Eyes for text-only DeepSeek
xsoc1/dsh-image-vision · model-adapters
Add the DeepSeek to the plain text with the eyes: Maps in the chat + Image, a plugin is done.
★ 2GitHub stars
0Forks
2026-08-25Last updated
TypeScriptLanguage
NOASSERTIONLicense
Key features
- View image toolModels call with questions (OCR, numbers, graph reading, UI layout ... Any visual question), plugins forward pictures and questions to any OpenAI compatible VLM endpoint (local Ollama/ Intellectual / DashScape / Bean bag...) and the answer returns in text.
- Set of configurations (baseURL + apiKey + model) overwhelm all backends
- LegendThe dialogue box drags/ pastes dsh attachment service durability (/.dsh/attachments/v1/) message tape image block.
- Local peer (localhost) does not need key.
- Free-of-charge downgrading chainautomatically downgraded plugins in default configuration by type of gram-4.6v-flash ⁇ gl-4.1v-thinking-flash ⁇ gl-4v-flash (covered by fallbackModels).
Install command
dsh plugin --profile web add github:xsoc1/dsh-image-vision