ResearchGPT - Llama Research Assistant

A fully local, AI-powered research assistant that lets you upload PDFs and ask questions using retrieval-augmented generation (RAG). Powered by Ollama, ChromaDB, and Llama 3.2.

Features

• 📄 PDF Upload & Ingestion - Upload PDFs and automatically extract and index content
• 🤖 AI-Powered Q&A - Ask natural language questions about your documents
• 🔍 Source Attribution - Get page numbers and referenced chunks for every answer
• 📊 Real-time Status Monitoring - Check Ollama and ChromaDB health in real-time
• 🏠 Fully Local - No cloud dependencies, complete privacy
• ⚡ Fast & Responsive - Built with React, Vite, and FastAPI

Tech Stack

Frontend:

• React with TypeScript
• Vite (build tool)
• TailwindCSS (styling)
• Radix UI components

Backend:

• FastAPI (Python)
• ChromaDB (vector database)
• Ollama (LLM & embeddings)
• PyMuPDF (PDF processing)

Prerequisites

Before you begin, ensure you have:

1. Node.js (v16+) - Download
2. Python (v3.8+) - Download
3. Ollama - Download
   • Required models (pull these first):

     ollama pull nomic-embed-text
     ollama pull llama3.2

Installation

1. Clone or Navigate to Project

cd LLM_Research_Assistant

2. Install Backend Dependencies

pip install -r requirements.txt

3. Install Frontend Dependencies

npm install

Setup

1. Start Ollama

Before running the backend, ensure Ollama is running:

# macOS (if not running as daemon):
ollama serve

# Or ensure the Ollama app is running

Pull the required models (if not already done):

ollama pull nomic-embed-text
ollama pull llama3.2

2. Start Backend Server

In a terminal window, run:

python server.py

You should see:

INFO:     Uvicorn running on http://0.0.0.0:8000

3. Start Frontend Development Server

In another terminal window, run:

npm run dev

The frontend will be available at http://localhost:8080

Usage

1. Ingest a PDF:

   • Click "Step 1 — Choose a Book"
   • Select a PDF file from your computer
   • Click "Ingest Book"
   • Wait for the ingestion to complete

2. Ask Questions:

   • Once ingestion is complete, go to "Step 2 — Ask Your Book"
   • Type your question in the input field
   • Press Enter or click "Send"
   • View the AI response with page sources

3. Monitor System:

   • Check the "System Status" section to ensure Ollama and ChromaDB are running
   • All indicators should be green

API Endpoints

• GET /status - Check system health (Ollama, ChromaDB, models)
• POST /ingest - Upload and ingest a PDF file
• GET /collection-info - Get collection metadata (chunk count, pages)
• POST /query - Query the ingested documents

Configuration

You can customize the collection name using environment variables:

export COLLECTION_NAME=my_custom_collection
python server.py

Default collection name: my_book

Troubleshooting

Models not found error

Error: model "nomic-embed-text" not found

Solution: Pull the models first:

ollama pull nomic-embed-text
ollama pull llama3.2

Backend won't start

• Ensure Ollama is running
• Check that port 8000 is available
• Verify all dependencies are installed: pip install -r requirements.txt

Frontend can't connect to backend

• Make sure backend is running on http://localhost:8000
• Clear browser cache and refresh
• Check browser console (F12) for network errors

Ingestion fails

• Ensure the PDF file is valid
• Check that ChromaDB is accessible
• Monitor backend terminal for error messages

Folder Structure

LLM_Research_Assistant/
├── src/
│   ├── components/
│   │   └── rag/           # RAG-specific components
│   ├── lib/
│   │   └── api.ts         # API client
│   └── pages/
│       └── Index.tsx      # Main application page
├── server.py              # FastAPI backend
├── Research_Assistant_RAG.py  # RAG logic
├── requirements.txt       # Python dependencies
└── package.json          # Node.js dependencies

Development

Build for Production

Frontend:

npm run build

Run Tests

npm run test

Performance Notes

• First query may take 10-30 seconds as the LLM initializes
• Subsequent queries are faster as the model stays in memory
• Ingestion time depends on PDF size (typically 5-15 seconds)

License

MIT License - See LICENSE file for details

Support

For issues or questions, please check the troubleshooting section or review the browser console (F12) and backend terminal logs for error messages.

---

Happy researching! 🚀