Solution for long term memory for agent coding CLIs and to facilitate handoff between different agent vendors
Rust
docs: point the changelog entries at this PR
Djalma Júnior259d8fa96e16
diff --git a/CHANGELOG.md b/CHANGELOG.mdindex 93eea781..780e56e7 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -37,7 +37,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0a NULL-collapsing unique index, so one operator's pending suggestion stopsblocking everybody else's for the same page while every unattributed caller— single-operator servers, the scheduler, the telemetry report, the curator- — keeps the original one-per-page rule unchanged (#310).+ — keeps the original one-per-page rule unchanged (#336).- Page reinforcement is recorded per operator (V43 `page_access`, keyed on theidentity storage key) beside the untouched shared counter, and the retentionformula gains an opt-in breadth term: `[decay] breadth_weight` (default@@ -46,7 +46,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0distinct readers under any weight, so no existing eviction decision movesuntil an operator deliberately turns it on; the per-operator access-bumpthrottle keys on (page, operator) so one reader cannot swallow a- colleague's reinforcement inside the cooldown window (#310).+ colleague's reinforcement inside the cooldown window (#336).### Fixed- A staged auto-improve proposal colliding with one already pending no longer@@ -55,7 +55,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0staging surface — `memory_auto_improve`, `/admin/auto-improve`, thetelemetry report, the curator, the CLI and the scheduler's log — names theskipped target and the reason instead of silently returning N-1 proposals- (#310).+ (#336).## [1.21.0] - 2026-07-31