dsh-vision
UnverifiedAdds vision to text-only DeepSeek via OpenAI-compatible VLM
Local OCR for images and scanned documents
This plugin uses PaddleOCR and FastAPI locally to read PNG/JPEG/WebP images, screenshots, and scanned documents, enabling text-only DeepSeek models to obtain text, coordinates, and confidence via vision_read and vision_read_region. Supports CPU, region OCR, text/structured/markdown output, Bearer Token, and secure file validation, with no cloud vision API.
Before you install
This plugin has not completed plugin-level verification. Review the upstream README, required permissions, dependencies, and data flow first.
Taken from the author’s README. Follow the upstream instructions to install.
Loading upstream installation instructions…
Original documentation
Read the author’s README in place. Documentation and directory verification remain separate facts.
Loading the original README…
Adds vision to text-only DeepSeek via OpenAI-compatible VLM
Chrome sidebar extension for direct browser control
Vision bridge for image description
