DeepEye vision plugin for DeepSeek Harness (DSH)
Favio8/dsh-plugin-deepeye · model-adapters
DeepEye visual plugin for DeepSeek Harness (DSH): Image description, OCR, VQA, UI layout and clipboard analysis.
★ 4GitHub stars
0Forks
2026-08-25Last updated
TypeScriptLanguage
MITLicense
Key features
- Primary Isolation - Direct registration to ctx.tools, no MCP intermediate expenses
- Local path/path/to/image.png or relative path
- Multibackend - Support OpenAI (GPT-4o), Google Gemini, Custom OpenAI-compatible endpoint
- Smart preprocession - Auto-shrimp large pictures, convert JPEG to save token
- Result Cache - LRU Cache Reduce Repeated API Call (with TTL key over backend/model, toggle backend does not miss)
Requirements
- The DSH frontend parent supports pasting/ drag-in of pictures into input boxes (stored as attachments, message is an image reference block).
- But there's two levels that won't get us out of here.
Install command
dsh plugin --profile web add github:Favio8/dsh-plugin-deepeye