xing666173/dsh-vision-hub#tool-vision

GitHub: Scorp1o117/dsh-tool-vision · npm: dsh-tool-vision

Part of the DeepSeek Harness Enhancement Suite — Vision · Soul/Persona · Long-term Memory · Plugin Marketplace.

External vision model for DeepSeek Harness.

DeepSeek's own models are text-only, and the harness derives every model request strictly from the session log (llm/stream requests must equal the durable derivation — the agent-loop invariant). This plugin bridges the gap in two ways:

1. inspect_image tool — sends an image (local file, or http(s) URL) to any OpenAI-compatible /chat/completions endpoint that supports image_url content parts, and returns the vision model's textual answer into the agent loop.
2. Image bridge (v0.2.1) — pasted images are turned into inspect_image hints before they enter the durable log, on the agent/pre-step waterfall (the one seam where the harness lets a plugin replace the messages of a proposed step). Images already logged by an older version are repaired lazily with a surface replace on the session's first pre-step. Only models listed in multimodalModels receive image blocks directly; a model's declared inputModalities are never consulted, because profiles routinely declare input: [text, image] on text-only models just to pass the harness's prompt-admission check.

• Zero dependencies beyond the dsh SDK — works with any compatible endpoint: OpenAI GPT-4o, Qwen-VL (DashScope), GLM-4V (Zhipu), Moonshot, Gemini compatible endpoints, local Ollama, etc.
• Registered on the global tools layer: every agent in the process can call inspect_image.
• Web UI settings section (v0.3.0): Settings → 视觉模型 edits the tool-vision namespace (API endpoint, write-only key, model, bridge options) in settings.yaml; changes hot-apply without a restart. The API key lives in settings.yaml, not the profile patch. Mount by package name (name: 'dsh-tool-vision') so the web client bundle is discovered.

Install

Mount in a profile patch ($DSH_HOME/profiles/<name>/cordis.patch.yml):

- insert:
    - id: tool-vision
      name: 'dsh-tool-vision'     # after: pnpm add dsh-tool-vision in the profile
      config:
        baseURL: 'https://api.openai.com/v1'
        apiKeyEnv: 'VISION_API_KEY'
        model: 'gpt-4o-mini'

Or load it from a local path without npm:

    - id: tool-vision
      name: './plugins/dsh-tool-vision/index.js'

Config

Field	Default	Meaning
baseURL	https://api.openai.com/v1	OpenAI-compatible API base URL.
apiKey	''	API key (takes precedence over env).
apiKeyEnv	VISION_API_KEY	Env var holding the key.
model	gpt-4o-mini	Vision model id.
maxTokens	1024	Max output tokens.
timeoutMs	60000	Per-request timeout.
maxImageBytes	10MB	Largest accepted local image.
description	default	Tool description shown to the model.
bridgeTextOnly	true	Bridge pasted images to text hints on models that cannot see images.
bridgeExportDir	temp	Export dir for bridged images (os.tmpdir()/dsh-vision-bridge).
multimodalModels	[]	Model ids that receive image blocks directly (e.g. mimo-v2.5).

Image bridge setup

1. In your model settings, declare image input on the models you paste images onto, so the harness admits image messages (pi-ai style):

   llm-pi-ai:
     providers:
       your-provider:
         models:
           - id: deepseek-v4-flash
             input: [text, image]
   

2. List genuinely multimodal models in the plugin config so they receive image blocks untouched:

   - id: tool-vision
     name: 'dsh-tool-vision'
     config:
       multimodalModels: ['mimo-v2.5', 'grok-4.5']
   

Then pasting an image while on a text-only model stores a hint like [User sent an image, exported to: <path>. Inspect it with the inspect_image tool...] in the transcript (the pasted image no longer renders as pixels in that message), and the agent inspects it through the configured vision endpoint.

Why not llm/stream? The harness freezes every request and the agent-loop invariant fails any request whose messages diverge from the session-log derivation (log-reconstruction desync), and this cordis waterfall's next() cannot replace request arguments. The agent/pre-step waterfall is the supported seam: its decision messages become the durable log, so the invariant stays satisfied.

