Running tests

riverbank has three test suites with different requirements.

Unit tests

Fast, no external dependencies. Run in CI on every push.

pytest tests/unit/ -v

Unit tests cover:

• Configuration parsing
• Pipeline logic (mocked extractors)
• Plugin discovery
• Profile parsing
• Fragment hashing
• Editorial policy gates
• Cost calculations
• CLI command parsing

Golden corpus tests

Validate the full pipeline against the example corpus using the noop extractor. No database required.

pytest tests/golden/ -v

Golden tests verify:

• Profile registration
• Corpus ingestion with the noop extractor
• Competency question assertions
• Fragment deduplication

Integration tests

Require a running PostgreSQL instance (via testcontainers). Test real database operations.

pytest tests/integration/ -v

Integration tests cover:

• Alembic migrations
• Graph writes via pg-ripple
• SPARQL queries
• Advisory locks
• Tenant RLS
• Audit logging

Running with testcontainers

Integration tests use testcontainers[postgres] to spin up a temporary PostgreSQL instance:

export TESTCONTAINERS_RYUK_DISABLED=true
pytest tests/integration/ -v

All tests

pytest tests/ -v --tb=short

Coverage

coverage run -m pytest tests/
coverage report --show-missing

Writing tests

• Place unit tests in tests/unit/test_<module>.py
• Use fixtures from tests/conftest.py
• Mock external dependencies (database, LLM) in unit tests
• Integration tests can use real database connections via testcontainers
• Golden tests should be self-contained (no external state)