig {
const publicOrigin = TRIM_TRAILING_SLASH(t.publicOrigin)
return {
...t,
iamUrl: TRIM_TRAILING_SLASH(t.iamUrl),
iamIssuer: TRIM_TRAILING_SLASH(t.iamIssuer || t.iamUrl),
publicOrigin,
// The social OAuth hop's redirect_uri must hit the provider's registered
// callback host. Default to this host; brands sharing a single OAuth client
// override it (via the catalog) to that client's registered origin.
oauthCallbackOrigin: TRIM_TRAILING_SLASH(t.oauthCallbackOrigin || publicOrigin),
}
}
/** Parse the runtime catalog JSON safely; returns {} on any error. */
export function parseCatalog(raw: string | undefined | null): Record<string, Partial<TenantConfig>> {
if (!raw) return {}
try {
const parsed = JSON.parse(raw)
return parsed && typeof parsed === 'object' ? parsed : {}
} catch {
return {}
}
}
+20 -27
View File
@@ -1,12 +1,12 @@
/**
* Per-tenant configuration resolved at runtime.
* Per-org configuration resolved at runtime.
*
* One image, many hosts. The portal resolves a TenantConfig for each
* One image, many hosts. The portal resolves a OrgConfig for each
* incoming request by hostname; the IAM backend, OAuth client id, and
* brand package are all wired from this single object.
*/
export interface TenantConfig {
/** Tenant org slug (matches the JWT `owner` claim and the IAM `<org>-<app>` namespace). */
export interface OrgConfig {
/** Org org slug (matches the JWT `owner` claim and the IAM `<org>-<app>` namespace). */
readonly orgId: string
/**
* OPTIONAL org-resolution anchor for PASSWORD LOGIN only. Unset (the default)
@@ -16,7 +16,7 @@ export interface TenantConfig {
* their own org). Pinning `orgId` here would resolve a colliding brand-org row
* and truncate a global admin to a single org so the portal leaves this
* unset. Set it ONLY for a brand that deliberately scopes its portal login to
* one tenant. Does NOT affect signup (which always targets `orgId`) or the
* one org. Does NOT affect signup (which always targets `orgId`) or the
* apps launcher (which is brand-scoped by `orgId`).
*/
readonly loginOrg?: string
@@ -45,29 +45,22 @@ export interface TenantConfig {
/** Optional absolute URL to brand.json (e.g. a jsDelivr-hosted copy from
* config.json). Preferred over the app-local /brand/<pkg>/brand.json. */
readonly brandUrl?: string
/**
* Pay origin for this brand the top-up / plan-purchase surface and its
* billing catalog (`GET <payUrl>/v1/billing/plans`). Onboarding's plan step
* reads the catalog from here and hands off to it after the choice. Defaults
* to `https://pay.hanzo.ai`; a white-label brand sets its own in the runtime
* catalog. NO trailing slash. */
readonly payUrl?: string
}
/**
* Brand contract that all per-org brand packages MUST satisfy.
* Matches the consumer contract in `@hanzo/brand` / `@luxfi/brand` /
* `@zooai/brand` / `@parsdao/brand`. Read from each pkg's `brand.json`.
* Brand contract that all per-org brand packages MUST satisfy the shape the
* portal reads from each pkg's `brand.json` (`@hanzo/brand` / `@luxfi/brand` /
* `@zooai/brand` / `@parsdao/brand`).
*
* DRY: this type is NO LONGER defined here. `@hanzo/brand` is the canonical
* home (`toBrandContract` projects the registry onto exactly this shape); we
* re-export it so there is one contract, not two that can drift.
*/
export interface BrandContract {
/** Org display name shown in headings ("Hanzo", "Lux", "Zoo", "Pars"). */
readonly name: string
/** Browser tab title prefix. */
readonly title: string
/** Short tagline rendered on the portal hero. */
readonly description: string
/** Marketing site (footer link target). */
readonly appDomain: string
/** Logo + favicon URLs (CDN or data URI). */
readonly logoUrl: string
readonly faviconUrl: string
/** Primary accent (CSS color string, e.g. "#ff6b35" or "var(--brand)"). */
readonly accentColor?: string
/** Optional social links rendered in the footer. */
readonly twitter?: string
readonly github?: string
readonly discord?: string
}
export type { BrandContract } from '@hanzo/brand/registry'
+820 -6165
View File
File diff suppressed because it is too large Load Diff
+12
View File
@@ -0,0 +1,12 @@
import { defineConfig } from 'vitest/config'
// Single test runner for the whole monorepo. Every `*.test.ts` under any
// package's `src/` runs here — the connect crypto/connector suites (vitest
// describe/it/expect) and the auth / shared / onboarding suites (bare
// `test()` + node:assert). One config, one `pnpm test`, one way.
export default defineConfig({
test: {
include: ['pkgs/**/src/**/*.test.ts', 'apps/**/src/**/*.test.ts'],
environment: 'node',
},
})