NotebookLM RAG Web App

A RAG (Retrieval-Augmented Generation) web application similar to Google NotebookLM, built with FastAPI and using Hugging Face Inference API for both embeddings and LLM responses.

Features

• Three-Pane Layout:

  • Left Pane: List of uploaded documents
  • Center Pane: Chat/conversation interface with follow-up questions
  • Right Pane: Document content viewer with citation linking

• Document Upload:

  • Upload PDF, DOCX, and TXT files directly from the browser
  • Real-time upload and indexing progress indicators
  • Automatic text extraction and chunking

• RAG Pipeline:

  • Uses Hugging Face Inference API for embeddings (free API)
  • Uses Hugging Face Inference API for LLM responses (free API)
  • Local vector storage (no external vector DBs like Pinecone, Weaviate, Chroma, FAISS)
  • Semantic search and retrieval
  • Citations linking to source document chunks

Note: Uses Hugging Face's free Inference API for both embeddings and LLM responses.

Architecture

The application follows a clean, modular structure:

• fileParser.py - Handles document parsing (PDF, DOCX, TXT)
• chunker.py - Splits text into manageable chunks
• groqClient.py - All Hugging Face API interactions (embeddings & LLM)
• retriever.py - Vector storage and retrieval
• ragPipeline.py - Coordinates the end-to-end RAG process
• app.py - FastAPI backend with REST endpoints
• static/ - Frontend HTML, CSS, and JavaScript

Installation

1. Install dependencies:

pip install -r requirements.txt

2. Set up Hugging Face API token:

Get a free API token from Hugging Face Settings

export HUGGINGFACE_API_KEY=your_huggingface_token_here

Or create a .env file:

HUGGINGFACE_API_KEY=your_huggingface_token_here

Usage

1. Start the server:

python app.py

Or using uvicorn directly:

uvicorn app:app --host 0.0.0.0 --port 8000 --reload

2. Open in browser:

Navigate to http://localhost:8000

3. Upload documents:

• Click the "+ Upload" button in the left pane
• Select PDF, DOCX, or TXT files
• Wait for upload and indexing to complete

4. Ask questions:

• Type your question in the chat input
• The bot will retrieve relevant chunks and generate an answer
• Click on citations to view the source chunks in the right pane

API Endpoints

• GET / - Serve the main web interface
• POST /api/upload - Upload and process a document
• POST /api/chat - Send a chat message and get response
• GET /api/documents - Get list of all uploaded documents
• GET /api/documents/{document_id} - Get full document content
• GET /api/health - Health check endpoint

Configuration

You can modify the following in ragPipeline.py:

• chunk_size - Size of text chunks (default: 1000 characters)
• chunk_overlap - Overlap between chunks (default: 200 characters)
• storage_path - Path for vector storage

In groqClient.py:

• embedding_model - Hugging Face embedding model name (default: "sentence-transformers/all-MiniLM-L6-v2")
• llm_model - Hugging Face LLM model name (default: "mistralai/Mistral-7B-Instruct-v0.2")

Requirements

• Python 3.8+
• Hugging Face API token (free at https://huggingface.co/settings/tokens)
• See requirements.txt for Python dependencies

Notes

• Documents are stored locally in ./vectorstore/
• Uploaded files are temporarily stored in ./uploads/ during processing
• The vector store persists between sessions
• Hugging Face API has rate limits on free tier - suitable for development/testing

Troubleshooting

Error: HUGGINGFACE_API_KEY not set

• Make sure you've set the API token as an environment variable or in a .env file
• Get a free token at: https://huggingface.co/settings/tokens
• Note: Free tier has rate limits, suitable for development/testing

Upload fails

• Check file format (PDF, DOCX, TXT, MD only)
• Ensure file is not corrupted or password-protected
• Check server logs for detailed error messages

License

This project is provided as-is for educational and development purposes.