promptledger-verifiable-provenance-and-l/README.md

promptledger-verifiable-provenance-and-l

This project provides a minimal MVP for a cross-tool, offline-first provenance ledger intended for generative AI creative workflows. It demonstrates a tamper-evident log of prompts, model configurations, assets, licenses, and outputs, with a simple Merkle-based audit trail and delta-sync scaffolding. Adapters map Blender and Figma provenance into a canonical model.

How to run locally

• Prereqs: Python 3.8+, pip, and build tooling (setuptools, wheel)
• Build metadata: pyproject.toml (name: promptledger-verifiable-provenance-and-l)
• Run tests: ./test.sh (requires pytest; should pass without extra dependencies)
• Build package: ./test.sh (will run python3 -m build)

Project structure (核心)

• promptledger_verifiable_provenance_and_l/: Python package containing:
  • LocalProvenanceBlock: a single provenance step with author, tool, action, metadata, and license
  • MerkleAuditLog: a simple, tamper-evident log built on a Merkle tree of blocks
  • DeltaSync: lightweight export/import of provenance deltas for offline-first operation
• BlenderAdapter, FigmaAdapter: sample adapters emitting provenance blocks

Notes

• This is a minimal MVP intended to bootstrap the architecture. Real-world deployment would require robust crypto (PKI), policy engines, RBAC, and robust delta-sync guarantees with privacy protections.

Extensibility

• The MVP now includes a lightweight LicenseContract, SchemaRegistry, and ContractMarketplace to begin modeling a cross-tool governance layer.
• LocalProvenanceBlock supports optional fields (prompt, model_version, seed, parameters, sources, outputs) to capture richer provenance without breaking existing usage.
• Adapters and the ledger can emit and sign blocks; a registry/marketplace can be used to publish and verify reusable contracts and licensing templates.

License: MIT

READY_TO_PUBLISH marker is created when the repo is ready to publish.