[pytest] markers = unit: Fast tests with mocked dependencies - run in CI and git hooks integration: Tests requiring complex setup/external services - run in git hooks only e2e: End-to-end workflow tests with real configurations - run in git hooks only eval: Quality evaluations testing LLM response quality - run manually only performance: Timing harness against a live Ollama - run manually only (needs Ollama reachable) testpaths = tests # Evals are excluded by default, run them explicitly with: pytest evals/ -v # Performance tests are excluded by default, run them explicitly with: pytest tests/performance/ -v -m performance addopts = -m "not performance"