# Changelog
All notable changes to this project will be documented in this file.
The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
## [Unreleased]
## [0.11.5](https://github.com/nambok/mentedb/compare/mentedb-v0.11.4...mentedb-v0.11.5) - 2026-07-11
### Fixed
- remove redundant format arg borrows (clippy 1.97 useless_borrows_in_formatting) ([#125](https://github.com/nambok/mentedb/pull/125))
### Other
- correct LongMemEval to the real 92.0% (460/500), replace cherry-picked composite with the honest baseline run + committed judge labels ([#124](https://github.com/nambok/mentedb/pull/124))
## [0.11.4](https://github.com/nambok/mentedb/compare/mentedb-v0.11.3...mentedb-v0.11.4) - 2026-07-10
### Fixed
- decay actually affects recall ranking (recompute decayed salience in hybrid recall hot path) ([#121](https://github.com/nambok/mentedb/pull/121))
## [0.11.3](https://github.com/nambok/mentedb/compare/mentedb-v0.11.1...mentedb-v0.11.3) - 2026-07-06
### Added
- close_quick for fast lock release on shutdown ([#119](https://github.com/nambok/mentedb/pull/119))
### Other
- release v0.11.2 ([#118](https://github.com/nambok/mentedb/pull/118))
## [0.11.2](https://github.com/nambok/mentedb/compare/mentedb-v0.11.1...mentedb-v0.11.2) - 2026-07-06
### Added
- close_quick for fast lock release on shutdown ([#119](https://github.com/nambok/mentedb/pull/119))
## [0.11.1](https://github.com/nambok/mentedb/compare/mentedb-v0.11.0...mentedb-v0.11.1) - 2026-07-06
### Fixed
- ghost memories never inject as knowledge ([#116](https://github.com/nambok/mentedb/pull/116))
## [0.11.0](https://github.com/nambok/mentedb/compare/mentedb-v0.10.8...mentedb-v0.11.0) - 2026-07-06
### Other
- amortize snapshot writes across flushes, reconcile stale snapshots at open ([#113](https://github.com/nambok/mentedb/pull/113))
## [0.10.8](https://github.com/nambok/mentedb/compare/mentedb-v0.10.6...mentedb-v0.10.8) - 2026-07-06
### Fixed
- integration tests carry the new InjectionQuery agent_id field ([#110](https://github.com/nambok/mentedb/pull/110))
### Other
- release v0.10.6 ([#109](https://github.com/nambok/mentedb/pull/109))
## [0.10.7](https://github.com/nambok/mentedb/compare/mentedb-v0.10.6...mentedb-v0.10.7) - 2026-07-06
### Fixed
- integration tests carry the new InjectionQuery agent_id field ([#110](https://github.com/nambok/mentedb/pull/110))
## [0.10.5](https://github.com/nambok/mentedb/compare/mentedb-v0.10.4...mentedb-v0.10.5) - 2026-07-06
### Fixed
- installable on debian and arm linux, publish mentedb-server ([#102](https://github.com/nambok/mentedb/pull/102))
## [0.10.3](https://github.com/nambok/mentedb/compare/mentedb-v0.10.2...mentedb-v0.10.3) - 2026-07-05
### Fixed
- enforce single process ownership with an exclusive lock held from open to close ([#97](https://github.com/nambok/mentedb/pull/97))
## [0.10.2](https://github.com/nambok/mentedb/compare/mentedb-v0.10.1...mentedb-v0.10.2) - 2026-07-04
### Fixed
- correction detection scans only the user message and tags the episodic turn instead of storing a verbatim semantic node ([#96](https://github.com/nambok/mentedb/pull/96))
## [0.10.1](https://github.com/nambok/mentedb/compare/mentedb-v0.10.0...mentedb-v0.10.1) - 2026-07-03
### Other
- lead with Claude Code hooks and connector integration paths, bump benchmark version to 0.10.0 ([#92](https://github.com/nambok/mentedb/pull/92))
## [0.10.0](https://github.com/nambok/mentedb/compare/mentedb-v0.9.2...mentedb-v0.10.0) - 2026-07-03
### Fixed
- dedup write-inference edges, make similarity bands exclusive, run inference in store_batch
- wire embedding provider into server and stop silent zero-vector fallback in process_turn
- crash-durable graph edges via edge log and rebuild graph nodes from storage on open
- make forget durable via WAL-logged page free and update memories in place
### Other
- remove unimplemented vscode setup target from README
- align README and architecture docs with actual engine behavior
## [0.9.2](https://github.com/nambok/mentedb/compare/mentedb-v0.9.1...mentedb-v0.9.2) - 2026-05-13
### Added
- benchmark improvements + engine fixes
### Fixed
- resolve clippy and fmt warnings in engine/SDK
### Other
- update LongMemEval benchmark results to 95.2% (476/500)
## [0.9.0](https://github.com/nambok/mentedb/compare/mentedb-v0.8.2...mentedb-v0.9.0) - 2026-04-27
### Fixed
- WAL durability, page checksums, graph cleanup, salience O(1)
