perf(chroma): cache backfill watermarks in JSON to skip per-restart Chroma scans

Browse Source

Worker restarts triggered a full Chroma metadata scan for every project on every
boot to figure out which sqlite ids were already embedded. With 253 projects and
~92k embeddings, this pegged chroma-mcp at 100-422% CPU on every spawn.

Replace the scan with ~/.claude-mem/chroma-sync-state.json — per-project highest
synced sqlite_id watermarks for observations/summaries/prompts. Backfill switches
from "id NOT IN (huge list)" to "id > watermark"; live syncs bump the watermark
on success; one-time bootstrap derives initial watermarks from a single Chroma
scan if the state file is missing.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

...

This commit is contained in:

Alex Newman

2026-04-25 14:05:48 -07:00

parent d7c7eccd7f

commit 5769f00827

3 changed files with 430 additions and 222 deletions

Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL

Download Patch File Download Diff File Expand all files Collapse all files

plugin/scripts/worker-service.cjs

+159 -159

File diff suppressed because one or more lines are too long