The Python client now reaches what the Rust crate does and the server
serves. Research covers every /v1/research route: record() files
benchmarks, studies, runs and papers in one batch a kind; runs(),
get_run() and compare() read them; grant(), revise() and retract() move
consent and versions; the experiment plane and artifacts as before.
Arena covers /v1/benchmark: catalog, leaderboard, claims (written with
protocol and source), history, compare, presets, and runs, which queue a
paid execution and are refused without confirm=True.
wire holds the crate's Run, Benchmark, Measure, Study, Paper and
Comparison field for field; canonical writes RFC 8785 bytes equal to the
crate's (tests/vectors/rust.json, written by the crate); lineage and wal
are the crate's records, and a WAL the crate wrote verifies here.
capture() reads the harness commit (the diff's SHA-256 when dirty), the
weights' SHA-256, host, CPU, GPU, device, dtype and packages, and
apply() folds them into a run.
runs(org=) sends no credential: the server answers ?org= only to a
caller with no org. An org the token does not belong to is refused
before the request, since the server would act in the home org. The
command is hanzo-research; the native CLI owns `hanzo`.
The research client spoke only the experiments wire. The record
hanzo.ai/benchmarks draws from is /v1/research/{benchmarks,runs,studies,
papers,compare}, and producers posted it by hand (Kai's bench/code
post.py). record() posts each non-empty batch under its own key; runs(),
benchmarks(), studies() and papers() read with the API's own filters,
org= included for another org's public runs; compare() reads two runs
side by side. Tested against a stand-in server and read live.
Claude-Session: https://claude.ai/code/session_01M8RoubcPN71KgJuZ8LdBUL
The default surface was 34 tools, 61 KB, ~15,300 tokens of schema in every
prompt of every agent. The mode system already names what a session should
carry, but only the tools register_all_tools enumerates obeyed it: zsh,
curl, playwright and the helpers (ps, npx, uvx, jq, wget, cdp, open) stayed
on in every mode. A mode now decides every discovered tool, the `hanzo`
mode lists what a coding agent calls every few steps (18 tools, 32 KB,
~8,000 tokens), and the rest is a name away: HANZO_MCP_ENABLED_TOOLS=a,b
or =all, `tool enable`, or `mode`.
hanzo-tools-shell 0.6.6: run_shell kept every byte a foreground command
wrote until the command ended. `yes` for thirty seconds is gigabytes. It
keeps the first 16 MiB per stream and drains the rest.
The tests' session-wide patch of asyncio.create_subprocess_exec handed
every shell command a stream that never reached EOF; the first test to run
one read it in a loop until the machine was out of memory (22 GB, killed).
The patch is now opt-in and its process has exited.
bench/latency.py measures the cold start, the tools/list bill and the common
calls through a real MCP client over stdio. Interleaved rounds, wall and CPU
seconds, p50 over 6 cold starts:
cold wall cold CPU tools bytes ~tokens
installed 0.15.18 9.34 s 10.46 s 30 52,601 13,150
main 0.15.20 1.52 s 1.13 s 34 61,287 15,321
this 0.96 s 1.06 s 18 32,222 8,055
The installed environment's 9 s is its hanzoai 8.5.156, whose package init
imports the whole generated cloud client; hanzo-mcp >= 0.15.20 requires
hanzoai >= 8.5.624, which loads it lazily. Per-call latency is spawn-bound
(zsh 17.6 ms, git status 33 ms, fs.read 2.5 ms at p50) and moves with the
machine's load; no call path changed.
Every mode disables a registered tool it does not list, and llm was
listed nowhere but HEAVY_TOOLS, written when it pulled a 100 MB client.
It is an HTTP call to api.hanzo.ai now, so llm and kai_decide join the
essential set every personality carries, and the llm package maps to the
two tools it ships.
hanzo-mcp 0.15.19, on hanzo-tools 0.3.8 and hanzo-tools-llm 0.2.1.
Co-authored-by: Hanzo Dev <[email protected]>
The llm tool called upstream providers through the litellm fork, read
provider keys, and never reached the MCP surface: its register() was a
no-op and it skipped BaseTool's constructor. Kai had no Python tool.
hanzo-tools-llm now carries two HIP-0300 tools over HanzoCloud:
- llm (default action query): POST /v1/chat/completions with model
enso-auto unless named, "auto" for Enso's pick, max_cost and
max_latency_ms as X-Max-Cost and X-Max-Latency-Ms, returning id and the
model that served; models: GET /v1/models, one row per model with the
catalog's per-million prices, narrowed by family; feedback: POST
/v1/ai/feedback with request_id, signal and a 1-3 rating.
