This release includes 2 breaking changes for platform teams planning a safe upgrade.
✓ No known CVEs patched in this version
Topics
+4 more
Summary
AI summaryRemoved the staging layer and drift bug that left projects without playbooks when they had many memories.
Full changelog
Playbooks now derive directly from memories with a staleness hash. Fixes the drift bug where projects with many memories ended up with no playbook because the old staging layer only populated during active mining batches.
Changed
_playbook_refine(project)reads memories directly (no staging layer)- Source-hash gate: re-running with unchanged memories is a cheap no-op, no Haiku call
_playbook_sweep()refreshes every project with ≥5 memories in one pass — mining now calls this at end-of-batch, so no project is left behind due to batch gating- Minimum threshold of 5 memories before a playbook is written
--rebuild-playbooksnow uses the sweep, with optional--forceflag
Removed
_playbook_appendand staging jsonl files — the staging layer that caused the drift bug
Migration
On the next mining batch (or python3 -m memem.server --rebuild-playbooks), the sweep rebuilds all drifted projects from current memories. Expect ~12 Haiku calls for a typical drifted install.
Breaking Changes
- Removed `_playbook_append` function and staging JSONL files that implemented the staging layer
- Playbooks now require a minimum of 5 memories before being written
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About TT-Wang/cortex-plugin
Persistent, self-evolving memory plugin for Claude Code. Background miner extracts durable lessons (decisions, conventions, bug fixes)
Related context
Beta — feedback welcome: [email protected]