DeepSeek-OCR-WebUI
DeepSeek-OCR-WebUI is a ready-to-use web interface for the DeepSeek-OCR-2 optical character recognition model. Built with Vue 3 and TypeScript, it provides a modern, fully responsive UI for extracting text, tables, charts, and structured content from images and documents. It supports seven recognition modes including document conversion, OCR, figure parsing, freeform extraction, and custom prompts. Batch processing allows multiple files to be handled simultaneously, while real-time logging provides live feedback on processing status. The backend features non-blocking inference with ThreadPoolExecutor, per-client and per-IP rate limiting, queue management with position tracking, and an enhanced health API exposing model version, queue depth, and rate limit metrics. DeepSeek-OCR-2 introduces Visual Causal Flow architecture, dynamic resolution up to 1024x1024, native Flash Attention 2 support on CUDA, and improved accuracy for document understanding and chart parsing. Distribution is available via Docker with a