AI API receipt · vllm-project/vllm

/v1/completions returns HTTP 500 for use_beam_search + echo + logprobs

vllm-project/vllm#60875 · upstream bug reproduced by running the issue's own reproduction

✓ Badgr rejects this response

use_beam_search + echo + logprobs(1) on facebook/opt-125m answered 500 with an empty error message (an unhandled AssertionError in beam search's prompt-logprobs path) instead of a 400 naming the conflicting parameters.

Outcome contract verdict: response is empty. The attempt is recorded as failed (invalid_outcome) instead of a verified success, so it is not billed as a good answer and can fall back to another route.

What the upstream returned (replayed)

The output the bug produces, taken from the issue’s reproduction and fed to Badgr’s validators. Not a live provider call.

text: none
tool_calls: none

Badgr receipt

Written by Badgr’s receipt ledger and read back from GET /v1/receipts/replay-vllm-60875-beam-search-echo-logprobs-500.

{
  "request_id": "replay-vllm-60875-beam-search-echo-logprobs-500",
  "status": "failed",
  "endpoint": "/v1/completions",
  "stream": false,
  "failure_stage": "outcome_contract",
  "error_code": "invalid_outcome",
  "error_message": "Response did not satisfy the outcome contract: response is empty",
  "retryable": true,
  "provider_mode": "replay",
  "routing_decision": "replayed_upstream_output",
  "created_at": "1791632037.517389"
}

Recorded 2026-10-10 in Badgr’s local development environment. Replay cases and verdicts are pinned by backend/test_ai_api_replay_cases.py.

← All AI API receipts