Advanced RAG Chatbot

An advanced Retrieval-Augmented Generation (RAG) system built with FastAPI, LangGraph, and Next.js. This application allows users to upload documents and chat with them using a sophisticated graph-based orchestration engine that ensures high-quality answers through grading, hallucination checks, and query rewriting.

🚀 Features

• Graph-Based RAG: Uses LangGraph to orchestrate a corrective RAG flow:
  • Retrieval: Fetches relevant context from Qdrant.
  • Grading: Evaluates document relevance before generation.
  • Generation: Produces answers using OpenAI LLMs.
  • Hallucination Check: Verifies if the answer is grounded in documents.
  • Answer Grading: Checks if the answer actually resolves the user's question.
  • Query Rewriting: Optimizes queries if initial retrieval is poor.
• Dual-LLM Strategy: Uses a faster model for routing/grading and a smarter model for generation.
• Document Ingestion: Supports uploading PDF, TXT, and MD files with automatic chunking and embedding.
• Secure Authentication: JWT-based authentication with role-based access control (User vs. Admin).
• Modern Frontend: Built with Next.js 16, TypeScript, Tailwind CSS, and Shadcn/UI.

🛠️ Tech Stack

Backend

• Framework: FastAPI
• Orchestration: LangGraph, LangChain
• LLM: OpenAI (GPT-3.5/4o)
• Vector Store: Qdrant
• Database: PostgreSQL (for user data)
• Dependency Manager: Poetry

Frontend

• Framework: Next.js 16 (App Router)
• UI Library: Shadcn/UI, Tailwind CSS
• State/Fetching: React Server Components, Axios
• Form Handling: React Hook Form (implied)

🏗️ Architecture

graph TD
    Start([Start]) --> Retrieve
    Retrieve --> GradeDocuments
    GradeDocuments -->|Relevance High| Generate
    GradeDocuments -->|Relevance Low| RewriteQuery
    RewriteQuery --> Retrieve
    Generate --> HallucinationCheck
    HallucinationCheck -->|Grounded| AnswerCheck
    HallucinationCheck -->|Hallucination| Generate
    AnswerCheck -->|Useful| End([End])
    AnswerCheck -->|Not Useful| RewriteQuery

Loading

🚀 Getting Started

Prerequisites

• Python 3.11+
• Node.js 18+
• Docker (optional, for running Qdrant locally)
• OpenAI API Key
• Qdrant Cloud Cluster or Local Instance

Backend Setup

1. Navigate to the backend directory:

   cd backend

2. Install dependencies using Poetry:

   poetry install

3. Configure Environment Variables: Create a .env file in the backend directory:

   OPENAI_API_KEY=sk-...
   QDRANT_URL=your-qdrant-url
   QDRANT_API_KEY=your-qdrant-key
   QDRANT_COLLECTION_NAME=rag_collection
   LLM_MODEL_FAST=gpt-3.5-turbo
   LLM_MODEL_SMART=gpt-4o
   SECRET_KEY=your_jwt_secret_key
   ALGORITHM=HS256
   ACCESS_TOKEN_EXPIRE_MINUTES=30
   DATABASE_URL=postgresql://user:password@localhost/dbname

4. Run the Server:

   poetry run uvicorn app.server:app --reload

Frontend Setup

1. Navigate to the frontend directory:

   cd frontend

2. Install dependencies:

   npm install

3. Run the Development Server:

   npm run dev

4. Access the App: Open http://localhost:3000 in your browser.

📝 API Endpoints

• GET /: Health check.
• POST /auth/token: Login and get JWT token.
• POST /chat: Send a message and get a RAG-based response (Protected).
• POST /ingest: Ingest raw text (Admin only).
• POST /ingest/file: Upload and ingest a file (Admin only).

📄 License

This project is licensed under the Apache License 2.0.

---

Created by EternalKnight002 — Built to learn, optimized for scale.