v3.0.0

What's Changed

• feat(wonder): research-agent dispatch core — analyze_gaps + axes (#551) by @robotrocketscience in #558
• feat(wonder): wonder_ingest + wonder_gc lifecycle (closes #229, #548) by @robotrocketscience in #559
• test(session_id): ≥80% population guard via two-session ingest run (#554) by @robotrocketscience in #561
• docs(hrr): v2.1 integration spec — persistence + split-format + opt-out (#553) by @robotrocketscience in #556
• docs(readme): add 'What makes aelfrice different' section by @robotrocketscience in #564
• docs(readme): rewrite layered-retrieval bullet for plain-language clarity by @robotrocketscience in #567
• test(implicit_feedback): nonlinear age-vs-alpha-drift correlation guard (#555) by @robotrocketscience in #566
• feat(doctor): document + flag missing v2.1 auto-capture hooks (#557) by @robotrocketscience in #560
• feat(bench): temporal_half_life_seconds sweep harness (#487) by @yoshi280 in #574
• feat(ci): issue-creation content audit — flag duplicates of shipped surface (#569) by @robotrocketscience in #573
• ci: auto-unblock issues on blocker-close — flip Blocked → Todo when all blockers resolve by @robotrocketscience in #576
• fix(extraction): preserve underscored identifiers in markdown strip by @robotrocketscience in #571
• perf(lifecycle): drop update-check TTL 6h → 15min by @robotrocketscience in #565
• docs: scrub 2026-05-09 — refresh user-facing docs against v2.1.0 ship state by @robotrocketscience in #563
• feat(retrieval): use_intentional_clustering flips to default-on (#436) by @robotrocketscience in #577
• feat(hook): session-start enrichment with locked + core (#578) by @robotrocketscience in #584
• ci: auto-populate + auto-status v2.1 project board (#579) by @robotrocketscience in #586
• feat(cli,mcp): wonder --persist + gc surface (#549) by @robotrocketscience in #575
• ci: render Mermaid dep graph from Blocked-by links (#581) by @robotrocketscience in #585
• chore: add FUNDING.yml by @robotrocketscience in #588
• feat(hook): session-end Stop hook prompts to lock session corrections (#582) by @robotrocketscience in #590
• feat(hooks): pre-push branch-freshness check (#580) by @robotrocketscience in #583
• chore: remove FUNDING.yml by @robotrocketscience in #595
• feat(eval-harness): wire context-rebuilder eval harness commit-1 (#592) by @robotrocketscience in #596
• feat(benchmark): add multi-hop section to validate L2.5 entity-index lift (#597) by @robotrocketscience in #598
• feat(doctor): nag when per-project DBs are on pre-v1.x schema (#589) by @robotrocketscience in #599
• feat(vocab-bridge): remove vocab_bridge.py module (closes #536, #433) by @robotrocketscience in #540
• feat(rebuilder): post-compact hot-start (#587) by @robotrocketscience in #591
• docs(readme): tighten tagline + bullet wording by @robotrocketscience in #568
• ci(merge-thrash): pr-size-soft-cap advisory workflow (#602) by @robotrocketscience in #603
• ci(merge-train): label-driven serialized FF merger (#602) by @robotrocketscience in #604
• docs: cut v3.0 scope (#608) — replace stale v2.2 line with substrate-completion + design tickets by @robotrocketscience in #609
• feat(eval-harness): judges/llm_judge.py — opt-in LLM-judge stage (#592) by @robotrocketscience in #613
• docs(philosophy): v3.0 relatedness ratification — stay deterministic (closes #605) by @robotrocketscience in #610
• feat(hrr): split-format save/load on HRRStructIndex (#553 substrate) by @robotrocketscience in #617
• docs(v3-scope): multimodel scope memo — subsumed by wonder-dispatch by @robotrocketscience in #615
• feat(sentiment-hook): production wire-up of sentiment-feedback into UPS lane (closes #606) by @robotrocketscience in #612
• ci: wire workflows for GitHub native merge queue (closes #624) by @robotrocketscience in #625
• feat(doctor|cli): aelf doctor --prune-dormant interactive deletion (#594) by @robotrocketscience in #628
• revert: aborted GitHub native merge queue migration (#624) by @robotrocketscience in #631
• fix(merge-train): dedup check-runs by name + drop CANCELLED (#632) by @robotrocketscience in #633
• feat(setup): self-installing hook manifest at aelf-CLI first run (closes #623) by @robotrocketscience in #627
• feat(doctor): auto-migrate pre-v1.x per-project DBs in place (closes #593) by @robotrocketscience in #626
• fix(slash): /aelf:upgrade keeps step-2 execution imperative (closes #611) by @robotrocketscience in #614
• fix(merge-train): auto-close linked issues post-FF-push (#634) by @robotrocketscience in #635
• feat(eval-harness): replay_post_fork host-agent JSONL split (closes #600) by @robotrocketscience in #601
• fix(merge-train): grant issues:write to allow auto-close (closes #637) by @robotrocketscience in #638
• feat(tests): A2 recall@k bench gate for type-aware compression (#434) by @robotrocketscience in #641
