Serverless deployments running into error with 7.9-rc1 version

[mansiibm]

What happened?

Hi Danny, for the version 7.9-rc1 and 62b4f3b I tried deploying the serverless models from the same azure foundry, so naturally all the models share the same base URL and the same API key. My config is as follows-

endpoints:
custom:
- name: "Deepseek"
apiKey: "${DEEPSEEK_API_KEY}"
baseURL: "https://regionname.api.cognitive.microsoft.com/models"
models:
default: ["DeepSeek-R1"]
fetch: false
titleConvo: true
directEndpoint: false
titleModel: "DeepSeek-V3-0324"
modelDisplayLabel: "Deepseek"
- name: "xai"
apiKey: "${XAI_API_KEY}"
baseURL: "https://regionname.api.cognitive.microsoft.com/models"
models:
default: ["grok-3", "grok-3-mini"]
fetch: false
titleConvo: true
directEndpoint: false
titleModel: "grok-3"
modelDisplayLabel: "xAI"

so the deepseek and xai api keys I have defined separately, have the same value as the models in the azure foundry have the same base url and same key. Deepseek works perfectly fine, xai runs in incorrect api key but the key is correct and was working until this version.

I had even kept the complete URL as -

custom:
- name: "Deepseek"
apiKey: "${DEEPSEEK_API_KEY}"
baseURL: "https://region.api.cognitive.microsoft.com/models/chat/completions?api-version=2024-05-01-preview"
models:
default: ["DeepSeek-R1"]
fetch: false
titleConvo: true
directEndpoint: true
titleModel: "DeepSeek-V3-0324"
modelDisplayLabel: "Deepseek"
- name: "xai"
apiKey: "${XAI_API_KEY}"
baseURL: "https://region.api.cognitive.microsoft.com/models/chat/completions?api-version=2024-05-01-preview"
models:
default: ["grok-3","grok-3-mini"]
fetch: false
titleConvo: true
directEndpoint: true
titleModel: "grok-3"
modelDisplayLabel: "xAI"

I still face the same issue-

The fun part is that the same models as serverless endpoints are performing well in the azure ai endpoint

• group: "server-less models"
  apiKey: "${DEEPSEEK_API_KEY}" /// even if i keep the key deepseek or xai its the same and all the models work
  baseURL: "https://regionname.api.cognitive.microsoft.com/models/"
  version: "2024-05-01-preview"
  serverless: true
  models:
  grok-3: true
  grok-3-mini: true
  DeepSeek-R1: true

Can you please suggest something here............thanks!!!!!

Version Information

Using the LC version- 62b4f3b

Steps to Reproduce

1. Add the xai and deepseek models using the config above
2. The api key and the base url are same for both the providers
3. Try chatting with the models, you will run into incorrect api key error.

What browsers are you seeing the problem on?

Microsoft Edge

Relevant log output

2025-07-18T04:15:20.298958796Z {"conversationId":"fab253b1-420f-4a8e-99cc-945ef496d0da","isCreatedByUser":true,"level":"debug","message":"[BaseClient] userMessage","messageId":"3a9a5225-f040-4932-a746-7eb0e736251e","parentMessageId":"00000000-0000-0000-0000-000000000000","sender":"User","text":"hi","timestamp":"2025-07-18T04:15:20.298Z","tokenCount":5}
2025-07-18T04:15:20.445496990Z {"attemptNumber":1,"error":"Incorrect API key provided: Eq***Oy. You can obtain an API key from https://console.x.ai.","headers":{},"level":"error","message":"[api/server/controllers/agents/client.js #sendCompletion] Operation aborted 400 \"Incorrect API key provided: Eq***Oy. You can obtain an API key from https://console.x.ai.\"","pregelTaskId":"80fac82b-d2e3-533c-baac-2048f4d894b7","requestID":null,"retriesLeft":6,"stack":"Error: 400 \"Incorrect API key provided: Eq***Oy. You can obtain an API key from https://console.x.ai.\"\n    at APIError.generate (/app/node_modules/@librechat/agents/node_modules/openai/core/error.js:45:20)\n    at CustomOpenAIClient.makeStatusError (/app/n...","status":400,"timestamp":"2025-07-18T04:15:20.445Z"}
2025-07-18T04:15:20.445664403Z {"attemptNumber":1,"error":"Incorrect API key provided: Eq***Oy. You can obtain an API key from https://console.x.ai.","headers":{},"level":"error","message":"[api/server/controllers/agents/client.js #sendCompletion] Unhandled error type 400 \"Incorrect API key provided: Eq***Oy. You can obtain an API key from https://console.x.ai.\"","pregelTaskId":"80fac82b-d2e3-533c-baac-2048f4d894b7","requestID":null,"retriesLeft":6,"stack":"Error: 400 \"Incorrect API key provided: Eq***Oy. You can obtain an API key from https://console.x.ai.\"\n    at APIError.generate (/app/node_modules/@librechat/agents/node_modules/openai/core/error.js:45:20)\n    at CustomOpenAIClient.makeStatusError (/app/n...","status":400,"timestamp":"2025-07-18T04:15:20.445Z"}



on the UI- 

Something went wrong. Here's the specific error message we encountered: An error occurred while processing the request: 400 "Incorrect API key provided: Eq***Oy. You can obtain an API key from https://console.x.ai."

Screenshots

No response

Code of Conduct

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

[mansiibm]

Hi Danny, this I found after some experimentation.

If the name of the model is changed -

like instead of grok-3 I use- grok-3-version-1 then also error persists.

If the endpoint is changed for example -

- name: "x-AI"
  apiKey: "${XAI_API_KEY}"
  baseURL: "https://region.api.cognitive.microsoft.com/models"
  models:
    default: ["grok-3"]
    fetch: false
  titleConvo: true
  directEndpoint: false
  titleModel: "grok-3"
  modelDisplayLabel: "x-AI"

then the grok-3 model works absolutely fine.

So just by changing the endpoint name the models from xai in azure foundry started working, can you please fix this as before 7.9-rc1, the grok models from xai used to work normally.

The config for this set up (the one above working is)-

endpoints:
custom:
- name: "Deepseek"
apiKey: "${DEEPSEEK_API_KEY}"
baseURL: "https://region.api.cognitive.microsoft.com/models/chat/completions?api-version=2024-05-01-preview"
models:
default: ["DeepSeek-R1"]
fetch: false
titleConvo: true
directEndpoint: true
titleModel: "DeepSeek-V3-0324"
modelDisplayLabel: "Deepseek"
- name: "x-AI"
apiKey: "${XAI_API_KEY}"
baseURL: "https://region.api.cognitive.microsoft.com/models/chat/completions?api-version=2024-05-01-preview"
models:
default: ["grok-3","grok-3-mini"]
fetch: false
titleConvo: true
directEndpoint: true
titleModel: "grok-3"
modelDisplayLabel: "x-AI"

Thanks !!

[mansiibm]

As the grok models are not working with the name of the custom endpoint- xai, and x-AI works as shared above, the icon is also not rendered.