Testing

Unit tests cover the pure-function logic: pipeline resolver matching, verifier confidence-combination and parallel join strategies, step-library use: deep-merge, and webhook dedup fingerprinting/settings.

# from repo root — installs requirements.txt plus pytest/pytest-asyncio
pip install -r requirements-dev.txt

cd service
pytest

By default the suite runs against SQLite — a fresh per-test temp-file DB, no setup required. To run the same tests against Postgres instead (exercising the Postgres-only migration branch in create_tables()), point VECTORSTEP_TEST_DATABASE_URL at a throwaway database:

createdb vectorstep_test
VECTORSTEP_TEST_DATABASE_URL=postgresql+asyncpg://localhost:5432/vectorstep_test pytest

Isolation on Postgres is via a DROP SCHEMA public CASCADE; CREATE SCHEMA public; reset around each test (see tests/conftest.py’s db fixture) rather than a fresh file, since there’s no per-test temp file on a shared server. CI (.github/workflows/tests.yml) runs both backends on every push.