Shuddhi

Prove what your model was trained on. Shuddhi is an open-source (Apache-2.0) data factory for LLM training corpora: provenance gating, deduplication, quality scoring, PII redaction and contamination screening — sealed with a reproducible build hash a training run can cite, plus a draft EU AI Act Article 53(1)(d) training-content summary.

docker run --rm ghcr.io/agentanywhere/shuddhi:1.2.0 demo

• Source: github.com/agentanywhere/shuddhi · Apache-2.0
• Image: ghcr.io/agentanywhere/shuddhi

Where to go next

• Quickstart — from nothing to a corpus you can prove the contents of
• User Guide — every stage explained
• CLI Reference — every command and flag
• Running in Docker
• Extending Shuddhi — write a filter plugin
• Engine Internals — how the hashes are defined
• Troubleshooting · FAQ