pdf-rag-chat

pdf-rag-chat is a Retrieval-Augmented Generation (RAG) application that allows users to upload PDF documents and chat with them using a large language model. The application is built with FastAPI and extract just text from the pdf document.

Features

• PDF Upload: Users can upload PDF documents through a simple web interface.
• Document Processing: The application uses the unstructured library to partition the PDF into text and images. The text is then chunked by title to maintain semantic context.
• Vector Store: The processed chunks are stored in a ChromaDB vector store. The embeddings are generated using Ollama.
• Chat Interface: A simple chat interface allows users to ask questions about the uploaded document.
• Background Processing: The PDF processing and indexing is done in the background to avoid blocking the UI.
• Status Polling: The UI polls the server to check the status of the indexing process.
• Overwrite Confirmation: The user is warned if an existing index will be overwritten.

Technology Stack

• Backend: FastAPI, Uvicorn
• Frontend: HTML, CSS, JavaScript
• Document Processing: unstructured, langchain, pypdf
• Vector Store: ChromaDB
• Embeddings: OpenAI or Ollama based on USE_OLLAMA_EMBED flag in .env
• LLM: OpenAI (for summarization and chat)

Installation

1. Clone the repository:

   git clone https://github.com/your-username/pdf-rag-chat.git
   cd pdf-rag-chat

2. Install the dependencies:

   uv sync

3. Set up your environment variables:

   Create a .env file in the root of the project and add your OpenAI API key:

   OPENAI_API_KEY=your-api-key
   USE_OLLAMA_EMBED=true/false
   

Usage

1. Run the application:

   uv run main.py

   Note: There will be a delay of few seconds When you start the app for the first

2. Open your browser:

   Navigate to http://127.0.0.1:8000.

3. Upload a PDF:

   • Click on the "Upload PDF" link in the navigation.
   • Select a PDF file to upload.
   • The application will start processing the document in the background.

4. Chat with the document:

   • Once the indexing is complete, you can go to the "Chat" page.
   • Ask questions about the document in the chat interface.

API Endpoints

• GET /: The home page.
• GET /upload: The file upload page.
• POST /upload: The endpoint for uploading a PDF file.
• GET /status/{task_id}: The endpoint to check the status of the indexing process.
• GET /index-status: The endpoint to check if an index is already available.
• GET /chat: The chat page.
• POST /chat: The endpoint for sending a chat message.

Future Improvements

• Support for more file types: The application could be extended to support other file types like DOCX, TXT, etc.
• More advanced chat features: The chat interface could be improved with features like chat history, user authentication, etc.
• Better error handling: The error handling could be made more robust.
• More sophisticated RAG pipeline: The RAG pipeline could be improved with more advanced techniques like re-ranking, query transformations, etc.