fleet-memory/hindsight-api/hindsight_api/engine/consolidation
Nicolò Boschi 8781c9fbfe
feat: add real-time timing breakdown logging for consolidation (#235)
- Log timing breakdown after each batch (every 50 memories by default)
- Log timing breakdown in progress logs (every 10 memories)
- Shows recall, llm, embedding, db_write times incrementally
- Includes avg time per memory for quick diagnosis
- Helps diagnose performance issues in production without waiting for job completion

Example output (every 10 memories):
[CONSOLIDATION] bank=xyz progress: 10/39303 memories processed | recall=2.09s, llm=11.03s, embedding=0.48s, db_write=0.02s

Example output (per batch):
[CONSOLIDATION] bank=xyz batch 1/50 memories: recall=7.3s, llm=57.5s, embedding=2.0s, db_write=0.09s | avg=1.3s/memory
2026-01-29 17:50:27 +01:00
..
__init__.py feat: revisit mental models, directives and reflections (#179) 2026-01-22 17:13:16 +01:00
consolidator.py feat: add real-time timing breakdown logging for consolidation (#235) 2026-01-29 17:50:27 +01:00
prompts.py feat: consolidation performance benchmark and optimization (#227) 2026-01-29 11:24:15 +01:00