mltrail-verifiable-provenan.../README.md

29 lines
1.6 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# MLTrail Verifiable Provenance Ledger (MVP)
This repository contains a minimal, productionoriented MVP for a verifiable provenance ledger
tailored to federated machine learning experiments. It offers a compact, append-only hash-chain ledger
with a contract-driven schema registry and a delta-sync protocol to keep distributed networks eventually consistent.
Key components (MVP):
- Data contracts and schemas for Experiment, Run, Dataset, Model, Environment, EvaluationMetric, Policy.
- Append-only ledger with hash-chain, genesis block, and delta-sync capability.
- Lightweight contract registry (GoC bridge) concepts to enable canonical primitives and replay protection.
- Adapters (MLFlow-like, WandB-like) that publish to the ledger.
- Reproducibility groundwork via environment hashing and deterministic record layout.
- Governance and security scaffolding via signatures (plausible placeholder for MVP).
How to use (quick start):
- Run tests and build: ./test.sh
- Publish a registry entry: use the ledger to add a ContractRegistryEntry payload.
- Explore delta-sync: create two ledgers, publish on one, and retrieve deltas on the other with delta_sync.
Note: This is an MVP and intentionally lightweight. It focuses on correct data flow and verifiable append-only logs
while keeping the surface area small for rapid iteration and extension (adapters, delta-sync, and governance).
For contributors: see mltrail_verifiable_provenance_ledger_for/goC_bridge.py for structure of the canonical bridge and
mltrail_verifiable_provenance_ledger_for/ledger.py for the delta-sync primitives.
Version: 0.1.0
"README.md" hooks into packaging via pyproject.toml readme field.