v3.1.0

What's Changed

• feat(retrieval): adaptive expansion-gate for broad prompts (#741) by @robotrocketscience in #743
• feat(rebuilder): default trigger_mode threshold (#746) by @robotrocketscience in #750
• fix(transcript): filter harness-wrapper user turns at logger + broaden noise filter for
  Monitor (#747) by @robotrocketscience in #751
• fix(store): retype legacy belief_corroborations.belief_id INTEGER -> TEXT (#762) by @robotrocketscience in #765
• feat: cross-fire injection dedup ring (#740) by @robotrocketscience in #744
• feat(bench): PR-smoke fixtures + CI for benchmark adapters (#476) by @robotrocketscience in #752
• feat(onboard): add 'aelf onboard --check' read-only pre-scan (#761) by @robotrocketscience in #767
• feat(canvas): JSON Canvas 1.0 exporter probe for Obsidian mapping (#763) by @robotrocketscience in #766
• feat(meta_beliefs): adaptive meta-belief substrate (#755, umbrella #480) by @robotrocketscience in #771
• refactor(store): install_meta_belief → INSERT OR IGNORE for atomic idempotence (#772) by @robotrocketscience in #773
• docs: sweep stale Default-OFF claims for already-flipped retrieval flags by @robotrocketscience in #770
• docs: correct stale Default-OFF claim in retrieve_v2 docstring by @robotrocketscience in #768
• fix(lifecycle): runnable uv install one-liner when maybe_migrate_to_uv hits uv-not-found by @robotrocketscience in #774
• feat(bench): expansion-gate latency micro-bench harness (#741) by @robotrocketscience in #764
• feat(bench): publish v3.0.1 parity results by @robotrocketscience in #780
• feat(bench): v3.0 BFS latency harness + first run (#739) by @robotrocketscience in #754
• feat(eval-harness): v3.1 close — sharpen JUDGE_PROMPT_TEMPLATE + widen hot_start fixture (#778) by @robotrocketscience in #782
• feat(bench): wire A4 continuation-fidelity bench gate for type-aware compression (#775) by @robotrocketscience in #776
• feat(status): show aelfrice version in stats snapshot by @robotrocketscience in #783
• feat(retrieval): first meta-belief consumer — adaptive temporal_half_life_seconds (#756) by @robotrocketscience in #784
• feat(retrieval): second meta-belief consumer — adaptive bm25f_anchor_weight (#757) by @robotrocketscience in #787
• docs(ingest): spec for speaker-gate + sentiment-routing + SVO floor (#785) by @robotrocketscience in #786
• fix(setup): collapse hook accumulation + prune dead aelf-* entries (#781) by @robotrocketscience in #788
• feat(retrieval): third meta-belief consumer — adaptive bfs_depth_budget (#759) by @robotrocketscience in #790
• fix(test): bfs_latency_v3 smoke flake — bump iterations + decouple from gate verdict (#792) by @robotrocketscience in #793
• docs: v3.0.1 sweep — README Status + ROADMAP rows (#791) by @robotrocketscience in #794
• feat(ingest): speaker-attribution gate excludes assistant rows from belief creation (#785 §1) by @robotrocketscience in #795
• feat(retrieval): live close-the-loop relevance-signal infrastructure (#779) by @robotrocketscience in #789
• feat(retrieval): adaptive expansion_gate token threshold (#760) by @robotrocketscience in #799
• feat(onboard): rejection ledger so re-runs don't re-classify rejected sentences (#801) by @robotrocketscience in #806
• ci(publish): promote release-drafter draft on tag push (#804) by @robotrocketscience in #805
• feat(bench): tighten post-clear replay template + ratify 0.80 AC scope (#797 P3) by @robotrocketscience in #802
• release: v3.1.0 by @robotrocketscience in #808

Full Changelog: v3.0.1...v3.1.0