Testing

Running Unit Tests

# Run all tests
uv run pytest

# Run a specific test file
uv run pytest tests/py_tests/test_utils.py

# Exclude slow tests
uv run pytest -m "not slow"

# Run with randomized order (enabled by pytest-randomly)
uv run pytest -p randomly

Test Markers

Marker	Description
@pytest.mark.slow	Long-running tests, skipped with -m "not slow"

Type Checking

uv run mypy

Configured in pyproject.toml with strict mode and the pydantic plugin.

Linting and Formatting

# Lint
uv run ruff check bsllmner2/ tests/ scripts/

# Format
uv run ruff format bsllmner2/ tests/ scripts/

# Format check (for CI)
uv run ruff format --check bsllmner2/ tests/ scripts/

Mutation Testing

mutmut validates that tests can detect code mutations.

# Run mutation testing
uv run mutmut run

# Show results
uv run mutmut results

Target modules are configured in pyproject.toml:

[tool.mutmut]
paths_to_mutate = "bsllmner2/"
tests_dir = "tests/py_tests/"

Model Evaluation

The tests/model-evaluation/ directory contains scripts for benchmarking LLM models on ontology mapping accuracy.

Datasets (hosted on Zenodo):

• https://zenodo.org/records/14881142
• https://zenodo.org/records/14643285

600 BioSample entries evaluated against a human-curated gold standard. Data files are stored in tests/data/ (see tests/data/README.md for details).

Evaluated models: deepseek-r1 (8b/32b), gemma3 (4b/12b/27b), gpt-oss (20b), llama3.1 (8b), phi4 (14b), qwen3 (4b/8b/32b)

Metrics: Precision, Recall, F1-score, Accuracy (for the cell_line field)

For the full evaluation workflow (batch execution, metric computation, result aggregation), see tests/model-evaluation/README.md.