Holon

Open-source AI assistants on your own computer or server. Assign tasks, track progress, and let them follow up on ongoing work.
Skip to content
Browse docs
On this page

Generate an image

Agents can create images from a text prompt with the GenerateImage tool. You describe what you want; the agent calls the tool and saves the result.

Before you start

  • The agent can use GenerateImage.
  • An image-generation route and its credentials are configured. Follow Configure models; a conversation model alone may not support image generation.
  • The runtime can reach the provider and write to the agent's media directory.

Steps

  1. Ask for the image and include the details that matter:

    Generate a 1024x1024 product shot of a matte black desk lamp on a white
background. Save it as desk-lamp.png.

    The agent calls GenerateImage with your prompt and supported output options. Each call generates one image; request separate calls for distinct images.

  2. Ask the agent to report the returned file reference. Generated images are written under agent_home/media/generated; the requested name is not a substitute for the actual saved path.

  3. Open the image from the response or the client's file browser. If you need the agent to check its content, follow Inspect an image.

Confirm it worked

Confirm that the tool succeeded and returned an image reference and metadata, then open the saved artifact. Check the image against your prompt; a successful provider call does not guarantee that every requested detail appears.

Troubleshooting

  • No generation model: Check the image-generation route and provider credentials in Configure models.
  • Unsupported option: Use a size, format, and background supported by the configured model. The reference lists the tool's accepted values.
  • Provider or network error: Read the returned error before retrying. Do not report an image as generated without a successful tool result.
  • Cannot open the result: Check the returned path on the runtime host. A remote host's filesystem path does not identify a file on your laptop; use the client's artifact viewer or download action.

Sizes and formats

Size, background, and format options, plus supported models, are in the Image generation reference.