Can i get a seperate model for vision as my main one doesnt support it

[subtle fiber]

Can i get a seperate model for vision as my main one doesnt support it

[devout sable]

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

[forest solstice]

Yes — set a separate vision analysis model with agents.defaults.imageModel.

Example in ~/.openclaw/openclaw.json:

{
  agents: {
    defaults: {
      model: {
        primary: "provider/text-model",
      },
      imageModel: {
        primary: "provider/vision-model",
        fallbacks: ["provider/backup-vision-model"],
      },
    },

  },
}

Or via CLI:

openclaw config set agents.defaults.imageModel '{"primary":"provider/vision-model"}' --strict-json

Important distinction:

• agents.defaults.imageModel = analyze/understand images
• agents.defaults.imageGenerationModel = generate images

Docs: https://docs.openclaw.ai/gateway/config-agents and https://docs.openclaw.ai/nodes/media-understanding

If you’re using a custom/local provider, make sure that model is marked image-capable with input: ["text", "image"]. For picking the exact model, #models will have fresher recommendations. 🦐

[subtle fiber]

Give me a command to search for avalible vision models in oc config

[forest solstice]

Use this:

openclaw models list --json | jq -r '.models[] | select((.input // "") | contains("image")) | .key'

If you only want models with auth available:

openclaw models list --json | jq -r '.models[] | select(((.input // "") | contains("image")) and (.available == true)) | .key'

Then set one as the vision model:

openclaw models set-image provider/model