abiramavarshini/rag-backend

RAG Copilot with Authentication

A full-stack AI Copilot application featuring Retrieval-Augmented Generation (RAG) for PDF knowledge bases, complete with user authentication, precision scoring, and a modern React frontend.

Features

• —Document Q&A: Upload PDFs and ask questions about them.
• —RAG Engine: Hybrid search (BM25 + Semantic) for high-accuracy retrieval.
• —Precision Scoring: Displays relevance scores (0-100%) and transparency for every answer.
• —Verification: Refuses to answer if context is insufficient ("I don't have enough information").
• —Authentication: JWT-based Signup and Login flows.
• —Modern UI: Glassmorphism design using React and CSS modules.

Architecture

graph TD
    User[User] -->|HTTPS| FE[React Frontend]
    FE -->|Requests| BE[FastAPI Backend]
    
    subgraph Backend
        BE -->|Auth| DB[(SQLite/Postgres)]
        BE -->|Ask| RAG[RAG Engine]
    end
    
    subgraph RAG Pipeline
        Docs[PDF Documents] -->|Ingest| Chunks[Text Chunks]
        Chunks -->|Embed| Chroma[(ChromaDB)]
        Chunks -->|Tokenize| BM25[BM25 Index]
        Chroma & BM25 -->|Retrieve| Hyb[Hybrid Search]
        Hyb -->|Context| LLM["LLM (Llama/OpenAI)"]
    end

Quick Start / Quality Gate

Run the automated quality gate to verify the system:

Windows (PowerShell):

./scripts/test_all.bat

Linux/Mac:

make test

Manual Setup

Backend:

   cd backend
   pip install -r requirements.txt
   python -m uvicorn app.main:app --reload

Frontend:

   cd frontend
   npm install
   npm run dev

Live Deployment

🚀 Production App (Single Link)

• —Live URL: https://huggingface.co/spaces/abiramavarshini/rag-backend
• —Direct App Link: https://abiramavarshini-rag-backend.hf.space
• —Status: Running 🟢

The entire application (Frontend + Backend + RAG) is hosted here. Anyone with this link can sign up, log in, and use the RAG copilot immediately.

API Documentation

Auth

• —POST /auth/signup: Register a new user.
• —POST /auth/login: Get access token.

RAG

• —POST /rag/ingest: Trigger PDF ingestion.
• —POST /rag/ask: Ask a question.

  {
    "query": "What is the policy on X?"
  }

Ingestion & Evaluation

• —Ingestion: PDFs in backend/data/ are read, split into 150-word chunks with 30-word overlap, and indexed.
• —Evaluation: We use a hybrid retrieval approach (Vector + Keyword) and verify answers by checking if retrieval scores exceed a 0.60 threshold.