Grounded Knowledge Assistant (Governed RAG System)

A governed Retrieval-Augmented Generation (RAG) system designed for regulated industries. This solution delivers traceable, source-grounded answers with built-in compliance checks, hallucination mitigation, and auditability.

🚀 Key Features

• Source-grounded answers with citations
• Governance and compliance layer
• Trust score + explainability
• Audit trail logging

🏗️ Architecture

See /docs/architecture.md

📊 Evaluation

See /evaluation/

🛡️ Governance

See /governance/

👥 Team

• Portia — Product, Architecture, Governance
• Nadia — Evaluation, Metrics
• Megan — Backend, Security
• Esthefany — Data, RAG Pipeline
• Neha — Governance, Risk

---

▶️ Run Locally

1) Configure environment variables

Copy .env.example (if present) or create .env with:

• PORT=3000
• AZURE_OPENAI_ENDPOINT=https://<your-azure-openai>.openai.azure.com/
• AZURE_OPENAI_KEY=<your-azure-openai-key>
• AZURE_OPENAI_DEPLOYMENT=<your-deployment-name>
• AZURE_SEARCH_ENDPOINT=https://<your-search-service>.search.windows.net
• AZURE_SEARCH_KEY=<your-search-key>
• AZURE_SEARCH_INDEX=<your-index-name>

If OPENAI_API_KEY is not set, this code uses Azure OpenAI when available and otherwise falls back to a local mock response for faster local development.

2) Start backend

cd /Users/testtaker/governed-rag-backend
npm install
npm start

Verify backend:

curl http://localhost:3000/health

3) Start frontend

cd /Users/testtaker/governed-rag-backend/frontend
npm install
npm run dev

Open browser at URL printed by Vite (usually http://localhost:5173).

4) API endpoints

• GET /health - health check
• POST /query - search tools (body: { "query": "..." })
• POST /govern - policy validation (body: { "type":"query","content":"..."})
• POST /rag - full RAG+governance pipeline (body: { "query": "...", "user_role": "user" })
• GET /rag?query=...&user_role=... - same pipeline

5) Troubleshooting

• If 403/404 from Azure Search/OpenAI, fallback is used and you still get static results.
• Set OPENAI_API_KEY to use public OpenAI APIs instead of Azure.
• Confirm .env values are correctly loaded and re-start backend after edits.

---

🔗 Broader Initiative

🔗 This project is part of a broader initiative under AI for Regulated Enterprises, focused on building compliant AI systems for real-world use cases.