DeepSeek-OCR Dockerized API

Use it when

• Scanned or image-based PDFs have no text layer and you need structured markdown out of them
• Confidential documents cannot be uploaded to a cloud OCR service, so extraction must run entirely on your own hardware

What it solves

Not the fit when

• machines without an NVIDIA GPU
• audio or video transcription
• extracting article text from live web pages
• no CPU-only mode; hard requirement on an NVIDIA GPU with 12GB+ VRAM and CUDA 11.8
• output quality depends on the DeepSeek-OCR model; no accuracy guarantees documented for handwriting

Install

git clone https://github.com/Bogdanovich77/DeekSeek-OCR---Dockerized-API && cd DeekSeek-OCR---Dockerized-API && huggingface-cli download deepseek-ai/DeepSeek-OCR --local-dir models/deepseek-ai/DeepSeek-OCR && docker-compose build && docker-compose up -d (requires NVIDIA GPU with 12GB+ VRAM, CUDA 11.8, NVIDIA Container Toolkit)

Invoke

curl -X POST http://localhost:8000/ocr/pdf -F "file=@document.pdf" (also /ocr/image and /ocr/batch), or drop PDFs in data/ and run python pdf_to_markdown_processor.py; verify with curl http://localhost:8000/health

Alternatives

No reviewed alternatives recorded yet.