AI API receipt · BerriAI/litellm
BerriAI/litellm#45546 · upstream bug reproduced by running the issue's own reproduction
Badgr cannot tell this response is wrong
The outgoing assistant turn had no reasoning, so the model never saw its own earlier thinking in a multi-turn conversation.
the reply is well-formed; the loss is in the request sent to the provider, which the response check never sees. 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: "I will call get_number again." tool_calls: none
Written by Badgr’s receipt ledger and read back from GET /v1/receipts/replay-litellm-45546-mistral-reasoning-not-replayed.
{
"request_id": "replay-litellm-45546-mistral-reasoning-not-replayed",
"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": "1791542228.515253"
}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