Skip to content

LORE 0.62.30 — Memory maintenance

Latest

Choose a tag to compare

@docwilde docwilde released this 15 Nov 06:01
797e8a1

LORE 0.62.30 adds bounded pending review and protected memory maintenance. Proposal approval and signed-sync checks retain their existing authority requirements.

  • Review large pending queues in pages with scoped previews, authenticated continuations, portable resumption, and bounded output.
  • Plan diagnostic retention with reference and recent-capture protection; apply codegraph cleanup only to explicitly reviewed orphan keys. Automatic trace retention remains off by default.
  • Sync opt-in signed skill usage counters across authors, including reset generations, seed, and offline transfer. Imported counts cannot authorize skill updates or retirement.
  • Use retained-history replay, private-hub verification, and isolated subscription coding-suite tools with explicit attempt budgets and independent verifiers.

Validation: 294 native workspace tests, the complete Python 3.12 suite, native launcher and interoperability checks, macOS pagination runtime tests, and regression negative controls.

Known limitations: #112 remains open for exact state parity and missing contemporaneous file snapshots. Live Codex benchmark compatibility and strict Tailnet-only hub verification remain unresolved. Codex hook migration still requires UI discovery and trust; retain working custom hooks until that check passes. The release does not establish model skill improvement.

Update the plugin, then run ./task install from the updated checkout to install the matching native carrier. Review changed Codex hooks with /hooks.