PDF Q&A System

A multilingual (Hebrew/English) PDF question-answering system with grounded answers and citations.

Quick Start

Model Setup

The application uses embedding models for semantic search. You have two options:

1. Automatic online download (default):

   • Models will be downloaded automatically from Hugging Face on first run
   • Requires internet connection
   • Can be slower for initial startup

2. Local models (recommended):

   • Run the included script: download_models.bat
   • This downloads models to the configured local directory (default: D:/models/)
   • Faster startup and no internet dependency

To use local models, ensure use_local_models = True in app/core/config.py.

Running Locally

1. Setup environment:

   cp .env.example .env
   # Edit .env with your OpenAI API key

2. Place your PDFs:

   • Put your PDF files in the data/pdfs folder

3. Start the application:

   • Run the included batch file: run_app.bat
   • Or manually: venv\Scripts\python -m uvicorn app.main:app --host 0.0.0.0 --port 8000
   • Access the UI at: http://localhost:8000

4. If you encounter issues:

   • Try the development version: venv\Scripts\python run_dev.py
   • Access the development UI at: http://localhost:8001

Common Issues and Solutions

• Hebrew text issues: The application now supports Hebrew text properly.
• Connection timeout: For large PDFs, increase the timeout in app/ui/app_ui.py.
• Slow first run: The first run will download language models, which may take time.

Using Docker (Coming Soon)

Docker support is planned for a future update.

Features

• Multilingual support for Hebrew and English PDFs
• Contextual Q&A with source citations
• Evidence panel showing exact sources
• Language selection for answers