sqlmodel-ai-service-production-kit

Local-first SQLModel kit for AI services that need tenant-safe records, prompt/run storage, migration fixtures, and evaluation reports before using a production database.

The quickstart uses SQLite fixture data only. It does not call model providers or require a database server.

What It Includes

• SQLModel tables for tenants, projects, datasets, model runs, and evaluations
• fixture data for two tenants
• tenant-safe repository functions
• migration plan fixture with rollback notes
• evaluation fixture storage and summary report
• FastAPI endpoints, tests, CI, local demo, and production docs

Quickstart

python -m venv .venv
. .venv/bin/activate
python -m pip install -e ".[dev]"
pytest -q
python examples/local-only-demo/demo_client.py

Run the API locally:

uvicorn prodkit_sqlmodel_ai.app:app --reload

Try fixture endpoints:

curl http://127.0.0.1:8000/health
curl http://127.0.0.1:8000/tenants
curl http://127.0.0.1:8000/tenants/tenant-acme/datasets
curl http://127.0.0.1:8000/tenants/tenant-acme/runs
curl http://127.0.0.1:8000/migration-plan
curl http://127.0.0.1:8000/evaluation-report

Local Boundaries

• Fixture prompts are synthetic.
• Tenant-owned reads require a tenant id.
• No provider tokens or external database URLs are needed.
• Migration output is a planning fixture, not an automatic database migration.

Production Docs

Start with production/README.md.

License

MIT