@@ -4,22 +4,53 @@ All notable changes to this project will be documented in this file.
44
55## [ Unreleased]
66
7+ ## [ 0.8.0] - 2026-03-30
8+
79### Added
810
911** Three-Layer Distillation Pipeline**
1012- Automatic distillation via NullDistiller in ingest pipeline (Layer 1: regex-based, P95 < 5ms)
1113- Context hook injection for LLM-based extraction at SessionStart (Layer 2: Claude Code as distiller, zero extra cost)
1214- ` /distill-transcripts ` skill for manual deep extraction (Layer 3: on-demand, depth-aware prompts)
13- - ` memory.undistilled ` MCP tool to list content items not yet deeply distilled
14- - ` memory.mark_distilled ` MCP tool to mark content items as distilled after extraction
15+ - ` memory.undistilled ` and ` memory.mark_distilled ` MCP tools for distillation tracking
16+ - ` Hook::DistillationRunner ` extracted from Handler for context hook injection
1517- ` TaskCompleted ` and ` TeammateIdle ` hook events for ingest triggers
1618- Distillation metrics backfill on database initialization
1719- Doctor check for undistilled content
20+ - Pending distillation count in ` memory.status ` output
21+
22+ ** Recall Enhancements**
23+ - Intent parameter for recall query disambiguation (#3 )
24+ - Retrieval score traces for semantic search (#5 )
25+ - Configurable embedding providers with dimension checking
26+
27+ ** Hook Enhancements**
28+ - ` statusMessage ` on all hooks for descriptive spinner text during hook execution
29+ - ` StopFailure ` hook to capture transcript data even on session errors (rate limits, server errors)
30+ - ` Notification ` hook with ` idle_prompt ` matcher for opportunistic sweep during idle
31+
32+ ** New Commands & Skills**
33+ - ` install-skill ` command and ` memory-recall ` agent (#8 , #12 )
34+ - Shell completion command for bash and zsh (#18 )
1835
1936** Distillation Benchmark Results**
2037- NullDistiller: Concept Recall 0.952, Fact Precision/Recall 1.000 (31 test cases)
2138- Claude Code LLM: Concept Recall 0.902 (all 41 cases), 0.900 on semantic cases (vs 0.333 for regex)
2239- Average 1.6 facts stored per case across LLM extraction
40+ - E2E distillation recall benchmark and extraction quality benchmarks
41+ - Concept-based matching for distiller-agnostic benchmark comparison
42+
43+ ### Fixed
44+
45+ - ` --allowedTools ` added to ` ClaudeCliRunner ` for MCP tool permissions
46+ - Test isolation for context hook when global database has facts
47+
48+ ### Internal
49+ - Extracted ` RetryHandler ` and ` SchemaManager ` modules from ` SQLiteStore `
50+ - Extracted ` Recall ` into engine strategy pattern with ` DualEngine ` , ` LegacyEngine ` , and shared ` QueryCore `
51+ - Extracted ` Tools ` god object into 6 handler modules
52+ - Added 36 specs for 5 previously untested files
53+ - All 3 god objects eliminated, 0 files over 500 lines
2354
2455## [ 0.7.1] - 2026-03-17
2556
0 commit comments