opencode-image-vision

OpenCode plugin that gives text-only models (GLM-5, DeepSeek V4, MiniMax, etc.) the ability to understand pasted images. Images are analyzed by a vision model in the background and replaced with text descriptions before the chat model runs — paste → ask → done.

Requires OpenCode 1.14+, Node 18+, and a signed-in vision provider (e.g. glm-4.6v).

Install

Add to ~/.config/opencode/opencode.json (or opencode.jsonc):

{
  "plugin": [
    [
      "@showlotus/opencode-image-vision",
      {
        "model": "zhipuai-coding-plan/glm-4.6v",
        "timeout": 120000,
      },
    ],
  ],
}

model is required (providerId/modelId). API keys are read from OpenCode's auth.json — no extra key setup.

For local development, use a file:// absolute path (symlinks in plugins/ are skipped):

["file:///Users/YOU/path/to/opencode-image-vision", { "model": "zhipuai-coding-plan/glm-4.6v" }]

Restart OpenCode, paste an image, and ask about it.

Options

| Option | Required | Default | Description | | --------- | -------- | -------- | ----------- | | model | Yes | — | Vision model, e.g. zhipuai-coding-plan/glm-4.6v | | prompt | No | built-in | Analysis prompt | | timeout | No | 120000 | Vision API timeout (ms) | | debug | No | false | Log to /tmp/iv-debug.log |

Debug via env: IMAGE_VISION_DEBUG=1 (optional IMAGE_VISION_DEBUG_PATH).

See config.example.json for more examples.

How it works

On each message, the experimental.chat.messages.transform hook:

1. Scans only the current user message for image parts
2. Skips if the chat model already supports image input
3. Calls the vision API and replaces each image with [图片识别结果]\n… text

Failed images become [图片识别失败: reason] and do not block others. Identical images are cached by MD5.

Providers

Any provider registered in src/opencode.js works. Common examples:

| Provider ID | Example model | | ----------- | ------------- | | zhipuai-coding-plan | glm-4.6v | | openai | gpt-4o | | qwen / dashscope | qwen-vl-max | | anthropic / claude | claude-3-5-sonnet-20241022 |

Base URLs are resolved from OpenCode at runtime when available. To add a provider, update PROVIDER_REGISTRY and PROVIDER_MAP in src/opencode.js and src/providers/index.js.

Troubleshooting

• Plugin not loading — Confirm model is set; check logs for [image-vision] 初始化失败; run opencode auth
• Timeout on large images — Increase "timeout": 120000 (or higher)
• No API key — Sign in to the vision provider in OpenCode (opencode auth)
• Local dev — Use file:///absolute/path, not symlinks; set "debug": true and check /tmp/iv-debug.log

License

MIT — see LICENSE.