Python Testing
Use this when adding or updating DeepSearch tests.
Workflow
- Locate nearby tests for the touched source.
- Add the smallest test that captures the behavior.
- Mock LLM/search/storage/network boundaries by default.
- Run targeted pytest before broad tests.
Commands
uv run pytest tests/path/to/test_file.py
uv run pytest -m "not llm"
RUN_LLM_TESTS=1 uv run pytest -m llm
Run live tests only when explicitly requested and credentials are configured.
Fixtures
- Prefer the nearest
conftest.py. - Use
tmp_pathfor generated reports, markdown/html/docx files, logs, and databases. - Use
monkeypatchfor environment variables and runtime config. - Use
AsyncMockfor async LLM/search calls.
Async Tests
Pytest uses strict asyncio mode. Mark async tests explicitly and ensure background tasks are awaited or cancelled.
import pytest
from unittest.mock import AsyncMock
@pytest.mark.asyncio
async def test_node_uses_llm(monkeypatch):
fake_call = AsyncMock(return_value={"result": "ok"})
monkeypatch.setattr("module.path.call_llm", fake_call)
result = await run_node()
assert result["result"] == "ok"
fake_call.assert_awaited_once()
Live-Test Boundary
- Mark real LLM/search tests with
@pytest.mark.llm. - Skip unless
RUN_LLM_TESTS=1. - Never make live credentials required for normal
uv run pytest.