## [0.8.2](https://github.com/nambok/mentedb/compare/mentedb-v0.8.1...mentedb-v0.8.2) - 2026-04-26
### Added
- add enrichment orchestrator module to engine
### Other
- Merge pull request #76 from nambok/feat/enrichment-phase3-4
- fix cargo fmt in enrichment.rs
- add sleeptime enrichment to READMEs
## [0.8.1](https://github.com/nambok/mentedb/compare/mentedb-v0.8.0...mentedb-v0.8.1) - 2026-04-26
### Added
- add community detection and user model generation (Phase 3 & 4)
### Fixed
- address code review findings for Phase 3 & 4
## [0.8.0](https://github.com/nambok/mentedb/compare/mentedb-v0.7.2...mentedb-v0.8.0) - 2026-04-26
### Fixed
- entity resolver hyphen matching, context truncation overflow, doc comments
- drop page_map read lock before loading memories to prevent potential deadlock
- fix duplicate entity link edges on repeated runs, add idempotency test
### Other
- fmt
- LLM-powered entity linking: replace threshold heuristics with EntityResolver + LLM pipeline
- add entity linking: name+embedding disambiguation, entity tags, SDK methods, tests
- add sleeptime enrichment foundation: config, API, always-scope, trigger
- fix READMEs — process_turn as primary API, remove deprecated ingest, fix class names
- rewrite quick start around process_turn
- remove unused MIN_WORD_LEN constant
## [0.7.2](https://github.com/nambok/mentedb/compare/mentedb-v0.7.1...mentedb-v0.7.2) - 2026-04-26
### Other
- add installation section, update Docker and SDK examples
## [0.7.1](https://github.com/nambok/mentedb/compare/mentedb-v0.7.0...mentedb-v0.7.1) - 2026-04-26
### Added
- unified process_turn() engine API ([#67](https://github.com/nambok/mentedb/pull/67))
### Fixed
- sync SDK versions to 0.7.0 and add process_turn ([#69](https://github.com/nambok/mentedb/pull/69))
## [0.6.2](https://github.com/nambok/mentedb/compare/mentedb-v0.6.1...mentedb-v0.6.2) - 2026-04-26
### Added
- wire all 9 remaining cognitive subsystems into MenteDb ([#66](https://github.com/nambok/mentedb/pull/66))
- wire cognitive engine subsystems into MenteDb ([#65](https://github.com/nambok/mentedb/pull/65))
- token efficiency benchmark ([#63](https://github.com/nambok/mentedb/pull/63))
## [0.6.1](https://github.com/nambok/mentedb/compare/mentedb-v0.6.0...mentedb-v0.6.1) - 2026-04-23
### Added
- export VERSION const from mentedb crate
## [0.6.0](https://github.com/nambok/mentedb/compare/mentedb-v0.5.2...mentedb-v0.6.0) - 2026-04-20
### Added
- bincode index persistence, async LLM extraction, SDK concurrency
### Fixed
- remove global RwLock — each component uses interior mutability
### Other
- update README test count, ARCHITECTURE server section, lib.rs module list
- eliminate global write lock for read operations
## [0.5.1](https://github.com/nambok/mentedb/compare/mentedb-v0.5.0...mentedb-v0.5.1) - 2026-04-19
### Fixed
- resolve all clippy warnings across workspace
- use rustls-tls instead of native-tls in extraction crate
### Other
- update MCP install to npx as primary method
## [0.5.0](https://github.com/nambok/mentedb/compare/mentedb-v0.4.2...mentedb-v0.5.0) - 2026-04-13
### Added
- add LongMemEval benchmark harness ([#48](https://github.com/nambok/mentedb/pull/48))
## [0.4.2](https://github.com/nambok/mentedb/compare/mentedb-v0.4.1...mentedb-v0.4.2) - 2026-04-07
### Other
- Add entity resolution with three tier strategy ([#43](https://github.com/nambok/mentedb/pull/43))
## [0.4.1](https://github.com/nambok/mentedb/compare/mentedb-v0.3.2...mentedb-v0.4.1) - 2026-04-07
### Other
- Wire LLM topic canonicalization into TransitionMap ([#42](https://github.com/nambok/mentedb/pull/42))
- release v0.4.0 ([#41](https://github.com/nambok/mentedb/pull/41))
- Add bi-temporal validity to memories and edges ([#35](https://github.com/nambok/mentedb/pull/35))
## [0.4.0](https://github.com/nambok/mentedb/compare/mentedb-v0.3.2...mentedb-v0.4.0) - 2026-04-07
### Other
- Add bi-temporal validity to memories and edges ([#35](https://github.com/nambok/mentedb/pull/35))
## [0.3.2](https://github.com/nambok/mentedb/compare/mentedb-v0.3.1...mentedb-v0.3.2) - 2026-04-07
### Other
- release v0.3.2 ([#39](https://github.com/nambok/mentedb/pull/39))
- Add LLM accuracy benchmark results to README ([#38](https://github.com/nambok/mentedb/pull/38))