Maintain the documentation

User-facing documentation must agree with the public API, runnable examples and statistical meaning of outputs. Read repository contribution guidance and the standards index before changing it.

Keep the documentation index organised around Python and YAML workflows. Use ordinary Markdown; no generated API site or notebook gallery is required. Link to public interfaces and explain units, prerequisites, unsupported operations and interpretation limits where they affect a user’s decision.

Use local validation for links and examples. Numerical parity, causal validity and production capacity need separate evidence; do not infer them from a formatting check or a short example run.

Find implementation ownership

Read architecture and contracts before extending model construction or runner stages. The typing and performance note records its own evidence limits. The empty docs-site/ directory is unused; maintained pages remain under docs/source/ and require no documentation build.