Skip to content

Cognithor 0.96.0

Choose a tag to compare

View all tags
@github-actions github-actions released this 01 May 16:33
· 269 commits to main since this release
v0.96.0
9a5f2c6

What's Changed

Features

  • feat(pse): Phase-2 Sprint-8 Track B — 100 % on every ARC corpus subset (#271) (1a85c36)
  • feat(pse): Phase-2 Sprint-6 Track A — first real Score-Lift via Symbolic-Repair cascade (#269) (30acee6)
  • feat(pse): Phase-2 Sprint-5 Track 1 — A/B-Test on hard subset reveals partial activation (#268) (c968b27)
  • feat(pse): Phase-2 Sprint-4 — ARC-AGI-3 corpus + Phase-1 reality-check (#267) (2aca8e8)
  • feat(pse): Phase-2 Sprint-3 Track 1 — A/B benchmark with --phase2 flag (#266) (5636899)
  • feat(pse): Phase-2 Sprint-2 Track D — Benchmark report + nightly CI (#264) (3e74358)
  • feat(pse): Phase-2 Sprint-2 Track C — 20-Task Leak-Free fixture set (#263) (605a0a4)
  • feat(pse): Phase-2 Sprint-2 Track E — MCTS extensions (parallel + restart + diversity) (#265) (ce87090)
  • feat(pse): Phase-2 Sprint-2 Track B — Production wiring (#262) (546c72d)
  • feat(pse): Phase-2 Sprint-2 — End-to-end Verifier evaluator (§7.3.3, Sprint-2 Track A) (#261) (b0444cc)
  • feat(pse): Phase-2 Sprint-1 — Benchmark runner (plan task 12, §12.3 / §12.4) (#260) (be8bfd6)
  • feat(pse): Phase-2 Sprint-1 — Top-level synthesis engine (plan task 10, §3.2) (#259) (3dfd2bc)
  • feat(pse): Phase-2 Sprint-1 — Module B MCTS controller (plan task 7, §5) (#258) (696d549)
  • feat(pse): Phase-2 Sprint-1 — Symbolic-Prior heuristic catalog (plan task 4, §4.4) (#257) (89ad234)
  • feat(pse): Phase-2 Sprint-1 — Refiner Escalation logic (plan task 9 final slice, §6.6) (#256) (80fb615)
  • feat(pse): Phase-2 Sprint-1 — Hybrid-Repair (plan task 9 slice, §6.5.2 Zone-2) (#255) (366d122)
  • feat(pse): Phase-2 Sprint-1 — Trace-Replay (plan task 9 slice, §6.4) (#253) (fa89ae1)
  • feat(pse): Phase-2 Sprint-1 — LLM-Repair Two-Stage (plan task 9 slice, §6.5.2 Zone-1) (#254) (33da64d)
  • feat(pse): Phase-2 Sprint-1 — Symbolic-Repair Advisor (plan task 9 slice, §6.5.2) (#252) (1847670)
  • feat(pse): Phase-2 Sprint-1 — Diff-Analyzer (plan task 9 slice, §6.3) (#251) (3f05639)
  • feat(pse): Phase-2 Sprint-1 — CEGIS Refiner stage (plan task 9 slice, §6.5.3) (#250) (6dbf420)
  • feat(pse): Phase-2 Sprint-1 — Local-Edit Repair (plan task 9 slice, §6.5.1) (#249) (f702f27)
  • feat(pse): Phase-2 Sprint-1 — triviality rules (plan task 8 slice, §7.3.1) (#248) (bf2b327)
  • feat(pse): Phase-2 Sprint-1 — partial_pixel_match metric (plan task 8 slice) (#247) (2852c7a)
  • feat(pse): Phase-2 Sprint-1 — α-Controller + PriorPerformanceTracker (plan task 6) (#246) (61ba82f)
  • feat(pse): Phase-2 Sprint-1 — verifier final-score aggregation (plan task 8, §7.2) (#245) (402776d)
  • feat(pse): Phase-2 Sprint-1 — typed datatypes (plan task 2, spec §9) (#244) (39bb251)
  • feat(pse): Phase-2 Sprint-1 — YAML heuristics loader (plan task 1) (#243) (0cb3fb3)
  • feat(pse): Phase-2 Sprint-1 — F3 interaction tests (spec v1.4 §12.2) (#242) (387ef6c)
  • feat(pse): Phase-2 Sprint-1 — Dual-Prior mixer + Symbolic-Prior interface (#241) (142034d)
  • feat(pse): Phase-2 Sprint-1 — Module A LLM-Prior over vLLM/Qwen 3.6 27B (#240) (46c7419)
