hyungi
5dca5b5d28
ops(pipeline): embed/chunk 고속 컨슈머 분리 + 배치 1→10 — LLM 사이클 인질 해소
...
진단(2026-06-12 용량 평가): 단일 루프에서 classify(~190s×3)가 사이클을 점유,
건당 <1s 인 embed/chunk 가 사이클당 1건 캡 → 실효 ~580/일 vs 수요 최대 2,700/일,
적체 3,570 + 신규 문서 벡터 미적재(RAG 검색 누락). 4070 가동률 0% = 순수 구조 캡.
수리 = markdown 분리(05-01) 선례: consume_fast_queue 1분 잡 + 배치 10(GPU 공유 보수값,
캡 ~14,400/일). 세 컨슈머 stage 집합 disjoint(stale reset 이중 복구 방지). retrieval
로직·임베딩 모델 무접촉.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com >
2026-06-12 07:50:07 +09:00
..
2026-05-22 13:43:47 +00:00
2026-04-17 08:11:06 +09:00
2026-06-11 17:19:35 +09:00
2026-06-11 12:55:16 +09:00
2026-06-09 10:12:26 +09:00
2026-05-14 09:42:07 +09:00
2026-04-10 08:49:11 +09:00
2026-04-24 09:42:24 +09:00
2026-04-08 12:31:29 +09:00
2026-05-25 07:02:46 +00:00
2026-05-22 13:43:47 +00:00
2026-04-02 10:20:15 +09:00
2026-05-19 12:43:53 +09:00
2026-04-02 10:20:15 +09:00
2026-04-17 08:11:06 +09:00
2026-05-12 13:15:26 +09:00
2026-05-25 05:37:15 +00:00
2026-06-11 06:32:15 +09:00
2026-06-11 07:54:13 +09:00
2026-05-02 08:35:34 +09:00
2026-06-12 06:56:02 +09:00
2026-05-03 08:38:09 +09:00
2026-04-17 08:11:06 +09:00
2026-05-19 12:43:53 +09:00
2026-05-19 12:43:53 +09:00
2026-05-19 08:54:07 +09:00
2026-06-12 07:22:47 +09:00
2026-05-10 14:05:41 +09:00
2026-06-12 07:50:07 +09:00
2026-06-12 06:56:02 +09:00
2026-05-24 04:48:50 +00:00
2026-06-11 14:36:10 +09:00
2026-06-08 03:05:30 +00:00
2026-06-08 03:05:30 +00:00
2026-05-02 07:33:57 +09:00
2026-06-07 10:11:38 +09:00
2026-06-07 08:08:55 +09:00
2026-04-17 08:29:49 +09:00
2026-04-17 08:11:06 +09:00
2026-05-19 12:43:53 +09:00
2026-05-19 12:43:53 +09:00
2026-05-19 12:43:53 +09:00
2026-05-19 12:43:53 +09:00
2026-05-19 12:55:51 +09:00
2026-05-19 12:55:51 +09:00