Skip to content

v0.4.0: LangChain Integration + First PyPI Packages

Choose a tag to compare

View all tags
@jigangz jigangz released this 21 Apr 07:40
· 26 commits to main since this release
v0.4.0-langchain
a5dcae1

🎯 3 PyPI Packages Now Live

pip install trustrag-langchain   # LangChain + LangGraph integration
pip install trustrag-mcp         # MCP server for Claude/Cursor
pip install trustrag-eval        # RAGAS-based evaluation pipeline

✨ trustrag-langchain Highlights

L1 — TrustRAGRetriever: LangChain BaseRetriever with built-in trust filtering. Answers below min_trust_score return empty list (keeps agent context clean of untrusted data).

L2 — TrustRAGTool: BaseTool wrapper for any LangChain agent. Returns formatted [Trust: X/100] answers.

L3 — TrustBudgetAgent: LangGraph multi-hop agent with cumulative trust budget. Three outcomes:

  • answer — cumulative trust ≥ threshold (default 150)
  • stop_low_trust — max_retrievals reached but evidence insufficient
  • error — retriever exception

📦 Ecosystem

  • 3 PyPI packages (langchain / mcp / eval)
  • 3 n8n workflow templates (doc ingestion, Slack trust-gated Q&A, daily digest)
  • GitHub Actions auto-publish on pkg-<name>-v* tags
  • 77 tests passing (backend + frontend + 3 packages + eval)

🙏 Notes

  • trustrag-mcp Claude Desktop integration test pending (P5-GATE)
  • Production benchmark pending Groq quota refresh (P3-GATE)
  • Deployment to Railway + Vercel pending (P7-GATE / v1.0.0)
Assets 2
Loading