[BUG] Unable to enum o1 models on Azure OpenAI

[alkampfergit]

Environment

big-agi.com (production website)

Description

When I try to refresh my deployment, if I deploy an o1-mini model I got this error

[ { "received": "failed", "code": "invalid_enum_value", "options": [ "succeeded" ], "path": [ "data", 3, "status" ], "message": "Invalid enum value. Expected 'succeeded', received 'failed'" } ]

I've deployed o1-mini on another datacenter, this time I do not get error, but the model does not show in the list of models.

Device and browser

WIndows, edge browser.

Screenshots and more

Willingness to Contribute

• 🙋‍♂️ Yes, I would like to contribute a fix.

[enricoros]

Hi @alkampfergit can you try now? I've just relaxed the parser and should proceed.

[alkampfergit]

I'll upgrade and try Il mer 29 gen 2025, 19:40 Enrico Ros ***@***.***> ha scritto:

…

Hi @alkampfergit <https://github.com/alkampfergit> can you try now? I've just relaxed the parser and should proceed. — Reply to this email directly, view it on GitHub <#744 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACXREIQ5BDZZF6NDVJYWKL2NEOCVAVCNFSM6AAAAABWDHMVH6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMRSGUZTQOJRGI> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[alkampfergit]

I tried to update docker image but I didn't see any new image, still have the same problem

[ { "received": "failed", "code": "invalid_enum_value", "options": [ "succeeded" ], "path": [ "data", 3, "status" ], "message": "Invalid enum value. Expected 'succeeded', received 'failed'" } ]

[rafaelpierre]

same here, I don't get any error messages but my o1 endpoints from Azure don't show up

[enricoros]

Reopening because evidently there's an issue with Azure models. Note: the complexity with azure it's always been the dynamic listing, as the names of deployments are likely non standard and one cannot guess the context window sizes, model capabilities, etc. The code right now is looking up the official OpenAI model names, but in this particular instance there could be other underlying issues.

[rafaelpierre]

Wouldnt make sense to ask users to include the deployment name in the URL for Azure models? That would solve the issue.

…

On Thu, Feb 6, 2025, 2:50 PM Enrico Ros ***@***.***> wrote: Reopening because evidently there's an issue with Azure models. Note: the complexity with azure it's always been the dynamic listing, as the names of deployments are likely non standard and one cannot guess the context window sizes, model capabilities, etc. The code right now is looking up the official OpenAI model names, but in this particular instance there could be other underlying issues. — Reply to this email directly, view it on GitHub <#744 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADEPZYVTHJB5YRFJRLSVJJT2ONSCRAVCNFSM6AAAAABWDHMVH6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMZZHA4DIOJRHE> . You are receiving this because you commented.Message ID: ***@***.***>

[enricoros]

@rafaelpierre it does, but everything else is auto detected. I welcome any pull requests as I don't have azure deployments so it would take me longer to test. 1. For the missing o1, I'll take a look. 2. For the missing UI, I welcome anyone's PR. Note: one must configure also the context window, and capabilities (image input, function calling, etc.) and QUIRKS of the model. For instance, o1 can't stream, and the user will need to say that (again, no auto detect with azure models) otherwise the LLM call will just fail (and note that O1 and P1 preview and O3 have all different quirks)