Docs hub

The morel documentation is split by audience. The homepage links to each of these in one click; this hub is the long-form version, with full reading paths and cross-references.

Sections at a glance

• Getting started — install, demo, smoke test, the minimal example.
• Installation — pinned lockfile, Docker, CUDA, dev extras.
• Method — notation, the seven stages, and the per-stage equations.
• Architecture — module layering, data flow, extension points, the machine-checked layering test.
• Experiments — reproducing a run end-to-end from its manifest and config.
• API reference — every public symbol, generated from each module's __all__. Updated by morel render-api.
• Tutorials — three ordered walk-throughs: synthetic data, real Amazon data, serve over HTTP.
• Contributing — dev setup, code style, commit messages, release process.
• Changelog — release-by-release record of every notable change.

Reading paths

Pick the path that matches your role; each one is a short ordered list.

New user

You just want to see morel run.

1. Getting started — five minutes from pip install to a working end-to-end pass.
2. Tutorial 1 — synthetic — what the demo does, step by step.
3. Tutorial 2 — real Amazon data — drop the synthetic features and run on a real catalogue.

Researcher

You want to understand the method, the implementation, and the fidelity registry before citing the library.

1. Method — notation, stages, equations.
2. Architecture — module layering and data flow.
3. Reference: paper fidelity — which paper equations map to which code paths.
4. Reference: benchmarks and Reference: production readiness — the evidence story.

Contributor

You want to add a stage, swap an implementation, or push a PR.

1. Architecture — the layering rules.
2. How-to: add a custom component — register a new factory and select it from config.
3. How-to: configure morel — every config field, with a plain-English description.
4. Contributing — workflow, commit messages, tests.

Reviewer

You want to audit a claim. The paper-fidelity registry is the canonical link between paper sections and code.

1. Reference: paper fidelity — every paper component bound to its implementation and test.
2. Operations: reproduce — re-run a saved experiment deterministically from its manifest.
3. Reference: production readiness — the audit trail across the codebase.

How the docs are organized

Home (index.md)
└── Docs hub (this page)
    ├── Getting started
    ├── Tutorials
    │   ├── Tutorial 1 — synthetic
    │   ├── Tutorial 2 — real Amazon data
    │   └── Tutorial 3 — serve over HTTP
    ├── How-to
    │   ├── Configure morel
    │   ├── Add a custom component
    │   └── Serve in production
    ├── Concepts
    │   ├── Method
    │   └── Architecture
    ├── Reference
    │   ├── API (auto-generated)
    │   ├── CLI
    │   ├── Config
    │   ├── Paper fidelity
    │   ├── Benchmarks
    │   ├── Production readiness
    │   └── Limitations
    ├── Operations
    │   ├── Install
    │   ├── Reproduce
    │   ├── Deploy
    │   ├── Observability
    │   └── GitHub Pages
    └── Project
        ├── Changelog
        ├── Security
        ├── Contributing
        └── License

Conventions used throughout the docs

• Code blocks carry a language tag (python, bash, yaml, …). Plain prose is the only thing without a tag.
• Tables carry parallel information in a fixed column order — never rearrange columns between rows of the same table.
• Admonitions are used sparingly:
  • !!! tip for things that help you decide.
  • !!! warning for things that break your workflow if ignored.
  • !!! danger for things that are security-critical.
• "See also" links at the bottom of every page point at the next page in the recommended reading order for each role.

Writing this documentation

If you spot something missing or wrong, open a GitHub issue or follow the Contributing guide and send a pull request. The Astro build renders the internal links on every build, so a broken link fails CI before merge.