Gemini not working over Google Vertex AI

[dannykorpan]

Hi community,

I’ve attempted to set up Gemini using Google Vertex AI, following this setup guide: https://www.librechat.ai/docs/configuration/pre_configured_ai/google

All required Google APIs are enabled, the IAM service account has been created, and I’ve exported the auth.json file. I also updated my docker-compose.yml to mount the auth.json file into the LibreChat API container.

.env file:

...
ENDPOINTS=azureOpenAI,bedrock,google,azureAssistants,agents
...
#GOOGLE_KEY=
GOOGLE_MODELS=gemini-2.5-pro-preview-03-25,gemini-2.0-flash
GOOGLE_TITLE_MODEL=gemini-2.0-flash-lite
GOOGLE_SAFETY_SEXUALLY_EXPLICIT=BLOCK_ONLY_HIGH
GOOGLE_SAFETY_HATE_SPEECH=BLOCK_ONLY_HIGH
GOOGLE_SAFETY_HARASSMENT=BLOCK_ONLY_HIGH
GOOGLE_SAFETY_DANGEROUS_CONTENT=BLOCK_ONLY_HIGH
...

docker-compose.override.yml:

version: '3.4'
 
services:
  api:
    volumes:
    - type: bind
      source: ./auth.json
      target: /app/api/data/auth.json

docker exec -it LibreChat cat /app/api/data/auth.json

{
  "type": "service_account",
  "project_id": "delta-deck-23344234",
  "private_key_id": "MY PRIVATE KEY IS",
  "private_key": "-----BEGIN PRIVATE KEY-----
XYZ
-----END PRIVATE KEY-----\n",
  "client_email": "[email protected]",
  "client_id": "343434324324",
  "auth_uri": "https://accounts.google.com/o/oauth2/auth",
  "token_uri": "https://oauth2.googleapis.com/token",
  "auth_provider_x509_cert_url": "https://www.googleapis.com/oauth2/v1/certs",
  "client_x509_cert_url": "https://www.googleapis.com/robot/v1/metadata/x509/dsfdsf%40delta-deck-sdfsdf.iam.gserviceaccount.com",
  "universe_domain": "googleapis.com"
}

However, whenever I try to use Gemini, I always encounter the following error.

2025-04-15 07:55:08 error: [GoogleClient] There was an issue generating the completion Google request failed with status code 401: [{
  "error": {
    "code": 401,
    "message": "API keys are not supported by this API. Expected OAuth2 access token or other authentication credentials that assert a principal. See https://cloud.google.com/docs/authentication",
    "status": "UNAUTHENTICATED",
    "details": [
      {
        "@type": "type.googleapis.com/google.rpc.ErrorInfo",
        "reason": "CREDENTIALS_MISSING",
        "domain": "googleapis.com",
        "metadata": {
          "method": "google.cloud.aiplatform.v1.PredictionService.StreamGenerateContent",
          "service": "aiplatform.googleapis.com"
        }
      }
    ]
  }
}
]
2025-04-15 07:55:08 error: [AskController] Error handling request { "type": "undefined", "info": "Google request failed with status code 401: [{
  "error": {
    "code": 401,
    "message": "API keys are not supported by this API. Expected OAuth2 access token or other authentication credentials that assert a principal. See https://cloud.google.com/docs/authentication",
    "status": "UNAUTHENTICATED",
    "details": [
      {
        "@type": "type.googleapis.com/google.rpc.ErrorInfo",
        "reason": "CREDENTIALS_MISSING",
        "domain": "googleapis.com",
        "metadata": {
          "method": "google.cloud.aiplatform.v1.PredictionService.StreamGenerateContent",
          "service": "aiplatform.googleapis.com"
        }
      }
    ]
  }
}
]" }
2025-04-15 07:55:08 error: [handleAbortError] AI response error; aborting request: { "type": "undefined", "info": "Google request failed with status code 401: [{
  "error": {
    "code": 401,
    "message": "API keys are not supported by this API. Expected OAuth2 access token or other authentication credentials that assert a principal. See https://cloud.google.com/docs/authentication",
    "status": "UNAUTHENTICATED",
    "details": [
      {
        "@type": "type.googleapis.com/google.rpc.ErrorInfo",
        "reason": "CREDENTIALS_MISSING",
        "domain": "googleapis.com",
        "metadata": {
          "method": "google.cloud.aiplatform.v1.PredictionService.StreamGenerateContent",
          "service": "aiplatform.googleapis.com"
        }
      }
    ]
  }
}
]" }

And here is a screenshot from LibreChat

I would be very grateful for any advice or help.

Kind Regards,
Danny

[alexhermida]

Hi Danny,

Not an expert but it has been working fine to me, I'd suggest just initially simplifying the env variables. From the .env.example I commented the # GOOGLE_KEY=user_provided and left everything as it was, then added the auth.json and updated the override.

Let's just comment everything you have, except the endpoints, since no conf it's needed as Vertex config relays only in the auth.json file. From your tests it looks like it's ok.

I'm wondering if any of those variables affects your config

[dannykorpan]

Thank you for your reply. Already did this, but still the same error.

[danny-avila]

make sure GOOGLE_API_KEY is not set

[dannykorpan]

Thank you, now it's working. GOOGLE_API_KEY was specified! Sorry for that.

How do I specify the server location, e.g. Frankfurt, Deutschland (europe-west3)?
https://firebase.google.com/docs/vertex-ai/locations?hl=de&platform=web#europe

[Denin-Siby]

Hi @dannykorpan ,

I was able to get the connection through, however I was facing issues about the agent's voice just getting cut-off before it finishes it's speech, was curious to know if it all works as expected for you?

Also, just to confirm, to use it with vertex ai you use the v1beta1 api-version of their's right?

[dannykorpan]

I’m only using Vertex AI for text generation. Where can I see which API version I’m using?