Unable to change default model for Gemini Integration

[kevhardy]

Checklist

• I'm running the newest version of LLM Vision https://github.com/valentinfrlch/ha-llmvision/releases/latest
• I have enabled debug logging for the integration.
• I have filled out the issue template to the best of my ability.
• This issue only contains 1 issue (if you have multiple issues, open one issue for each issue).
• This is a bug and not a feature request.
• I have searched open issues for my problem.

Describe the issue

I am unable to change the default model to use when configuring the Gemini Integration. This also is preventing me from updating the API key.

It appears that the POST request used to test the API key is using the gemini-2.0-flash model. Since I have a free version of billing on this account, this model is not available and returns an exceeded quota error. However, this key should be able to use other models I am attempting to put in the default model field.

Reproduction steps

1. Configure Google Gemini Integration
2. Input API key from a free billing account
3. Input a model available to free usage tier (e.g. gemini-2.5-flash)
4. Hit Submit and receive Invalid API key error (default field goes back to gemini-2.0-flash after error)

Debug logs

2025-12-18 23:41:16.330 DEBUG (MainThread) [custom_components.llmvision.providers] Provider initialized: Google(model=gemini-2.5-flash, endpoint={'base_url': 'https://generativelanguage.googleapis.com/v1beta/models/{model}:generateContent?key={api_key}'})
2025-12-18 23:41:16.330 DEBUG (MainThread) [custom_components.llmvision.providers] Request data: {'contents': [{'role': 'user', 'parts': [{'text': 'Hi'}]}], 'generationConfig': {'maxOutputTokens': 1, 'temperature': 0.5}}
2025-12-18 23:41:16.330 DEBUG (MainThread) [custom_components.llmvision.providers] Posting to https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent

...

2025-12-18 23:41:16.487 ERROR (MainThread) [custom_components.llmvision.config_flow] Validation failed: You exceeded your current quota, please check your plan and billing details. For more information on this error, head to: https://ai.google.dev/gemini-api/docs/rate-limits. To monitor your current usage, head to: https://ai.dev/usage?tab=rate-limit. 
* Quota exceeded for metric: generativelanguage.googleapis.com/generate_content_free_tier_input_token_count, limit: 0, model: gemini-2.0-flash
* Quota exceeded for metric: generativelanguage.googleapis.com/generate_content_free_tier_requests, limit: 0, model: gemini-2.0-flash
* Quota exceeded for metric: generativelanguage.googleapis.com/generate_content_free_tier_requests, limit: 0, model: gemini-2.0-flash
Please retry in 43.569762747s.

[Eischer]

I have exactly the same Problem. Tried to create a reshly new Gemini API Key but always get "Invalid key"

EDIT: Sorry after changing the prefilled "gemini-2.0-flash" to "gemini-2.5-flash" dem Api Key works

[miketahu]

@kevhardy
This looks like a streaming transport issue rather than a model failure. When responses pause between chunks, default proxy or client read timeouts can fire even though the upstream is still healthy. Check whether buffering is disabled and whether timeouts are per-chunk vs total-request. HTTP/2 stream resets can also surface as intermittent disconnects if not handled explicitly. To isolate whether this is client-side or upstream behavior, it can help to replay the same request against a schema-accurate OpenAI-compatible gateway (e.g., AI Badgr) and compare streaming behavior.

[TheRealFalseReality]

I cannot add a Gemini config anymore at all, get "Invalid API key" with newly generated keys

[kevhardy]

@kevhardy This looks like a streaming transport issue rather than a model failure. When responses pause between chunks, default proxy or client read timeouts can fire even though the upstream is still healthy. Check whether buffering is disabled and whether timeouts are per-chunk vs total-request. HTTP/2 stream resets can also surface as intermittent disconnects if not handled explicitly. To isolate whether this is client-side or upstream behavior, it can help to replay the same request against a schema-accurate OpenAI-compatible gateway (e.g., AI Badgr) and compare streaming behavior.

I'm not seeing any timeout related errors in the logs. In the snippet of logs I posted, I am directly receiving a message that it's hitting a rate that set at a limit of 0. I'm not sure what you are seeing in my logs that seem to indicate that could be the issue? I have since switched to using Groq's free tier and its working perfectly fine.

I'm assuming this is because the default gemini-2.0-flash is no longer on the free tier. However even if i switch the default model to one that is - gemini-2.5-flash - I am still seeing that its trying to test the API key using gemini-2.0-flash and receiving the same rate limit response.

Here are two sample requests I did to each model using the same free tier API key. The 2.5 returns a valid response, the 2.0 is immediately "rate-limited".

[dverleysen]

Tried with a new installation, also get the message "Invalid API key" when trying to setup a connection with Google. (Home Assistant).

When trying curl command via the terminal I get a correct answer from Google.

➜  ~ curl "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent?key=XXXXXXXXXXX" \
    -H "Content-Type: application/json" \
    -X POST \
    -d '{"contents": [{"parts":[{"text": "Say hello"}]}]}'

{
  "candidates": [
    {
      "content": {
        "parts": [
          {
            "text": "Hello!"
          }
        ],
        "role": "model"
      },
      "finishReason": "STOP",
      "index": 0
    }
  ],
  "usageMetadata": {
    "promptTokenCount": 2,
    "candidatesTokenCount": 2,
    "totalTokenCount": 21,
    "promptTokensDetails": [
      {
        "modality": "TEXT",
        "tokenCount": 2
      }
    ],
    "thoughtsTokenCount": 17
  },
  "modelVersion": "gemini-2.5-flash",
  "responseId": "XXXXXXXXXXX"
}

[kevhardy]

Tried with a new installation, also get the message "Invalid API key" when trying to setup a connection with Google. (Home Assistant).

When trying curl command via the terminal I get a correct answer from Google.

@dverleysen Are you using an API key on a free tier billing plan? If so, can you try enabling debug logging and trying again? I'm curious if you set the default model to gemini-2.5-flash is it still attempting to validate the key against gemini-2.0-flash and what response you get.

[github-actions]

This issue has been automatically marked as stale due to inactivity.