Attach input to a message
- Open the Chat or Code composer and confirm its selected model.
- Use Add to choose an available attachment action, or paste an image where the composer supports it.
- Inspect the attachment preview and remove anything you do not intend to send.
- Describe the relevant detail or requested analysis, then send.
- If the model rejects the input, read the error and check the format and model capability before retrying.
Opening an image in a workspace tab is not the same as attaching it to the model request. Attachment formats, size limits, and tool combinations depend on the selected route; this draft does not claim universal format support.
Configure image generation
- Open Settings → Image.
- Choose OpenAI image models or Foundry image deployment under Image provider.
- Select the offered model or deployment and configure the matching API connection under Models.
- Save settings, return to Chat, and use the image-generation action offered by the composer.
- Describe the image and inspect the returned result before saving or sharing it.
An external Code subscription does not automatically cover this API operation. A Foundry deployment name must refer to a suitable configured image deployment; selecting a text deployment does not make it image-capable.
Recover missing results
Separate upload/input validation from generation errors and result-display failures. Preserve the exact sanitized error and the selected model/deployment. Avoid repeatedly issuing the same generation request when completion is uncertain, since a retry can create another billable request.
See API connections, Files, and usage accounting.