Add `pre-commit` hook to sanitize Jupyter notebooks (strip noisy metadata / execution counts, keep cell source + outputs)

[conorheins]

Notebook diffs in pymdp often include noisy diffs from Jupyter metadata (e.g., metadata.kernelspec, metadata.language_info) and execution-count churn. Let’s add a pre-commit hook that automatically sanitizes notebooks before commit while still keeping cell source/markdown (and outputs) intact.

Proposed changes

• Add .pre-commit-config.yaml with nbstripout:

  repos:
    - repo: https://github.com/kynan/nbstripout
      rev: 0.8.2
      hooks:
        - id: nbstripout
          files: \.ipynb$
          args:
            - --keep-output
            - "--extra-keys=metadata.kernelspec metadata.language_info"

• Add pre-commit to a dev dependency group (e.g. [dependency-groups].test or a new [dependency-groups].dev) since we already ship nbstripout in pyproject.toml.

• Add a short note to CONTRIBUTING.md / README:

  • uv sync --group test (or --group dev)
  • uv run pre-commit install
  • (optional) uv run pre-commit run --all-files

Why this approach

• Hook runs locally and reduces notebook diff noise without changing our CI workflows.
• nbstripout supports running as a pre-commit hook and pre-commit uses .pre-commit-config.yaml at repo root. (link to nbstripout)