API documentation
Generated rustdoc for every crate in the workspace.
- auracle_grammar — the genome: typed PCFG, trace codec, compiler, structural edits, presets
- auracle_features — render, vet, LUFS-normalize, extract
- auracle_taste — the utility model, three likelihoods, MCMC posterior, standardization
- auracle_session — the two-loop engine, acquisition, calibration, persistence
- auracle_wasm —
WasmEngineandLivePoly
Built with cargo doc --workspace --no-deps, so the dependency crates are not
included. quiver, fugue-ppl
and fugue-evo have their own docs on docs.rs.
Where to start
The doc comments in this codebase carry a lot of the reasoning, and a few are worth reading directly rather than through this book's summary of them:
| For | Read |
|---|---|
| The grammar's site table | auracle_grammar::prior module docs |
| Why has families rather than per-module columns | auracle_features::structural module docs |
| The max-of-experts argument, and the correction | auracle_taste::model module docs |
| Why the standardizer's threshold is | auracle_taste::standardize::RUNAWAY_RATIO |
| Why refinement is 40 steps × 10 seeds | auracle_session::SessionConfig::refine_steps |
| The acquisition measurement, in full | auracle_session::Acquisition |
| Why accuracy was replaced by Brier skill | auracle_session::calib module docs |
Building it locally
cargo doc --workspace --no-deps --open
# or, the target this site uses:
make site-api
What rustdoc covers
The generated docs are authoritative about signatures and invariants, and they are where the numbers live. They are deliberately quiet about the pipeline: no rustdoc page explains why vetting has to run before normalization, because that fact belongs to no single item.
That is the division of labour between this book and the API docs: the book owns the reasoning that spans crates, and rustdoc owns the reasoning that fits beside a definition.