20:08:32 start (recording, release meetings)
20:08:32 fetch: http://172.18.0.1:8790/r/abendessen-2026-09-17/audio.mp3
20:08:32 convert: 0.37 h source, using 0.37 h
20:08:49 vad: 9 windows, continuous, 21 min speech
20:08:49 diar: pyannote diarize, naming by ReDimNet
20:08:54 diar: pyannote job 91cd44f2-0315-4ae6-adae-bf3a1068739b
20:08:54 asr: 9 windows, 9 to do
20:08:56 FAILED at asr: HTTP 400: {"error":{"message":"Access denied, please make sure your account is in good standing. For details, see: https://www.alibabacloud.com/help/en/model-studio/error-code#overdue-payment","type":"Arrearage","param":null,"code":"Arrearage"},"id":"chatcmpl-41ad954e-9196-931d-af55-cbe74cd7d72c","request_id"
20:13:46 retry: resuming asr with its partial results
20:13:46 start (recording, release meetings)
20:13:46 asr: 9 windows, 9 to do
20:14:07 preview: eval-abendessen-v4-baseline-2026-09-17 (transcript only, speakers follow)
20:14:25 asr: 421 lines
20:14:25 diar: pyannote diarize, naming by ReDimNet
20:14:29 diar: pyannote job 345f4180-72c7-4b48-937c-4a64ed839a1a
20:15:15 identify: merged over-split clusters SPEAKER_01→SPEAKER_00, SPEAKER_02→SPEAKER_00
20:15:15 Traceback (most recent call last):
  File "/opt/meetings/app/worker.py", line 138, in run_job
    ctx.check(); fns[stage](ctx); stamp(job, stage, "done", job["stages"][stage].get("note")); ctx.save(force=True); ok = True; break
                 ~~~~~~~~~~^^^^^
  File "/opt/meetings/app/worker.py", line 49, in stage_diar
    identify.identify(ctx, track, out, act, keys, to_abs); ctx.write_json(out, "p3.json"); diar.name_clusters(ctx, out, min_conf=identify.MATCH_MIN)
    ~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/meetings/app/pipeline/identify.py", line 109, in identify
    g = [x for x in plan if plan[x] == rep]; e = np.mean([embs[x]["emb"] for x in g], 0); embs[rep] = dict(embs[rep], emb=e / np.linalg.norm(e)); embs.pop(k, None)
                                                          ~~~~^^^
KeyError: 'SPEAKER_01'

20:15:15 FAILED at diar: 'SPEAKER_01'
20:17:46 re-run: previous recording moved aside
20:17:46 retry: resuming diar with its partial results
20:17:46 start (recording, release meetings)
20:18:17 identify: merged over-split clusters SPEAKER_01→SPEAKER_00, SPEAKER_02→SPEAKER_00
20:18:17 identify: 1 clusters vs 9 voiceprints in 31 s
20:18:17 diar: SPEAKER_00→Speaker 1
20:18:17 align: 237 turns from exclusive diarization
20:18:17 align: 2722 words spread over the Qwen line spans (no MMS)
20:18:17 judge: 236 turns (1 short empty turns skipped), 236 to do
20:18:19 FAILED at judge: HTTP 402: {"error":{"message":"This request requires at least $0.50 in balance for audio","code":402,"metadata":{"limit_source":"openrouter_credits","remedy_hint":"Add credits at https://openrouter.ai/settings/credits, or lower max_tokens / prompt size to fit your remaining balance."}}}