Restaurant Finder Demo: Gemini API Quota "0" Error in Europe-West1 (Turkey)

[EnderKaran]

Hi ,

I am trying to run the Restaurant Finder agent sample locally following the instructions in the README. However, I am encountering a persistent RESOURCE_EXHAUSTED error, specifically related to regional quotas, even though I created a fresh API Key from Google AI Studio.

Environment:

• OS: Windows
• Region: Turkey (Europe-West1)
• Model: Tried with default (gemini-2.5-flash) and gemini-1.5-flash.

The Issue: When running uv run . for the agent, the request fails with a 429 status code. The error details indicate that the quota limit value is "0" for this region.

Error Log snippet:

"error": {
    "code": 429,
    "message": "Quota exceeded for quota metric 'Generate Content API requests per minute' ...",
    "status": "RESOURCE_EXHAUSTED",
    "details": [
      {
        "reason": "RATE_LIMIT_EXCEEDED",
        "metadata": {
          "quota_location": "europe-west1",
          "quota_limit_value": "0"
        }
      }
    ]
}

Steps Taken:
Created a new API key via Google AI Studio.
Exported the key in the terminal.
Tried changing the model in agent.py to gemini-1.5-flash, but the issue persists with the same "quota_limit_value: 0" error.

[michaelbrinkworth]

@EnderKaran The symptoms here line up with rate-limit handling rather than request correctness. If 429s are being retried aggressively without backoff, the client can end up in a retry loop that never recovers. In production, retries should be gated to retryable errors only and combined with jittered backoff. Some teams also add an overflow path so traffic can fail fast or route elsewhere when a provider hard-limits requests. An OpenAI-compatible gateway (such as AI Badgr) is sometimes used as a reference point to verify whether the issue is provider throttling or client retry behavior.