An image generation agent produces images from instructions, commonly exposing an API that another agent can call as a tool. Unlike a synchronous chat reply, image creation may take tens of seconds to minutes, so requests continue asynchronously while the user keeps interacting. section 8.1.3
A practical mental model separates the agent-side flow from the infrastructure around it:
- Receive an image request through an API or a voice agent. section 8.1.3
- Translate the request into a concise image description and invoke the image-generation tool. section 8.1.3
- Return or display the finished image after processing completes. section 8.1.3
- Scale production deployment with queues, GPU-enabled workers, temporary storage, notifications, and persisted conversation state. section 8.1.3
Voice request to generated image
A user says, “Create an image of a playful robot painting a sunset.” The browser voice agent calls the image-generation tool, the API agent processes the request, and the resulting image appears in the browser while ordinary conversation remains available. section 8.5
