corrinehu/dsh-chat-imagine
Automatically generates and displays images in the DSH chat via API channels or local CLIs (supports mmx / codex / agy).
Generates and displays images inline in the DeepSeek Harness chat. It can use OpenAI-compatible API providers already configured in DSH (built-in providers with a blank base URL fall back to DSH's default endpoint), or local CLIs: MiniMax mmx, OpenAI Codex codex (spends ChatGPT Plus/Pro quota, not an API key) and Google Antigravity agy (spends Google account quota). When codex or agy is detected it also registers the cli-image-gen skill so the model can drive the CLIs when generate_image fails. The same CLI channels power image recognition: analyze_image returns structured JSON evidence (full OCR text, reading-order layout regions, semantic entities and relations, visual cues, uncertainty list) so any model — including text-only ones — can read an image without switching to a vision model.
Install
dsh plugin --profile web add dsh-chat-imaginenpm package dsh-chat-imagine 0.4.1 (registry-verified 2026-09-10; registry repository field -> github.com/corrinehu/dsh-chat-imagine) ships prebuilt output, which the README recommends. A source install from GitHub is also documented: dsh plugin --profile web add github:corrinehu/dsh-chat-imagine. Requires pnpm on PATH.
Compatibility
Tested in the DSH Web profile (the plugin's README states it has only been verified there). Image recognition additionally requires one local CLI installed and signed in: MiniMax mmx, OpenAI Codex codex, or Google Antigravity agy.
Details
- Repo: corrinehu/dsh-chat-imagine
- Category: Vision & Multimodal
- Stars: 10
- Version: npm dsh-chat-imagine 0.4.1 (registry-verified 2026-09-10)
- Last push: 2026-08-26
- First seen: 2026-08-15
FAQ
- How do I install dsh-chat-imagine?
- Run: dsh plugin --profile web add dsh-chat-imagine — the README recommends the npm channel because it ships prebuilt output. A GitHub source install is also documented.
- Which image-generation backends does dsh-chat-imagine support?
- OpenAI-compatible API providers already configured in DSH, plus the local CLIs mmx (MiniMax), codex (ChatGPT Plus/Pro quota) and agy (Google Antigravity quota). The plugin scans for installed CLIs and offers each one found.
- Do I need a vision model to read images?
- No. analyze_image uses the CLI channels' vision capabilities and returns structured JSON evidence — OCR text, layout regions, entities and relations — so any model in the session can use the result.
Alternatives
dickpy/dsh-imagegen · ConsoleSun/Gemini-Eyes · 54xkeee/dsh-vision