Apache Sourcelume is open-source instrumentation for AI training-data provenance — a metadata specification, a reference registry, and tooling that let dataset curators and model producers publish signed, independently verifiable records of where their data came from and what terms it carries.

Clone this repo:
  1. 691d26c CI: add Python validation job (uv, validate.py, unit tests) (#7) by NaveenKumar Namachivayam ⚡ · 5 days ago main
  2. b9ad844 SETUP.md: document mvn install for downstream Registry builds (#6) by NaveenKumar Namachivayam ⚡ · 5 days ago
  3. 7bd01b9 Update custodyChain usage information by jgoodyear · 6 days ago
  4. db111bf Update changelog and test_validate by jgoodyear · 6 days ago
  5. da90d59 Add 0.0.1 ProvenanceRecord spec (common, model-agnostic parts) (#5) by Achim Nierbeck · 6 days ago

Apache Sourcelume — Spec (sourcelume-spec)

Overview

This repository contains our Apache Sourcelume specification.

ssourcelume-spec/
├── LICENSE                     # Apache 2.0 (code/tooling)
├── README.md
├── NOTICE                      # required for ASF releases
├── CHANGELOG.md
├── context/
│   ├── 0.0.1/
│   │   └── sourcelume.jsonld    # the JSON-LD @context document, pinned
│   └── latest -> 0.0.1/        # symlink or redirect, NOT a moving target for actual releases
├── schema/
│   ├── 0.0.1/
│   │   ├── sourcelume.schema.json   # JSON Schema for structural validation
│   │   └── sourcelume.shacl.ttl     # SHACL shapes for RDF-level validation (optional but common for JSON-LD vocabs)
│   └── ...
├── spec/                         # the prose specification
│   ├── 0.0.1/
│   │   └── index.md (or .adoc)   # human-readable spec text, versioned alongside schema
├── examples/
│   ├── minimal-record.jsonld
│   ├── full-provenance-card.jsonld
│   └── crosswalk/                # worked examples showing mapping to Croissant / SPDX / OTDI fields
├── mappings/
│   ├── croissant-crosswalk.md
│   ├── spdx-ai-profile-crosswalk.md
│   └── otdi-crosswalk.md
├── tools/
│   ├── validate.py               # or JS — validates examples/ against schema/shacl on every PR
│   └── render-docs.py            # generates human-readable HTML/MD from schema
├── .github/workflows/
│   └── validate.yml              # CI: lint JSON-LD, run schema+SHACL validation, check examples
└── .gitignore

Versioning

This project follows Semantic Versioning. See VERSIONING.md for how MAJOR, MINOR, and PATCH apply to specification changes, and for the rules around the latest pointer and immutable released versions.

Get involved

License

This project is licensed under Apache License, Version 2.0.