ragadoc

AI document assistant that answers questions about your PDFs with citations and highlights.

Privacy-first Streamlit app for chatting with documents using local AI models.

✨ Features

• 🤖 AI Document Q&A - Natural language questions about your PDFs
• 📍 Citation Grounding - Answers include specific citations from your document
• 🎯 PDF Highlighting - Citations automatically highlighted in the PDF
• 🔒 Complete Privacy - Local AI models only, documents never leave your computer

⚠️ Early Development - This is a proof of concept. Expect incomplete features and potential breaking changes.

🚀 Quick Start

Prerequisites

1. Install and start Ollama:

# macOS
brew install ollama

# Or download from https://ollama.com

2. Pull models (recommendations: nomic-embed-text for embeddings, olmo3:7b or olmo3:32b for chat):

ollama serve
ollama pull <embedding-model>
ollama pull <chat-model>

Option 1: uv (Recommended)

Requires Python 3.12 and uv.

git clone https://github.com/clstaudt/ragadoc.git
cd ragadoc
uv sync
uv run streamlit run app.py

Open http://localhost:8501

Option 2: Docker

git clone https://github.com/clstaudt/ragadoc.git
cd ragadoc
docker-compose up --build

Open http://localhost:8501

⚙️ Configuration

Copy env.example to .env to configure Ollama instances (local or remote).

📖 Usage

1. Select Models - Choose chat and embedding models in the sidebar
2. Upload a PDF - Drag and drop your document (extraction and indexing happens automatically)
3. Chat - Ask questions in natural language
4. View Citations - Answers include citations with highlighted PDF passages shown below

Expert Mode (optional): Toggle in sidebar to adjust RAG parameters like chunk size, similarity threshold, and retrieval count.

🏗️ Architecture

PDF Upload → Text Extraction → Chunking → Vector Embeddings
                                               ↓
User Question → Semantic Search → Context Retrieval → AI Response
                                               ↓
                                    Citation Highlighting

Tech Stack: Streamlit • Ollama • ChromaDB • PyMuPDF4LLM • uv

🐛 Troubleshooting

Ollama Connection Error

curl http://localhost:11434/api/version
# For Docker: OLLAMA_HOST=0.0.0.0:11434 ollama serve

Slow Performance - Try a smaller model or reduce chunk size in RAG settings.

📄 License

GPL License - see LICENSE.

---

⭐ Star this repo if Ragadoc helps you work with your documents!