Commit graph

851 commits

Author SHA1 Message Date
Shay
03c87e680b fix(core): lock python and extend rust cga surface 2026-07-02 15:20:14 -07:00
Shay
8b12423dec
fix: green test-fast suite, consolidate ADR graph under docs/adr, and complete governance cohesion anchors
- Green make test-fast suite: fixed exemplar corpus issues, proposal validation, atomic state checkpointing (scheme=2), turn-scoped state leakage in ChatRuntime.chat
- ADR corpus consolidation: migrated all ADRs to docs/adr/, appended ADR-0225 governance cross-reference anchors to foundational ADRs (0001, 0027-0029, 0055-0057)
- Pack definitional closure: fixed en_arithmetic_v1 glosses.jsonl JSON error, updated manifest checksum, marked en_core_syntax_v1 definitional_layer: false
2026-06-30 17:56:12 -07:00
Shay
c6263f5a91
refactor(kernel): split ProblemFrame builder phases (#919)
* refactor(kernel): add ProblemFrame extraction phase module

* refactor(kernel): add ProblemFrame proposal phase module

* refactor(kernel): add ProblemFrame mention phase module

* refactor(kernel): add ProblemFrame bound relation phase module

* refactor(kernel): reduce ProblemFrame builder to phase orchestration

* test(kernel): pin ProblemFrame phase boundaries

* test(kernel): keep unary delta smoke within supported slice
2026-06-25 16:16:04 -07:00
Shay
3f15bc3b6b
test(kernel): pin proposal-first assessment boundary (#915) 2026-06-24 21:38:32 -07:00
Shay
1953a1d59d
feat(bench): add Apple UMA demo package builder (#913)
* feat(bench): add Apple UMA demo package builder

* test(bench): cover Apple UMA demo package builder

* docs(outreach): add Apple UMA demo package runbook

* feat(bench): harden Apple UMA demo package workflow
2026-06-24 15:51:45 -07:00
Shay
f8d882be48
feat(workbench): add Apple UMA report read model (#911)
* feat(workbench): add Apple UMA report read model

* test(workbench): cover Apple UMA report projection

* feat(workbench): expose Apple UMA report route
2026-06-24 14:47:36 -07:00
Shay
cd7579be32
feat(bench): wire MLX exact recall into Apple UMA report (#910) 2026-06-24 14:07:17 -07:00
Shay
b30716a19c
feat(bench): add MLX exact CGA recall experiment (#909)
* feat(bench): add MLX exact CGA recall experiment

* test(bench): cover MLX exact recall experiment contracts

* docs(bench): document MLX exact recall experiment

* docs(bench): record MLX experiment integration status

* docs(bench): add MLX local validation handoff
2026-06-24 13:53:50 -07:00
Shay
99a107dd39
feat(rust): scalar Cl(4,1) zero-copy input boundary (ADR-0235 PR C) (#907)
Replace extract_f32_slice list conversion on geometric_product, cga_inner,
versor_condition, versor_apply_with_closure, and versor_apply_with_closure_f64
with PyReadonlyArray1 zero-copy views. Wrong shape, dtype, and non-contiguous
layouts fail loudly. Update Apple UMA benchmark truth table and claim audit
after parity gates pass.
2026-06-24 13:11:38 -07:00
Shay
daa13684f8
feat(bench): Rust-enabled Apple UMA baseline report lane (PR B) (#906)
Add rust_backend_status helper, backend_status report fields, and
rust_backend_notes in claim safety audit. Improve diffusion_step skip
reasons and markdown backend summary. Document core_rs install and
CORE_BACKEND=rust activation in docs/benchmarks/apple-uma-rust-baseline.md.

Regenerate seed report under honest Python fallback (core_rs unavailable
locally). No scalar Rust binding changes.
2026-06-24 12:56:57 -07:00
Shay
7132997511
feat(bench): Apple Silicon UMA mechanical sympathy benchmark (#904)
* feat(bench): add Apple Silicon UMA mechanical sympathy benchmark

Engineering-grade reproducible benchmark measuring exact CGA recall,
Cl(4,1) scalar algebra, FrameVerdict TTFV, array_codec replay, and
honest Python/Rust copy/zero-copy boundaries. Runs without Rust;
skips Rust-only tracks with explicit reasons. Includes claim-safety
audit, CLI integration (core bench --suite apple-uma), and outreach brief.