  • feat(pse): Phase-2 Sprint-1 — telemetry counters (spec v1.4 §11) (#239) (289ff74)
  • feat(pse): Phase-2 Sprint-1 — α-mixer + sample-size-dampening (§4.4 / §4.4.4) (#238) (39f7460)
  • feat(pse): Phase-2 Sprint-1 — compute_suspicion helper (spec v1.4 §7.3.2) (#237) (4035a7a)
  • feat(pse): Phase-2 Sprint-1 — auto-classify primitives via @primitive decorator (#236) (6896960)
  • feat(pse): Phase-2 Sprint-1 — F1 (DSL classification) + F2 (Refiner mode controller) (#235) (0a37de8)
  • feat(pse): Phase-1 final close — K4 subprocess sandbox + D5/D7 + D13 (#234) (d6a4b74)
  • feat(pse): D15 — mypy --strict clean + pre-commit pytest hook (#233) (d408183)
  • feat(pse): K4 — flip on 6 type-system-blocked adversarial cases (#232) (2a2fd53)
  • feat(pse): ARC-AGI-3 eval-suite scaffold (D5 prep) (#231) (de2902b)
  • feat(pse): seed Predicate/Lambda bank so HO primitives are reachable (#228) (c4e5d3e)
  • feat(pse): channel-side telemetry + audit-trail wiring — Week 7 day 5 (#225) (11dfa4c)
  • feat(pse): telemetry counters + audit trail — Week 7 day 4-5 (D10 + D11) (#224) (f9192c2)
  • feat(pse): cognithor pse CLI — Week 7 day 4 (D8 — spec §19.2) (#223) (c728bfe)
  • feat(pse): deterministic Cost-Auto-Tuner — Week 7 day 1 (spec §7.6) (#222) (34cd4e9)
  • feat(pse): branch (H5) — conditional-Lambda combinator — Week 6 day 6 (v1.2 spec) (#221) (ae745a9)
  • feat(pse): sort_objects (H4) + SortKey enum — Week 6 day 5,5 (v1.2 spec addition) (#220) (49321e6)
  • feat(pse): align_to (H3) + AlignMode enum — Week 6 day 4-5 (#219) (ee82050)
  • feat(pse): Lambda type + map_objects + filter_objects — Week 6 day 3-4 (Phase 1.5 H1+H2) (#218) (dc6de65)
  • feat(pse): closed-set Predicate type + 13 constructors — Week 6 day 1-2 (Phase 1.5 prerequisite) (#217) (360cfcb)
  • feat(pse): trace + replay system — Week 5 day 4-5 — K9/K10 HARD GATES (#215) (745db7f)
  • feat(pse): PGE adapter + state-graph bridge + numpy fast-path — Week 5 day 1-2 (#214) (35918d2)
  • feat(pse): sandbox strategy + platform detection — Week 4 day 4-5 (spec §11.6) (#213) (76b7b24)
  • feat(pse): tactical-memory cache + capability-token constants — Week 4 day 3 (#212) (9585e55)
  • feat(pse): five-stage Verifier pipeline — Week 4 day 1-2 (spec §10) (#211) (0c69f99)
  • feat(pse): bottom-up enumerative search engine — Week 3 day 8 — FIRST END-TO-END SYNTHESIS (#210) (a525d5a)
  • feat(pse): in-process executor + observational-equivalence pruner — Week 3 day 3-7 (#209) (ce76a08)
  • feat(pse): search candidate-tree types — Week 3 day 1-2 (Program/InputRef/Const) (#208) (2edcc08)
  • feat(pse): add mask/logic + construction + color-constant primitives — Week 2 part D (21/56) (#207) (7d5a19a)
  • feat(pse): add object-detection DSL primitives — Week 2 part C (8/56) (#206) (de374b5)
  • feat(pse): add size/scale + spatial DSL primitives — Week 2 part B (12/56) (#205) (ed5c6e7)
  • feat(pse): add geometric + color DSL primitives — Week 2 part A (15/56) (#203) (5d7beea)
  • feat(pse): scaffold Program Synthesis Engine — Week 1 of Phase 1 spec (#202) (d355a84)
  • feat(bridge): risk caching, Planner-driven goal evaluator, HITL-resume fallback (#160) (075015b)

