Skip to content

Contributing

New transformers task? It already works - use_transformers reads the taxonomy at runtime. Found a model that needs special handling, or want a new example? PRs welcome.

Ground rules

  1. Real, runnable examples. Add to examples/ and drive actual inference - no mocks. Show the verified output.
  2. Keep smoke.py green. It's the fast E2E gate: 
    PYTHONPATH=. python examples/smoke.py     # → "18/18 checks passed"
  3. Verify claims. Any number/output in docs must come from a real run.

Dev setup

uv pip install -e ".[all]"

pip install -e ".[all]"

Building the docs

uv pip install -e ".[docs]"
mkdocs serve        # live preview at http://127.0.0.1:8000
mkdocs build        # static site → site/