https://www.boundaryml.com logo
🟡 Issue with configuring gpt-image-1 using openai...
# questions
s
Any guidance on configuring to use gpt-image-1? I am using openai-responses provider, but it doesn't construct the image input correctly: Unspecified error code: 400 Request failed with status code: 400 Bad Request, { "error": { "message": "Invalid value: 'input_image'. Supported values are: 'input_text'.", "type": "invalid_request_error", "param": "input[0].content[1]", "code": "invalid_value" } } Request options: {"model":"gpt-image-1"}
a
is this using a user role?
openai responses only accepts images in the 'user' field
in the user chat msg
s
Hmm, OK, I can try that.
But last I knew, you don't support images in the output yet?
In which case this won't entirely work for my use case.
a
Oh we cannot generate images yet
s
😭
Do you take PRs?
a
we definitely take PRs, that one will be challenging, but it would be doable to convert the output back into a BamlImage. The thing is we would want that BamlImage to now have more methods, etc. Not sure what format providers return images in
even a specification doc on what we would want the Image object in python to do may help
s
Why more methods? Not enough to just hand back an Image object?
a
I suppose we just have to figure out the properties of the image object
Like at bare minimum just the array of bytes and media type
s
I've got next week off, maybe I'll look at it.
a
It d be sweet 🙂