API call failed for model antigravity/gemini-2.5-flash-thinking and antigravity/gemini-2.5-flash

[techroy23]

Whats the correct ID for antigravity gemini 2.5 Flash?

[ERROR]

15:09 - 127.0.0.1:53736 - provider: antigravity, model: gemini-2.5-flash - N/A
INFO:     127.0.0.1:53736 - "POST /v1/chat/completions HTTP/1.1" 200 OK
Acquiring credential for model antigravity/gemini-2.5-flash. Tried credentials: 0/16(16)
Acquired key antigravity_oauth_10.json for model antigravity/gemini-2.5-flash (tier: free-tier, priority: 3, selection: sequential, quota: 0/3000 [100%])
Attempting stream with credential antigravity_oauth_10.json (Attempt 1/2)
Stream connection established for credential antigravity_oauth_10.json. Processing response.
15:09 - 127.0.0.1:53747 - provider: gemini_cli, model: gemini-3-flash-preview - N/A
INFO:     127.0.0.1:53747 - "POST /v1/chat/completions HTTP/1.1" 200 OK
Acquiring credential for model gemini_cli/gemini-3-flash-preview. Tried credentials: 0/16(16)
Acquired key gemini_cli_oauth_1.json for model gemini_cli/gemini-3-flash-preview (tier: free-tier, priority: 3, selection: sequential, quota: 33/1000 [96%])
Attempting stream with credential gemini_cli_oauth_1.json (Attempt 1/2)
Stream connection established for credential gemini_cli_oauth_1.json. Processing response.
Caught a critical API error mid-stream: HTTPStatusError. Signaling for credential rotation.
Released credential antigravity_oauth_10.json from model antigravity/gemini-2.5-flash (remaining concurrent: 0)
STREAM FINISHED and lock released for credential antigravity_oauth_10.json.
API call failed for model antigravity/gemini-2.5-flash with key antigravity_oauth_10.json. Error: StreamedAPIError. See failures.log for details.
Non-recoverable error (invalid_request) during custom stream. Failing.
An unhandled exception occurred in streaming retry logic: Provider error received in stream
Traceback (most recent call last):
  File "rotator_library\client.py", line 938, in _safe_streaming_wrapper
  File "rotator_library\providers\antigravity_provider.py", line 4602, in _streaming_with_retry
  File "rotator_library\providers\antigravity_provider.py", line 4444, in _handle_streaming
  File "httpx\_models.py", line 829, in raise_for_status
httpx.HTTPStatusError: Client error '400 Bad Request' for url 'https://daily-cloudcode-pa.sandbox.googleapis.com/v1internal:streamGenerateContent?alt=sse'
For more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/400

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "rotator_library\client.py", line 2275, in _streaming_acompletion_with_retry
  File "rotator_library\client.py", line 2238, in _streaming_acompletion_with_retry
  File "rotator_library\client.py", line 1141, in _transaction_logging_stream_wrapper
  File "rotator_library\client.py", line 1031, in _safe_streaming_wrapper
rotator_library.client.StreamedAPIError: Provider error received in stream
Recorded usage from response object for key gemini_cli_oauth_1.json
Released credential gemini_cli_oauth_1.json from model gemini_cli/gemini-3-flash-preview (remaining concurrent: 0)
STREAM FINISHED and lock released for credential gemini_cli_oauth_1.json.

opencode.json

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "llm-proxy": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "An2Kin-Proxy",
      "options": {
        "baseURL": "http://127.0.0.1:8888/v1",
        "apiKey": "myapikey",
        "timeout": 300000
      },
      "models": {
	"AntiGravity-gemini-2.5-flash": {
        "id": "antigravity/gemini-2.5-flash",
        "name": "XX.) antigravity gemini-2.5-flash (an2kin)",
        "limit": { "context": 1048576, "output": 64000 },
        "modalities": { "input": ["text"], "output": ["text"] }
        }
      }
    }
  }
}

[mirrobot-agent]

@techroy23 Thank you for submitting this issue. I am now beginning my analysis and will report back shortly.

[mirrobot-agent]

Initial Analysis Report

Summary: The API call failure (400 Bad Request) for antigravity/gemini-2.5-flash is likely due to an incorrect internal model ID being sent to the Antigravity API or an invalid thinkingBudget parameter configuration.

Issue Validation: Confirmed. The logs clearly show a 400 Bad Request from the Google internal API endpoint (daily-cloudcode-pa.sandbox.googleapis.com) when attempting to call gemini-2.5-flash.

Reproducibility Assessment: Reproducible (for users with Antigravity credentials).

Root Cause Analysis:

1. Model ID Mismatch: The proxy sends "model": "gemini-2.5-flash" in the request payload. However, the Antigravity backend might expect gemini-2.0-flash (which is supported in gemini_cli and mentioned in docs but missing from antigravity_provider.py) or gemini-2.5-flash-thinking exclusively.
2. Invalid Thinking Budget: For gemini-2.5-flash, the provider automatically injects thinkingBudget: -1 (auto) into the generationConfig. Some internal Google API variants are strict and may reject negative values or the thinkingConfig field entirely if the model ID doesn't explicitly support it.
3. Excluded Models: gemini-2.5-flash-thinking is explicitly listed in EXCLUDED_MODELS in antigravity_provider.py (line 220), which prevents it from being discovered dynamically, though it can still be called manually.

Suggested Labels: bug, provider:antigravity, models

Proposed Next Steps:

1. Try using the model ID antigravity/gemini-2.0-flash if available.
2. Try using antigravity/gemini-3-flash which is officially supported in the latest version.
3. If using gemini-2.5-flash, try specifying an explicit reasoning_effort (e.g., "low", "medium", "high") to avoid the thinkingBudget: -1 default.
4. Maintainers should verify the mapping in MODEL_ALIAS_MAP within src/rotator_library/providers/antigravity_provider.py and consider adding gemini-2.0-flash.

Missing Information (if any):

• Could you please try calling antigravity/gemini-2.0-flash and let us know if it works?
• Also, try specifying "reasoning_effort": "low" in your request to see if it bypasses the 400 error.

This analysis was generated by an AI assistant.