How to return an image from my tool?

[jareks]

I would like to return an image as a result of my custom tool so that LLM can analyze it.

It seems that the key is to specify type="image_url" and image_url with url or base64 inline data:
https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/gpt-with-vision?tabs=python

Here's an example how it can be done in langchain:
langchain-ai/langgraph#4384 (comment)

[HaoLi111]

https://medium.com/@tonyalfredsson/my-3-15-am-start-to-building-an-ai-image-agent-with-google-adk-gpt-image-1-dev-life-log-1-c0dce525ea85

Is it similar to this?

If multi agent, someone mentioned the problem

https://stackoverflow.com/questions/79633223/google-adk-ui-shows-raw-img-tag-instead-of-rendering-image-from-aws-s3-presign