Local Development¶

Setup¶

git clone <repo>
cd inference-engine
uv sync          # or: pip install -e ".[dev]"
cp .env.example .env

Running¶

# Minimal (SQLite + in-process async)
uvicorn app.adapters.http.app:app --reload

# Full stack (Postgres + Redis + arq worker)
bash dev.sh

dev.sh starts Docker services, waits for Postgres to be ready, runs the DB migration, launches the arq worker, and starts uvicorn.

Note

dev.sh now runs everything inside Docker Compose. No host-side venv path configuration is needed.

Tests¶

pytest                        # all tests + coverage report
pytest tests/test_phase1.py   # specific file

Tests use httpx.TestClient — no running server needed. SQLite uses :memory: for isolation. Coverage requires ≥ 70%.

Curl tests¶

bash tests/curl_test.sh http://localhost:8000

Runs end-to-end checks against a live server and writes results to tests/curl_results.md.

Useful commands¶

BASE=http://localhost:8000

curl $BASE/health
curl $BASE/ready
curl -H "X-API-Key: dev-key" $BASE/models

# Sync inference
curl -X POST $BASE/predict \
  -H "X-API-Key: dev-key" -H "Content-Type: application/json" \
  -d '{"model":"echo","version":"v1","data":"hello"}'

# Async inference
curl -X POST $BASE/predict/async \
  -H "X-API-Key: dev-key" -H "Content-Type: application/json" \
  -d '{"model":"echo","version":"v1","data":"hello"}'

curl $BASE/metrics
curl -H "X-API-Key: admin-key" $BASE/debug/models/loaded
curl -H "X-API-Key: admin-key" $BASE/admin/models/memory