Models And Provider Setup

List models: GET /api/models?organizationId= returns text models available to the org by default.

Use GET /api/models?organizationId=&modalities=text,image when you need generated-image model truth as well as normal text model discovery.

SDK helper: client.setup.listModels({ organizationId, includeUnavailable?, modalities? })

Run overrides: POST /api/skills/:id/run accepts optional modelKey in the body to select a provider/model for that execution.

Prefer model discovery over hardcoded model ids in builder flows.

Image-generation capability

Model discovery exposes image-generation capability as explicit metadata:

• supportsImageGeneration
• supportsImageInput
• imageGenerationMode
• inputModalities
• outputModalities
• capabilities

Do not infer image-generation support from a model name. Some direct image models have image-focused names, while some mainline models can request image generation through a hosted provider tool.

Existing skill execution selectors should continue using the default text-model discovery path. Clients that need to inspect generated-image capability should request modalities=text,image so direct image models and hosted-image-generation-tool-capable mainline models are visible together.

See also

• Models route (orchestration overview)
• Generated Image Artifacts
• Runtime vs Preview