Question: Do the SDKs supports image attachments?

[zavocc]

I can't seem to find any documentation or api reference about image inputs... as the SDK (I'm primarily going to use Python) seems to lack support for images?

Would really be useful especially involving certain use cases like UI to code

[github-actions]

Thanks for opening this issue, @zavocc!

I've labeled this as a question related to sdk/python and documentation. Your question about image attachment support is valid and the use case (UI to code) is interesting.

The team will review whether image inputs are currently supported in the Python SDK and update the documentation accordingly if needed. If this feature isn't available yet, this could become an enhancement request.

AI generated by Issue Triage Agent

[friggeri]

Thanks for the question @zavocc, this should be better documented, when you submit a prompt using CopilotSessionsend you can pass an attachments option, if you pass a path to an image it will be attached to your request, for example:

session.send({
  "prompt": "what's in this image?",
  "attachments": [
    Attachment(
      type="file",
      path="/path/to/image.jpg"
    )
  ]
)

In addition, the agent's view tool has the ability to read images, so you could also do:

session.send({"prompt": "what does the most recent jpg in this directory portray?"})