* fix(bench): patch apple-uma report paths, decode timing, CLI --report

- Use repo-relative report_path in JSON metadata (no absolute paths)
- Measure decode_array only; precompute encode payload before decode bench
- core bench apple-uma --report writes exactly to PATH; --write-report for defaults
- Add final newlines; regenerate seed report
2026-06-24 12:36:02 -07:00
Shay
8c3898c1e1
feat(workbench): wire Trace practice evidence route (#900)
* feat(workbench): route practice evidence through trace endpoint shim

* test(workbench): lock trace practice API route

* fix(workbench): remove stray practice API test token
2026-06-23 15:57:31 -07:00
Shay
b96c556701
feat(workbench): add sealed practice evidence read model (#899)
* feat(workbench): add practice evidence read model

* feat(workbench): add practice evidence endpoint seam

* feat(workbench): persist practice evidence field when present

* test(workbench): cover practice evidence read model seam
2026-06-23 15:41:13 -07:00
Shay
ec17fb7cdb
feat(workbench): persist construction evidence in trace journal (#896) 2026-06-23 13:58:12 -07:00
Shay
d7a809e82a
feat(workbench): wire construction evidence into Trace route (#895) 2026-06-23 13:28:17 -07:00
Shay
4c23aa4b00
feat(workbench): prepare construction evidence endpoint seam (#893)
* feat(workbench): add construction endpoint handler seam

* test(workbench): cover construction endpoint handler seam

* feat(workbench-ui): add construction evidence endpoint helpers

* test(workbench-ui): cover construction endpoint helpers
2026-06-23 10:11:21 -07:00
Shay
2c8258fe1a
feat(workbench): add UI catch-up evidence scaffolding (#891)
* feat(workbench): add construction evidence read model scaffolding

* feat(workbench-ui): add construction evidence TS types

* feat(workbench): add construction evidence journal projector

* test(workbench): cover construction evidence read model

* feat(workbench-ui): add construction evidence view helpers

* test(workbench-ui): cover construction evidence view helpers

* feat(workbench): add generalization audit evidence scaffolding

* test(workbench): cover generalization audit evidence view

* feat(workbench-ui): add generalization evidence TS types

* feat(workbench-ui): add generalization evidence view helpers

* test(workbench-ui): cover generalization evidence helpers

* feat(workbench): add proposal artifact authority scaffolding

* test(workbench): cover proposal artifact authority rules

* feat(workbench-ui): add proposal artifact TS types

* feat(workbench-ui): add proposal artifact view helpers

* test(workbench-ui): cover proposal artifact view helpers

* feat(workbench): add demo narrative evidence scaffolding

* test(workbench): cover demo narrative scaffolding

* feat(workbench-ui): add demo narrative TS types

* feat(workbench-ui): add demo narrative view helpers

* test(workbench-ui): cover demo narrative view helpers

* fix(evals): resolve discovery_candidates.jsonl via EngineStateStore
2026-06-23 10:09:45 -07:00
Shay
f8fe6e22ca test(evals): strengthen GSM1K repo-root guard 2026-06-23 07:57:34 -07:00
Shay
d34883b61a feat(evals): implement safe metadata and fail-closed evaluator check for GSM1K 2026-06-23 07:52:32 -07:00
Shay
db081fa5b3 feat(evals): add GSM1K local audit adapter 2026-06-23 07:40:53 -07:00
Shay
f823eb437b feat(evals): add generalization audit runner skeleton 2026-06-23 07:23:43 -07:00
Shay
ed8c0b0e43 feat(evals): verify local generalization benchmark cache 2026-06-23 07:03:56 -07:00
Shay
b51575a075
Merge branch 'main' into feat/generalization-manifest-policy-linter 2026-06-23 06:40:35 -07:00
Shay
83ba9a2e9b feat(evals): implement manifest schema and policy linter (#886) 2026-06-23 06:30:12 -07:00
Shay
fe4e2e6e33 feat(kernel): add quantity-entity replay target shell 2026-06-23 06:13:28 -07:00
Shay
cead96596c feat: implement quantity_entity_binding_candidate.v1 operator
- Implement CandidateOperatorPolicy and cue dataclass for quantity entity binding

- Add build_quantity_entity_binding_candidate constructor with strict refusal checks

