DocuBot — RAG-based AI Chatbot

Retrieval-Augmented Generation pipeline using React, Node.js, LangChain, Qdrant, and Google Gemini. Enhances prompts, retrieves semantic chunks, and generates grounded answers.

• Backend repo - https://github.com/zanzo2003/DocuBot_backend
• Frontend repo - https://github.com/zanzo2003/DocuBot_frontend

🚀 Key Features

• Advanced RAG pipeline: prompt enhancement → retrieval → grounded generation
• Prompt Enhancer: rewrite & normalize queries to improve retrieval recall
• Semantic retrieval with Qdrant (top-k chunk search)
• PDF ingestion: chunking + embeddings using LangChain loaders
• Modular backend: services, controllers, routes, utils, db, middleware for easy extension

🔍 How it Works

1. Ingest: PDF → chunk → embed → store in Qdrant
2. Enhance: user query rewritten (LLM) to increase retrieval quality
3. Retrieve: semantic search (top-k) returning relevant chunks
4. Answer: LLM generates response grounded on retrieved context

🏗️ Tech Stack

Frontend: React
Backend: Node.js, Express
AI/ML: LangChain, Google Generative AI (Gemini), GoogleGenerativeAIEmbeddings
Vector Database: Qdrant

📁 Project Structure

DocuBot/
├─ frontend/                 
├─ backend/
├─ screenshots/            
└─ README.md

⚙️ Installation

1. Clone the repository

   git clone https://github.com/your-username/DocuBot.git
   cd DocuBot

2. Setup Backend

   cd backend
   npm install

3. Environment Configuration

   Create a .env file in the backend directory:

   PORT=8080
   QDRANT_URL=http://localhost:6333
   QDRANT_COLLECTION=docubot_collection
   GOOGLE_API_KEY=your_google_api_key
   GENAI_BASE_URI=https://generativelanguage.googleapis.com/v1beta/openai
   FRONTEND_URL=http://localhost:3000
   NODE_ENV=development

4. Start Backend

   npm run dev

5. Setup Frontend

   cd ../frontend
   npm install
   npm start

📡 API Usage Examples

Upload File

curl -X POST http://localhost:8080/api/files/upload \
  -F "file=@/path/to/file.pdf"

Chat / RAG Query

curl -X POST http://localhost:8080/api/chat \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Summarize section 2",
    "fileName": "file.pdf"
  }'

Delete Vectors

curl -X DELETE http://localhost:8080/api/files/file.pdf

📂 Screenshots & Demo

Add your screenshots to the ./screenshots/ directory:

• 
• 
• 

🎥 Demo Video

YouTube video link

📌 Roadmap

• Support DOCX / MD ingestion and automatic OCR
• Streaming LLM responses + token-level highlights
• Docker + k8s deployment & infra templates
• Per-user sessions, multi-tenant support, and auth

Suggested branches:

• main (stable)
• dev (work-in-progress)

Built with LangChain • Qdrant • Gemini