10 Commits
Author SHA1 Message Date
zeekay 4a1aaffdf8 fix: importing the package no longer asks setuptools for a namespace
The first statement of __init__.py called pkg_resources.declare_namespace, and
setuptools 83 no longer ships pkg_resources, so importing datastore_orm raises
ModuleNotFoundError before any of its own modules load. Every consumer dies on
`import datastore_orm`; in Insights that is Django app startup, which takes the
image build down with it.

The declaration was for the `infi` namespace this distribution used to live
under — buildout.cfg still names that as the namespace package. The code now
ships one plain top-level package, src/datastore_orm, so there is no namespace
left to declare and the call could only ever fail. Removing it is what makes the
package importable again, not a workaround for the missing module.
2026-08-08 04:48:59 -07:00
zeekayandHanzo Dev 938fc361ed ci: canonical pair — hanzo.yml + forge caller
github.com has no runner for hanzo-build-linux-amd64, so a caller under
.github/workflows is a gate that can never be scheduled. This is the ~7-line
caller on the plane that can (git.hanzo.ai git-runner fleet), pinned @v1.

Co-authored-by: Hanzo Dev <[email protected]>
2026-08-06 02:32:04 -07:00
zeekayandHanzo Dev bf5f0c52b1 call the store the datastore
27 docstrings still said ClickHouse. The store is the datastore everywhere else
in the tree — the package, the module, the env vars, the table — so the prose
was the last place a reader met the old name.

Kept three kinds of reference, deliberately:
  - clickhouse.tech/... links — citations to the upstream engine's own docs.
    They explain what an engine argument MEANS; deleting them would remove the
    only pointer to the semantics and make the code harder, not cleaner.
  - "clickhouse-server generated" in the error parser — that string names the
    process that literally emits the message being parsed.
  - the one line in migrations.py recording what the table used to be called,
    which is the point of that comment.

A name we brand is ours to change. A name that identifies someone else's
software is a fact.

Co-authored-by: Hanzo Dev <[email protected]>
2026-07-30 12:28:06 -07:00
zeekayandHanzo Dev 045a476534 log under our own name, everywhere
Two more modules imported getLogger directly and were missed by the first pass.

Co-authored-by: Hanzo Dev <[email protected]>
2026-07-30 12:27:07 -07:00
zeekayandHanzo Dev 46e9dd3e34 log under our own name
The package is datastore_orm; two modules asked logging for 'clickhouse_orm',
so every line this library emitted was filed under a name that no longer exists
in the tree. Left the clickhouse.tech doc links and the 'clickhouse-server
generated' error prose alone — those name the actual upstream server and are
still true.

Co-authored-by: Hanzo Dev <[email protected]>
2026-07-30 12:26:39 -07:00
zeekayandHanzo Dev 710d672a52 migrations: one history model, named migrations
Three models — MigrationHistory, MigrationHistoryReplicated,
MigrationHistoryDistributed — held the same three fields for the same purpose,
and the first two returned the SAME table name, so at most one of them could
ever be the table anyone read. The Replicated/Distributed pair existed for a
sharded, ZooKeeper-coordinated cluster; the store is a single node with no
ZooKeeper, so `if replicated:` was a decision with one answer. Deleted both,
and the branch with them.

The table is `migrations`. It was `infi_clickhouse_orm_migrations`, which
restated three things the reader already has: the vendor of the library this is
forked from, an engine we no longer name, and a layer the database qualifier
already supplies. `insights.migrations` says all of it and nothing twice.

`replicated` stays in the two private signatures so callers need not change —
it simply selects nothing now. Dropped the then-unused Distributed import and
the docstring reference to the deleted class.

Co-authored-by: Hanzo Dev <[email protected]>
2026-07-30 10:57:51 -07:00
zeekay dda3979851 rename: the ORM is ours — datastore_orm, not infi.clickhouse_orm
This is our fork, so it should read like it. `import datastore_orm` instead of
`from infi.clickhouse_orm import ...`: `infi` was the original vendor's org
namespace, and carrying someone else's org name through every import in the
consuming codebase is not a fork, it is a vendored copy that never got renamed.

The `infi` namespace package is gone entirely (src/infi/__init__.py with it),
the distribution is `datastore-orm`, and the metadata points at Hanzo rather
than at the upstream repo and author.

NOT renamed, deliberately: the migration history tables are still
`infi_clickhouse_orm_migrations`. That name is persisted state — 448 rows in
production key off it — so changing it is a data migration, not a rename. Doing
it carelessly reproduces the exact failure this fork just hit, where a package
rename orphaned the history and silently blocked every migration.
2026-07-26 08:44:11 -07:00
zeekay 8a4cc4c025 docs(brand): add hero banner 2026-06-28 20:23:43 -07:00
zeekay bd95395ad8 chore(brand): dynamic hero banner 2026-06-28 20:23:42 -07:00
zeekayandGitHub 0b8b1ad8c9 Merge branch 'PostHog:develop' into develop 2026-06-03 13:04:10 -07:00