Image Generation
Image Generation creates an image from a text prompt and persists a successful result as a conversation attachment. The product tool currently uses the OpenAI Responses image-generation path.
Configure the Model
Open Dashboard → Tools → Image and configure Image Generation, or run:
anyy tools image generation
Choose the provider route and model, add the required credential, set a default aspect ratio, then probe the route before relying on it. auxiliary_models.image_generation takes priority. If no auxiliary model is configured, the runtime can use a main model that declares image_generation. Either selection requires api_mode: openai-responses and a resolvable credential; an ordinary OpenAI-compatible Chat endpoint alone does not satisfy this contract.
Dashboard → Models also configures image generation. Codemode’s models.generateImages() uses the separate image catalog type and openrouter-images API; it does not reuse this tool setting. See Model interfaces.
Generate an Image
Describe the subject, composition, style, lighting, and any text that must appear:
Create a clean editorial illustration of a resident AI assistant coordinating chat, calendar, and files. Use a landscape composition and no logos.
When the tool is available and the provider returns a valid successful result, Anyy delivers the file as a conversation attachment. Saving configuration or listing a model is not a live generation test for the account.
Aspect Ratio
Anyy accepts three aspect-ratio values: square, portrait, and landscape. If you omit the value, Anyy uses the configured default. The selected provider must also support the requested orientation; invalid or unsupported values are rejected rather than silently changed.
Name the intended use, orientation, visual hierarchy, exclusions, and exact text. Image providers may render text imperfectly, so inspect important wording.
Limits and Delivery
The default generation timeout is two minutes and the default accepted image payload limit is 5 MiB unless the Profile configuration changes them. A successful result is persisted as a scoped attachment so Dashboard and supported channels can deliver it.
Generation sends the prompt and required request data to the selected provider. Each request may consume paid quota.
Safety and Privacy
- Do not include credentials or private data in a generation prompt.
- Verify the provider's retention and content rules.
- Review images before publishing, especially faces, brands, medical content, and factual diagrams.
- Treat the output as generated content; do not present it as documentary evidence.
Troubleshooting
| Symptom | Check |
|---|---|
| Image tool is unavailable | Enable Image Generation and verify the auxiliary route |
| Authentication or model error | Re-test the selected provider, credential, endpoint, and model ID |
| Request times out | Simplify the request or review the configured timeout/provider health |
| Result is rejected as too large | Use a provider output within the configured maximum size |
| Attachment does not reach a channel | Verify that channel supports the generated attachment kind and size |