Skip to content

fix: improve memory footprint of recall#423

Merged
nicoloboschi merged 1 commit intomainfrom
mem-leak
Feb 23, 2026
Merged

fix: improve memory footprint of recall#423
nicoloboschi merged 1 commit intomainfrom
mem-leak

Conversation

@nicoloboschi
Copy link
Collaborator

@nicoloboschi nicoloboschi commented Feb 23, 2026

very high change to fix #418

@nicoloboschi nicoloboschi reopened this Feb 23, 2026
@nicoloboschi nicoloboschi merged commit 9f0c031 into main Feb 23, 2026
28 of 31 checks passed
r0gig0r pushed a commit to r0gig0r/hindsight that referenced this pull request Feb 23, 2026
…ptimization

Upstream changes (5 commits):
- vectorize-io#417: Fix bank config API for multi-tenant schema isolation
- vectorize-io#420: Add ZeroEntropy reranker provider support
- vectorize-io#421: Fix reflect based_on population and enforce full hierarchical retrieval
- vectorize-io#423: Improve memory footprint of recall (drop embedding from retrieval pipeline)
- vectorize-io#424: Include doc metadata in fact extraction prompts

Conflict resolution:
- fact_extraction.py: kept both our bank_mission and upstream's metadata params
- search/types.py: re-added optional embedding field for recall_exp diversity clustering
  (upstream removed it from standard recall path for memory savings)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Memory leak in consolidation worker during large batch processing

1 participant