Fixes

  • feat(pse): Phase-2 Sprint-8 Track B — 100 % on every ARC corpus subset (#271) (1a85c36)
  • fix(audit): batch 3 — hashline prev_hash chain, canvas-static-registration, catalog drift trigger (#186) (7ecb84d)
  • fix(audit): batch 2 — owner-gate, paths cleanup, CI hygiene, pre-commit hook (#185) (906c160)
  • fix(audit): software-side fixes for #2 #3 #4 from 2026-04-29 workflow audit (#184) (62591d1)
  • fix(identity): close two latent observations from identity-test sessions (442 total) (#173) (dd7711d)

Other

  • chore(release): v0.96.0 — ARC-AGI-3 Pass (9a5f2c6)
  • docs(pse): record D3/D9/D12 + fix HO-count drift (#230) (075ecc9)
  • docs(pse): benchmarks.md scaffold + drift gate (D9) (#229) (40529ac)
  • docs(pse): hello-world tutorial + drift-test (D12) (#227) (d99bcbc)
  • docs(pse): auto-generated DSL reference + overview + architecture (D9) (#226) (626d1c9)
  • test(pse): adversarial security test scaffold — Week 5 day 3 (K4 hard gate) (#216) (ba97d80)
  • test(flutter): unit tests for SecurityProvider (+12 tests) (#201) (f4728da)
  • test(flutter): unit tests for SkillsProvider (+17 tests) (#200) (76b6977)
  • perf(flutter): migrate workflows-categories tab to CustomScrollView+SliverList.builder (#199) (5c30a3e)
  • test(flutter): unit tests for 4 more ChangeNotifier providers (+22 tests) (#197) (bf893af)
  • chore(tooling): pre-commit auto-install + verify_readme_claims in CI + JARVIS_HOME fallback fix (#194) (f91a75f)
  • test(flutter): unit tests for 4 more ChangeNotifier providers (+45 tests) (#198) (19338a9)
  • test(flutter): add unit tests for 4 simple ChangeNotifier providers (#196) (1d2b1c8)
  • perf(flutter): migrate chat-history search results + models catalog to lazy slivers (#195) (640554e)
  • refactor(gateway): extract lifecycle.py from gateway.py — split COMPLETE (6/6) (#193) (31c9770)
  • refactor(gateway): extract message_handler.py from gateway.py (split 5/6) (#192) (fdf4650)
  • refactor(gateway): extract post_processing.py from gateway.py (split 3/6) (#190) (596b6b6)
  • refactor(gateway): extract pge_loop.py from gateway.py (split 4/6) (#191) (4a0102b)
  • refactor(gateway): extract session_mgmt.py from gateway.py (split 2/6) (#189) (fcdd0fe)
  • refactor(gateway): extract message_utils.py from gateway.py — split step 1/6 (#188) (87a6349)
  • perf(flutter): migrate leads + monitoring eager ListView to CustomScrollView+SliverList.builder (#187) (e2d1c05)
  • refactor(config_routes): extract learning.py + autonomous.py + social.py — split complete (8/8) (#183) (7e8c079)
  • refactor(config_routes): extract evolution.py + infrastructure.py + workflows.py from _factory.py (7/8) (#182) (89efb34)
  • refactor(config_routes): extract ui.py from _factory.py (6/8) (#181) (6488076)
  • refactor(config_routes): extract skills.py from _factory.py (5/8) (#180) (a150ea1)
  • refactor(config_routes): extract security.py + governance.py from _factory.py (4/8) (#179) (d1d3dbf)
  • refactor(config_routes): extract session.py + monitoring.py from _factory.py (3/8) (#178) (cf6769c)
  • refactor(config_routes): extract system.py + config.py from _factory.py (2/8) (#177) (dfeb749)
  • refactor(config_routes): backbone — convert single-file module to package + route-inventory drift test (1/8) (#176) (3fde396)
  • chore(benchmark): archive in-process benchmark/ module — no callers, duplicates cognithor_bench/ (#175) (d49ac3a)
  • docs(audit): benchmark/ archive recommendation (closes last VERIFY-WIRING item) (#174) (7fdbd6d)
  • test(identity): Session 4 Phase 4 (FINAL) — state mgmt + kill switch + admin (28 tests, 438 total) (#172) (79a62ed)
  • test(identity): Session 4 Phase 3 — retrieve_memories + build_context_for_llm (21 tests, 410 total) (#171) (2f0c4e8)
  • test(identity): Session 4 Phase 2 — process_interaction flow (27 tests, 389 total) (#170) (37344bf)
  • test(identity): Session 4 Phase 1 — engine init + Genesis Anchors (20 tests, 362 total) (#169) (dbdc1ff)
  • test(identity): Session 5 — attention + temporal + local_store + merkle_batcher (147 new tests, 342 total) (#168) (3cd6208)
  • test(identity): Session 3 — vector_store + working_memory (50 new tests, 195 total) (#167) (42436ac)
  • test(identity): Session 2 — adapter + llm_bridge (32 new tests, 145 total) (#166) (e5f1fe1)
  • test(identity): Session 1 — memory + epistemic + input_sanitizer (81 tests) (#165) (a93283d)
  • docs(audit): scope identity/ test-coverage gap before writing tests (#164) (f141e30)

Full Changelog: v0.95.0...v0.96.0

Contributors

primitive
Assets 8
Loading