Azure open ai responses api - text not being displayed

[MTR3K]

What happened?

@danny-avila, thank you guys! Configured deployments in librechat.yaml and params to use responses api. No errors. Also azure OpenAI studio monitoring indicates that request from librechat was processed and response returned. However the models answers are empty (in UI and in db).

Am I missing something?

Regards

Version Information

0.7.9-rc1

Steps to Reproduce

Configure azure open ai responses model in librechat.yaml
Select model and switch on responses api toggle
Wait for empty response
Check db messages and find empty text property

What browsers are you seeing the problem on?

No response

Relevant log output

See screenshots

Screenshots

Code of Conduct

• I agree to follow this project's Code of Conduct

[danny-avila]

No issue here:

Possibly related to codex-mini try another model?

[danny-avila]

Confirming with o3-pro, which only works with Responses API:

[MTR3K]

thanks for the quick reply. Strange. I tried with codex-mini and o3-pro. Could you share your yaml config? I am using GitHub codespaces but I don’t suspect this to be of any importance.

[danny-avila]

this is my azure config:

endpoints:
  azureOpenAI:
    streamRate: 35
    titleModel: "gpt-4o"
    titleConvo: true
    groups:
    - group: "region-eastus"
      apiKey: "${EASTUS_API_KEY}"
      instanceName: "eastus-instance"
      version: "2024-02-15-preview"
      models:
        gpt-4o:
          deploymentName: gpt-4o
    - group: "region-eastus2"
      apiKey: "${EASTUS2_API_KEY}"
      instanceName: "eastus2-instance"
      version: "2024-12-01-preview"
      models:
        gpt-4o-latest:
          deploymentName: "gpt-4o-mini"
          version: "2024-02-15-preview"
        o1-mini:
          deploymentName: "o1-mini"
        o1-preview:
          deploymentName: "o1-preview"
        o3-mini:
          deploymentName: "o3-mini"
        o4-mini-eus2:
          deploymentName: "o4-mini-eus2"
        gpt-4.1:
          deploymentName: "gpt-4.1"
        gpt-4.1-mini:
          deploymentName: "gpt-4.1-mini-eus2"
        gpt-4.1-nano:
          deploymentName: "gpt-4.1-nano"
        o3-pro:
          deploymentName: "o3-pro"
          version: "2025-04-01-preview"
    - group: "codex-mini"
      apiKey: "${AZURE_OAI_API_KEY}"
      baseURL: "https://SOME_RESOURCE.cognitiveservices.azure.com/openai/"
      version: "2025-04-01-preview"
      serverless: true
      models:
        codex-mini:
          deploymentName: "codex-mini"

Note I configured codex-mini as a serverless endpoint, since it seems it only works this way.

In this case, the preview API version is not accepted in the request, as detailed in their docs:
https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/responses?tabs=python-secure#api-support

This requires a small fix for codex-mini specifically: #8316

[MTR3K]

@danny-avila thank you for the feedback. Built an Image and deployed it. Its working as intended. So I am guessing it has to do with local npm builds? Or some weird behavior on codespaces.

Regards
Max

[cevatkerim]

It is still broken for me. It seems to not work if the reasoning mode kicks in. For trivial messages like "tell me a joke" it works fine. If you ask it something that will trigger extended thinking, then it fails.