RAG Document Assistant 📚

A Retrieval Augmented Generation (RAG) application that allows you to ask questions about documents. The system includes ManishKumar's resume as a permanent document and supports uploading additional PDFs for temporary use during your session.

Features

• Permanent Resume: ManishKumar's resume is always available in the vector database
• Temporary Upload: Upload additional PDFs for your current session
• Auto Cleanup: Session documents are automatically removed when the session ends
• Smart Q&A: Ask questions about any loaded document using Google Gemini
• Web Interface: User-friendly Streamlit frontend
• API Backend: FastAPI backend for document processing

Architecture

• Backend: FastAPI with ChromaDB for vector storage
• Frontend: Streamlit web interface
• AI Model: Google Gemini 2.0 Flash for text generation
• Embeddings: HuggingFace GIST-large-Embedding-v0
• Document Processing: Docling for PDF processing and chunking

Deployment

For Streamlit Cloud deployment, a separate streamlit_app.py file is provided that combines both frontend and backend functionality into a single Streamlit application.

Setup

Prerequisites

• Python 3.8+
• Google API Key (for Gemini)

Installation

1. Clone and navigate to the project:

   cd d:\Project\RAG

2. Install dependencies:

   pip install -r requirements.txt

3. Set up environment variables: Create a .env file in the project root:

   GOOGLE_API_KEY=your_google_api_key_here
   LANGSMITH_API_KEY=your_langsmith_api_key_here  # Optional

4. Ensure the resume file exists: Make sure ManishKumarResume.pdf is in the pdfs/ folder

Running the Application

Option 1: Using the startup script (Recommended)

python run_app.py

Option 2: Using the batch file (Windows)

start.bat

Option 3: Manual startup

1. Start the API server:

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

2. Start the Streamlit frontend (in another terminal):

   streamlit run frontend.py

Usage

1. Access the application:

   • Frontend: http://localhost:8501
   • API Documentation: http://localhost:8000/docs

2. Ask questions about the resume:

   • The system always has access to ManishKumar's resume
   • Ask questions like "What is ManishKumar's experience in Python?"

3. Upload additional documents:

   • Use the sidebar to upload PDF files
   • These files are processed and added to the current session
   • Ask questions about the uploaded content

4. Session management:

   • Uploaded documents are automatically removed when the session ends
   • Use the "Clear Session Documents" button to manually clean up
   • The resume file remains permanently available

API Endpoints

• GET /health - Health check
• POST /query - Ask questions about documents
• POST /upload - Upload a PDF for the current session
• POST /cleanup-session - Clean up session documents
• GET /loadedpdfs - List available documents

File Structure

d:\Project\RAG\
├── api.py              # FastAPI backend
├── frontend.py         # Streamlit frontend
├── textRAG.py         # Core RAG implementation
├── run_app.py         # Startup script
├── start.bat          # Windows batch file
├── requirements.txt   # Python dependencies
├── README.md          # Documentation
├── .env              # Environment variables (create this)
├── pdfs/             # PDF storage
│   └── ManishKumarResume.pdf
└── test_chroma_db/   # Vector database storage

Troubleshooting

1. API not starting: Check if port 8000 is available
2. Streamlit not starting: Check if port 8501 is available
3. Missing resume: Ensure ManishKumarResume.pdf is in the pdfs/ folder
4. Upload failures: Check file size and ensure it's a valid PDF
5. Google API errors: Verify your GOOGLE_API_KEY in the .env file

Development

To extend the application:

1. Add new document types: Modify the file upload validation in api.py
2. Change embedding models: Update the model in textRAG.py
3. Customize UI: Modify the Streamlit interface in frontend.py
4. Add new endpoints: Extend the FastAPI routes in api.py

Dependencies

Key dependencies include:

• fastapi - Web API framework
• streamlit - Web interface
• langchain - LLM framework
• chromadb - Vector database
• transformers - HuggingFace models
• docling - Document processing
• google-generativeai - Google Gemini integration

See requirements.txt for the complete list.