Conductor Eval measures how well AI models generate MIDI loops against musical constraints. It runs evaluations through the public conductor-core engine, applies deterministic MIDI checks, saves results, and provides an optional Plotly Dash dashboard for analysis.
The detailed documentation is built with MkDocs:
To preview the documentation locally, install the docs dependency group and start MkDocs:
uv sync --locked --group docs
uv run --locked --group docs mkdocs serveOpen http://127.0.0.1:8000/ in your browser.
Alternatively, Read the documentation source as plain markdown.
For a short local setup, install the locked development environment and run the checks:
uv sync --locked --all-extras
uv run --locked --all-extras pytest -qSee getting started for guidance on running an evaluation.
See analysis to run a dashboard displaying evaluation results.