@@ -11,6 +11,37 @@ log is the canonical source for those.
1111
1212## [ Unreleased]
1313
14+ ## [ 0.19.0] - 2026-06-08
15+
16+ ### Added
17+
18+ - Added planned semantic search for ` lore-query action='search' ` , including
19+ the ` strategy: "planned" | "direct" ` option, sanitized query variants,
20+ query-plan metadata in tool output, and capability-oriented candidate context
21+ for SkillRet agent/eval workflows. (#952 )
22+ - Published the sanitized 500-query SkillRet read-only agent checkpoint for
23+ planned search chunk 001, with corrected timeout-rerun provenance and report
24+ synthesis. (#952 )
25+
26+ ### Changed
27+
28+ - Query planning now stays enabled by default and is configurable only through
29+ ` .lore.yaml ` ` features.queryPlanning ` ; the environment rollback path was
30+ removed from feature-flag taxonomy and documentation. (#953 )
31+ - Confidence no longer affects retrieval ranking. Confidence and trust remain
32+ visible for display, auditing, and fact maintenance, while confidence audit
33+ and help text now describe trust-score/display impact rather than ranking
34+ impact. (#953 )
35+ - Fresh vault schemas no longer create legacy Memories ` Confidence ` /
36+ ` Confidence Score ` columns, while existing vaults retain compatibility for
37+ reads and migrations. (#953 )
38+
39+ ### Fixed
40+
41+ - Guarded the autosave-learning-source backfill so fresh schemas without the
42+ retired Memories ` Confidence ` column return zero candidates before issuing
43+ legacy-property queries. (#953 )
44+
1445## [ 0.18.1] - 2026-06-05
1546
1647### Added
0 commit comments