• build(deps): bump urllib3 to 2.7.0 — CVE patches (closes #640) by @robotrocketscience in #642
• feat(eval-harness): hot_start synthetic fixture (#592) by @robotrocketscience in #639
• feat(wonder): re-run bake-off + revisit memo with broadened verdict (closes #547) by @robotrocketscience in #643
• fix(merge-train): re-fire on rebased PRs (closes #646) by @robotrocketscience in #647
• test(bench-gate): #291 query-strategy p99 latency gate by @robotrocketscience in #648
• feat(cli): aelf wonder QUERY positional + N-agent shorthand (#645) by @robotrocketscience in #649
• feat(reason): VERDICT + IMPASSES classifiers on aelf reason (#645 R1) by @robotrocketscience in #660
• feat(wonder): WonderResult dataclass for structured wonder output (#656) by @robotrocketscience in #662
• fix(wonder): clamp WonderResult.coverage to [0, 1] (#667) by @robotrocketscience in #669
• feat(reason): R3 VERDICT-driven dispatch + feedback close-the-loop (#645) by @robotrocketscience in #665
• feat(retrieval): #N literal boost in BM25 scoring for issue/PR-number prompts (#677) by @robotrocketscience in #678
• feat(noise_filter): N-gram Jaccard similarity gate + aelf scan-derivation CLI (#681) by @robotrocketscience in #684
• docs: fix 6 broken doc links from link-check report (#683) by @robotrocketscience in #685
• docs(v3.0): release-gate documentation hygiene — ROADMAP + CHANGELOG + federation-primitives + LIMITATIONS (#663) by @robotrocketscience in #682
• docs(readme): add Lin memory-system framework table by @robotrocketscience in #666
• feat(hook): prompt-shape gate to skip BM25 on acks and system messages (closes #674) by @robotrocketscience in #680
• feat(reason): compound confidence + CONTRADICTS fork on aelf reason (#658 R2) by @robotrocketscience in #664
• fix(reason): ratio-based fork-tie threshold for compound_confidence (#668) by @robotrocketscience in #671
• feat(federation): read-only mechanics — knowledge_deps.json + peer ATTACH + foreign-ID rejection (#655) by @robotrocketscience in #670
• test(promotion): adversarial bench fixture for Surface B (#672, supersedes #673) by @robotrocketscience in #686
• feat(promotion): phantom promotion trigger — Surfaces A+B (closes #550) by @robotrocketscience in #616
• feat(hrr): HRRStructIndexCache persist-aware load+save+mmap (#691) by @robotrocketscience in #693
• feat(bench): Cohen's-κ runner for eval-judge calibration (#687) by @robotrocketscience in #702
• fix(bench): hot_start fixture meta — eval_turns at canonical answers (#687) by @robotrocketscience in #705
• feat(wonder): /aelf:wonder --axes skill-layer dispatch loop (closes #552) by @robotrocketscience in #644
• feat(config): [retrieval] hrr_persist TOML key (#698) by @robotrocketscience in #703
• feat(federation): scope field on beliefs (#688) by @robotrocketscience in #694
• feat(hrr): ephemeral-path auto-disable for HRR persistence (#695) by @robotrocketscience in #701
• test(hrr): cold-start bench gate at N=50k (#697) by @robotrocketscience in #706
• docs(bench): §Eval-judge calibration κ-protocol for #687 by @robotrocketscience in #700
• test(hrr): harden 2 vacuous-pass cache-persistence assertions (#707) by @robotrocketscience in #709
• feat(doctor): HRR persist/build reporter rows (#696) by @robotrocketscience in #704
• docs(hrr): CONFIG.md + COMMANDS.md + CHANGELOG bundle (#699, closes #553) by @robotrocketscience in #714
• fix(store): update_belief is now full-row + validates retention_class (#708) by @robotrocketscience in #710
• feat(reason): peer-aware graph walk (#690) by @robotrocketscience in #712
• feat(cli): aelf promote/demote --to-scope flag (#689) by @robotrocketscience in #711
• docs(wonder): refresh slash dedup-behaviour paragraph for shipped #644 v2 key by @robotrocketscience in #715
• feat: aelf clamp-ghosts — clamp α on audit-trail-less ghost belief rows by @robotrocketscience in #692
• feat(cli): wire aelf reason seed-fetcher to populate seed_scopes for foreign beliefs (#713) by @robotrocketscience in #716
• ci(merge-train): pre-check conversation-resolution before FF push (#721) by @robotrocketscience in #723
• docs(config): add [rebuilder] section to CONFIG.md (#720) by @robotrocketscience in #722
• feat(retrieval): flip query_strategy default legacy-bm25 → stack-r1-r3 (#718) by @robotrocketscience in #719
• perf(replay): N+1 belief loop → set-based query in _compute_replay_drift_report (#725) by @robotrocketscience in #726
• feat(ingest): transcript-noise filter to stop ingesting tool-output and agent-emit as beliefs (#675) by @robotrocketscience in #679
• fix(bench): debug fixture meta — eval_turns point at canonical answers (#687) by @robotrocketscience in #727
• release: v3.0.0 — CHANGELOG consolidation + docs scrub by @robotrocketscience in #728

Full Changelog: v2.1.0...v3.0.0