Key resolution order: config.apiKey → process.env[apiKeyEnv] → process.env.OPENAI_API_KEY.

Tool: inspect_image

Arg	Required	Meaning
path	✅	Image path (absolute, or relative to the current workspace) or http(s) URL.
question	–	Optional specific question about the image.
detail	–	auto / low / high resolution hint.

Example endpoints (baseURL):

• OpenAI: https://api.openai.com/v1 — gpt-4o, gpt-4o-mini
• Alibaba DashScope (Qwen-VL): https://dashscope.aliyuncs.com/compatible-mode/v1 — qwen-vl-plus, qwen-vl-max
• Zhipu (GLM-4V): https://open.bigmodel.cn/api/paas/v4 — glm-4v-flash (free tier), glm-4v-plus
• Moonshot (Kimi): https://api.moonshot.cn/v1 — moonshot-v1-8k-vision-preview
• Ollama local: http://localhost:11434/v1 — llama3.2-vision (no key)

Note for users

• This plugin is a standard profile bundle (dsh.bundle.patch): dsh plugin --profile web add dsh-tool-vision installs and mounts it in one step — no manual cordis.patch.yml edits needed.
• The settings section needs the dsh-host-apiproxy namespace allowlist; the plugin patches it automatically on first start — restart dsh web once more and the section appears. A dsh update overwrites the patch; the next plugin start re-applies it.
• Settings changes hot-apply (no restart needed).
• Tested against DSH 0.1.0-rc.6.

Pixel-level vision tools (v0.4.0, ported from dsh-vision-router)

14 vision_* tools driven by the same configured endpoint as inspect_image (baseURL/apiKey/model) — no provider chain, no local models, no extra settings:

Tool	Purpose
vision_describe	Image Q&A / multi-image comparison (optional structured JSON)
vision_ground	Locate a target and return its ORIGINAL-pixel bounding box
vision_detect	Enumerate elements (buttons, inputs, icons…) with numbered boxes
vision_crop	Crop a pixel region to a PNG artifact
vision_pixel_diff	Per-pixel comparison: ratio, worst regions, heatmap, report
vision_colors	Dominant-color quantization for palette matching
vision_ocr	Verbatim text transcription (letters only — not scene analysis)
vision_long_screenshot_ocr	Chunked long-screenshot transcription into Markdown
vision_trace	Potrace vectorization into colored SVG (worker-thread, safe)
vision_extract_foreground	Solid-background removal → transparent PNG
vision_html_screenshot	Headless render of a local .html (network blocked)
vision_screenshot	Desktop capture (Win: PowerShell; macOS: screencapture; Linux: import/scrot)
vision_present	Publish a generated image to the user via the host attachment store
vision_materialize	Copy an attachment/local image into the workspace as a real path

Quality & safety details:

• Content-hash cache keyed by endpoint+model+image+question (no stale answers across model switches, failures are never cached).
• Uniform 4MP downscale before every model call; oversized inputs are rejected with a clear error (stat pre-check, 20MB cap on both file and attachment paths).
• Rate-limit / 5xx auto-retry with Retry-After-aware backoff; endpoint content-safety rejections are surfaced as VISION_CONTENT_FILTERED instead of a generic backend error.
• Long-OCR bounds: 120s total budget, 40-chunk cap, cancellation checks, stop-on-first-backend-failure.
• Path containment for relative inputs; artifacts land in <workspace>/.dsh-tool-vision/.
• Bridge marker: pasted images become a short [图片: <path>] marker and the usage rule lives in a system-prompt section (clean transcript, same model behavior).

Requires sharp / potrace / puppeteer-core (declared in dependencies; missing ones degrade lazily with an install hint and never break other tools).

Limitations

• A bridged image enters the conversation as a text hint (a transcript, not pixels) — pixel-precise in-context reasoning is not available to text-only models; the vision model's description comes back through inspect_image.
• Images are base64-transferred; mind privacy and size limits.
• Independent of the dsh-llm routing/retry system; failures return clear errors to the agent.

License

MIT