Extract tables from images and PDFs using AI-powered OCR. Features local LLM (Ollama) and vector search (Qdrant) for intelligent document processing.
- Node.js 18+ (
node --version) - Docker (for Qdrant) – Install Docker
- Ollama – Install Ollama
git clone https://github.com/GAuravgiy87/OCR-.git
cd OCR-npm install
# If you encounter peer dependency errors, use:
npm install --legacy-peer-depsdocker run -d -p 6333:6333 --name qdrant qdrant/qdrant
# Verify it's running:
curl http://localhost:6333/collections # Should return {"result":{"collections":[]}}
# Or open http://localhost:6333/dashboard in your browserPull the required models:
ollama pull nomic-embed-text # For embeddings
ollama pull qwen2.5:1.5b # For chat assistant
# Verify installed models:
ollama listOllama serves models automatically on http://localhost:11434.
npm run devOpen http://localhost:3000 in your browser.
- Qdrant running:
docker ps | grep qdrant - Ollama models loaded:
ollama listshows both models - App accessible: http://localhost:3000 loads without errors
- Upload an image or PDF containing tables.
- Wait for OCR processing (automatic table detection).
- Edit extracted data if needed.
- Export as CSV or Excel.
- Port conflicts: Ensure ports 3000 (Next.js), 6333 (Qdrant), and 11434 (Ollama) are free.
- Ollama not responding: Run
ollama servein a separate terminal. - Qdrant connection error: Restart container:
docker restart qdrant.
Happy extracting!