dsh-tesseract-ocr

When interacting with AI models, image attachments often raise privacy and compatibility issues. The dsh-tesseract-ocr plugin solves this by enabling text-only models to handle images while performing local OCR recognition, sending only the extracted text to the model API, thus ensuring privacy by default. Its core capabilities include a local Tesseract OCR engine, multi-language support, caching for efficiency, and an opt-in passthrough mode for genuine vision models. Designed for DeepSeek Harness users, it is ideal for privacy-conscious developers, researchers, and everyday users who need OCR functionality without model configuration changes, offering a secure and flexible way to extend text-based model interactions with images.