- kai_decide (default action decide): POST /v1/decisions with the
TypeScript tool's schema and shape checks.
The consensus, llm_manage and provider modules wrapped the upstream
client and were never registered, so they go, with hanzo-consensus and
the [full] extra. hanzo-tools-llm 0.2.1.
Co-authored-by: Hanzo Dev <[email protected]>
Enso reads its routing bounds from X-Max-Cost and X-Max-Latency-Ms on
/v1/chat/completions, and HanzoCloud could send no header but its own, so
no tool could bound a routed request. post() and the shared request path
now take extra headers that ride beside the bearer.
BaseTool runs `help` for a call that names no action. A tool with one
obvious act (llm's query, kai_decide's decide) now declares
DEFAULT_ACTION, which the call, the advertised schema's default and its
required list all read, so it answers the same no-action call the
TypeScript and Rust runtimes do. Every other tool keeps `help`.
hanzo-tools 0.3.8.
Co-authored-by: Hanzo Dev <[email protected]>
In auto mode any extension error fell through to headless Playwright, so
a slow page in the user's own browser came back as Playwright's missing
executable. A connected browser's failure is now the answer; Playwright
stands in only when no browser is on the router. hanzo-tools-browser
0.5.16.
The release lane regenerates pkg/hanzoai/cloud from the served document
and refuses to publish when the committed client differs; it had drifted
since v8.5.458, so every python-sdk tag, hanzo-mcp and
hanzo-tools-browser included, stopped at that check. The lock now names
v8.5.491 and the hash api.hanzo.ai serves.
Firefox reads its host manifests from ~/.mozilla/native-messaging-hosts
(Mozilla/NativeMessagingHosts on macOS) and names the extension by its
gecko id in allowed_extensions, not an origin. With the extension now
asking for nativeMessaging, Firefox joins the router without a pairing
code, as Chromium does. hanzo-tools-browser 0.5.14.
hanzo-zap-host is the Chrome native messaging host ai.hanzo.zap: it relays
router envelopes between the extension's stdio and the router's unix socket
and stands for the router itself. hanzo-mcp registers it with every installed
Chromium-family browser on start. hanzo-tools-browser 0.5.13.
The served document moved past the one this projection was generated
from, so the client lane's check failed every release. uv.lock follows
hanzo-tools-browser onto zapd.
hanzo-mcp said HELLO as consumer:hanzo-mcp/<pid> to a separate zapd
daemon, which the router that replaced it refuses as bad_id. Every
process that speaks ZAP now embeds the router (zapd 1.1 on PyPI) and the
kernel lock elects one; hanzo-mcp stands for it at startup and takes the
seat mcp/hanzo-<pid>. A browser is a browser/<host>/<name> node, found
by engine or name and reached with Node.call, and the extension's
command codec is unchanged.
hanzo-mcp pair prints the code the extension's popup takes. The daemon
spawn and socket probing are gone, and so is the zap-proto dependency.
The client lane checks the committed client against the document
api.hanzo.ai serves now, not against the ref .spec-lock pins. Cloud
v1.833.257 (2774987c) went live after 8.5.163 was cut and rewrote the
POST /v1/decisions description (cad76e1a0: Jev by OpenRouter's vendor
ids), so the served document moved from 296b1f84 to 783f66a3 and every
push since has failed the check on ai_api.py's docstring. The drift
reproduces identically at e34ab7e1; nothing in this tree caused it.
This is the lane's own move, done by hand: .spec-lock names the served
revision and digest, the generator reruns over those bytes, uv.lock
takes what `uv run` wrote, and the version takes the next patch.
The browser tool types only its core fields; everything else rides in
`args`. navigate's state and console's level sat at the top level,
where the schema drops them, so errors opened the app without waiting
for networkidle and read every console message rather than errors.
The test's stand-in browser now has the real tool's shape.
Claude-Session: https://claude.ai/code/session_01Dc8DXc3fEQCRHUa6XWK4Ku
zeekay
z