src/db/schema.sql is not included in the built package — pipx installs cannot bootstrap the base schema

[bgimbel]

src/db/schema.sql is not included in the built package — pipx installs cannot bootstrap the base schema

Summary

The pip/pipx-installed package ships the incremental migrations (src/db/migrations/*.sql) but not the base schema file src/db/schema.sql. Any install path that doesn't go through the Docker stack has no way to create the core tables (memory, etc.) from the installed package alone.

Environment

• open-brain installed via pipx install 'git+https://github.com/benclawbot/open-brain.git' (v1.0.0/1.0.1, master as of 2026-07-31)
• Python 3.12, Ubuntu (WSL2), external PostgreSQL 17 + pgvector (no Docker)

Reproduction

$ pipx install 'git+https://github.com/benclawbot/open-brain.git'
$ ls ~/.local/share/pipx/venvs/openbrain/lib/python3.12/site-packages/src/db/
migrations/  ...          # <-- no schema.sql
$ ls .../src/db/migrations/
002_continuity_foundation.sql ... 015_embedding_dim_change.sql   # 002-015 only, no 001/base

With no base schema applied, openbrain store fails:

Error: relation "memory" does not exist

(or, if a partial/hand-made memory table exists, migrations 014/015 ALTER it and mask the problem until an INSERT fails with column "source" of relation "memory" does not exist.)

Root cause

pyproject.toml package-data only globs the migrations directory:

[tool.setuptools.package-data]
config = ["*.yaml", "*.yml"]
"src.db" = ["migrations/*.sql"]

src/db/schema.sql matches neither pattern, so it's excluded from the wheel. The migration chain also starts at 002_ — there is no 001_ migration that creates the base tables, so the migrations alone can never bootstrap an empty database; they assume schema.sql was applied first (which only the Docker startup path does).

Suggested fix

Either (a) include the schema in package-data and have openbrain configure apply it before migrations:

"src.db" = ["schema.sql", "migrations/*.sql"]

or (b) convert schema.sql into a proper 001_base_schema.sql migration so the existing migration runner bootstraps empty databases with no special case.

Workaround

Download src/db/schema.sql from the repo and apply it manually with psql -f before using the CLI.