Configuration of librechat.yaml for Azure OpenAI

[apamildner]

What happened?

When using a custom key for models( see example), the chat fails with: "Something went wrong. Here's the specific error message we encountered: An error occurred while processing your request. Please contact the Admin."

Version Information

v.0.7.8

Steps to Reproduce

This doesn't work:

version: 1.2.1
endpoints:
  azureOpenAI:
    plugins: true
    summarize: true
    titleConvo: true
    groups:
      - group: "some-grp"
        apiKey: ${OPENAI_API_KEY}
        instanceName: some-instance-name
        version: 2025-01-01-preview
        models:
          foo: # key differs from deploymentName
            deploymentName: "gpt-4.1"

But this works (notice that the name gpt-4.1 is same for the key and the deploymentName under models)

version: 1.2.1
endpoints:
  azureOpenAI:
    plugins: true
    summarize: true
    titleConvo: true
    groups:
      - group: "some-grp"
        apiKey: ${OPENAI_API_KEY}
        instanceName: some-instance-name
        version: 2025-01-01-preview
        models:
          gpt-4.1: # Here we use the same name as deploymentName
            deploymentName: "gpt-4.1"

What browsers are you seeing the problem on?

No response

Relevant log output

error: [handleAbortError] AI response error; aborting request: Failed to send message. HTTP 400 - Unsupported data type

Screenshots

No response

Code of Conduct

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

[danny-avila]

Thanks for the report but this is not a bug—it's expected behavior.

• In your config, the key under models: (the "model identifier") must match an actual LLM model name or identifier (like gpt-4.1).
• If you use a random key (like foo:), LibreChat can't recognize the model, resulting in errors.
• This matches what's stated in the docs:

  The model identifier must match its corresponding OpenAI model name.

Fix:
Change your model config to:

models:
  gpt-4.1:
    deploymentName: "gpt-4.1"

Arbitrary keys are not meant to be used under models:. The motivation behind this is to allow cases where the deployment name is not a valid LLM model name or identifier (like gpt-4.1).

Do you have a specific use case for customizing this key? As long as it's a partial match to a system-known model, it will work as expected.

endpoints:
  azureOpenAI:
    streamRate: 35
    titleModel: "gpt-4o"
    titleConvo: true
    groups:
    - group: "my-group-name"
      apiKey: "${MY_API_KEY_REFERENCE}"
      instanceName: "my-instance-name"
      version: "2024-12-01-preview"
      models:
        gpt-4.1:
          deploymentName: "gpt-4.1"

[apamildner]

I see! I think I just misunderstood it. My intuition was just that the names under models would be arbitrary, because then I could call them what I want when the user opens the drop down to select models. The model used is already decided by the deploymentName that I chose. So if I set this up for my customers, I could instead call the model for example recommended and I know that the model is gpt-4.1, but the user doesn't need to know:

      models:
        recommended:
          deploymentName: "gpt-4.1"