AI API receipt · BerriAI/litellm
BerriAI/litellm#45357 · upstream bug reproduced by running the issue's own reproduction
Badgr cannot tell this response is wrong
The transcript came back correct; the gateway's own spend record was computed from a bogus duration of about 5.8e14 seconds.
the response is correct; the wrong number is the gateway's spend accounting, which is not part of the response. The receipt therefore records success. This is a limit of what the contract can prove, shown here so it is not mistaken for protection.
The output the bug produces, taken from the issue’s reproduction and fed to Badgr’s validators. Not a live provider call.
text: "Welcome to the weekly planning meeting." tool_calls: none
Written by Badgr’s receipt ledger and read back from GET /v1/receipts/replay-litellm-45357-flac-transcription-cost-explodes.
{
"request_id": "replay-litellm-45357-flac-transcription-cost-explodes",
"status": "success",
"endpoint": "/v1/chat/completions",
"stream": false,
"failure_stage": null,
"error_code": null,
"error_message": null,
"retryable": null,
"provider_mode": "replay",
"routing_decision": "replayed_upstream_output",
"created_at": "1791540569.131029"
}Recorded 2026-10-09 in Badgr’s local development environment. Replay cases and verdicts are pinned by backend/test_ai_api_replay_cases.py.
← All AI API receipts