IssueTracker

Problem you have encountered:

When testing Vertex AI Express Mode with an invalid API Key (specifically, in this case, an API Key from AI Studio), I get the following error object:

    {
      "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.GenerateContent",
              "service": "aiplatform.googleapis.com"
            }
          }
        ]
      }
    }

The error code 401 isn't too surprising, tho disappointing. The message about API keys not being supported by this API is, since this is the entire purpose of Express Mode.

I tested this against the Express Mode endpoint (https://aiplatform.googleapis.com/v1/publishers/google/models/gemini-1.5-pro-002:generateContent) and my test code works correctly when it has a valid Express Mode API Key.

What you expected to happen:

An error message that indicated it was an invalid API Key.

Steps to reproduce:

1. Get a valid API Key from https://aistudio.google.com/app/apikey
2. Place the key in the appropriate place in the following UNIX shell script.
3. Run the script

#!/bin/bash

API_KEY=AIstudioKeyGoesHere
model=gemini-1.5-flash-002
method=generateContent

URL="https://aiplatform.googleapis.com/v1/publishers/google/models/${model}:${method}"

curl \
  -X POST "${URL}" \
  -H 'Content-Type: application/json' \
  -H "x-goog-api-key: ${API_KEY}" \
  -d @<(echo '{
  "contents": [
    {
      "role": "user",
      "parts": [
        {
          "text": "What is the answer to life the universe and everyitng? Answer briefly."
        }
      ]
    }
  ]
}')

Other information (workarounds you have tried, documentation consulted, etc):

Interestingly, if the API Key given is completely invalid (ie - something like an API Key of "bogus"), you get a different error message about the request containing an invalid argument.