- Bump CANDIDATE_OPERATOR_SET_VERSION to candidate_operators.v2

- Update exact export and static table shape tests

- Reconcile stale static guards across binding replay adapters

- Provide full explicit determinism and failure mode test coverage
2026-06-23 05:37:08 -07:00
Shay
df72c71182 test(kernel): reconcile bound evidence spine static guards 2026-06-23 00:28:25 -07:00
Shay
b9a6eb9676 fix(kernel): fail-closed bound trace structural validation
Return PracticeTraceRefusal before dereferencing bindings or operator
results when structural type/shape checks fail. Guard seal_bound zip
against binding_result_count_mismatch. Add five refusal-without-exception
tests for malformed bound inputs.
2026-06-23 00:11:14 -07:00
Shay
484e8ff9fe docs+kernel: support bound practice episode sealing
Extends SealedPracticeTrace for externally bound candidate episodes per
ADR-0233. Adds candidate_attempt_binding_ids to trace identity, bound trace
input/seal entry points, and end-to-end bound episode tests without mutating
GeometricSearchRun.candidate_attempts.
2026-06-23 00:04:26 -07:00
Shay
708237525c test(kernel): cover bound replay adapter input 2026-06-22 23:39:41 -07:00
Shay
76bb756519 test(kernel): cover run-attempt binding shell 2026-06-22 23:16:02 -07:00
Shay
d0d7df3857 fix: add generate/run_attempt_binding.py + complete real tests (no placeholders, full coverage) 2026-06-22 23:05:42 -07:00
Shay
fbf9b18868 fix: replace placeholder with full ADR-0232 implementation of run_attempt_binding (generate/ + tests/) 2026-06-22 23:03:04 -07:00
Shay
65505e059d feat(kernel): implement inert run-attempt binding shell per ADR-0232 2026-06-22 22:57:22 -07:00
Shay
1d9e40602c feat(kernel): implement missing-role candidate operator shell
Implements ADR-0231's inert missing_role_candidate.v1 operator shell for
direction_unbound residuals under unary_delta_transition. Adds deterministic
operator-set identity, candidate reconstruction with split digests,
CandidateAttempt-compatible output, typed refusals, and isolation tests.
2026-06-22 19:27:41 -07:00
Shay
a1de11e0e3 feat(kernel): implement inert SealedPracticeTrace shell
Adds immutable practice-trace input, sealed-trace, and trace-refusal
records with deterministic identity-chain validation after the Contract/
Proof Replay Adapter, as authorized by ADR-0230.
2026-06-22 18:20:52 -07:00
Shay
feb36cfdbb feat(kernel): implement diagnostic replay adapter shell
Adds the ADR-0229-authorized Contract/Proof Replay Adapter diagnostic
boundary after GeometricSearchRun: deterministic input/result/refusal
records, candidate_organ/contract_replay_target binding via static
allowlist, strict run/attempt/candidate identity validation, ADR-aligned
replay_disposition on refusals, vacuous-proof closure only with explicit
schema declaration, and injection-based disposition classification
without candidate generation, search execution, repair, answer
production, or runtime mutation.
2026-06-22 15:18:22 -07:00
Shay
93ba988c1e feat(kernel): implement inert GeometricSearchRun trace shell 2026-06-22 14:09:52 -07:00
Shay
fcac436f7d feat(kernel): implement diagnostic-only ComputeBudgetDecision 2026-06-22 13:13:49 -07:00
Shay
7b2456d402
feat(kernel): implement diagnostic-only SearchGateDecision
Implements a diagnostic-only SearchGateDecision adapter over ContractResidual records.

Preserves the authority boundary:
ContractAssessment -> ContractResidual -> SearchGateDecision.

SearchGateDecision is not proof, repair, budget, search, serving, candidate generation, or mutation authority.

Validation reported on head 1d68292c7f:
- tests/test_search_gate.py: 23 passed
- tests/test_contract_residuals.py: 75 passed
- adjacent contract/proposal tests: 48 passed
- ruff: passed
- compileall: passed
- smoke: 108 passed
- git diff --check: clean
2026-06-22 11:36:05 -07:00
Shay
8b0edb6915 test(kernel): cover ContractResidual projection 2026-06-22 08:07:27 -07:00
Shay
59a919e347 fix(kernel): complete unary-delta conformance alignment 2026-06-21 22:07:12 -07:00
Shay
61964c5cd4 fix(kernel): align unary-delta implementation with conformance decision 2026-06-21 20:19:53 -07:00
Shay
032c05f806 test(kernel): harden unary-delta isolation after proposal seam 2026-06-21 06:35:48 -07:00
Shay
0f6c51d0d4 test(kernel): pin missing-object unary-delta confuser 2026-06-21 05:48:11 -07:00
Shay
b6bde12e1e feat(kernel): introduce diagnostic unary-delta proposal seam 2026-06-21 05:39:38 -07:00
Shay
ca83d3846c
fix(kernel): harden quantity-entity proposal seam (#853) 2026-06-20 21:49:26 -07:00
Shay
5310bd6f3c
feat(kernel): introduce diagnostic quantity-entity proposal seam (#851) 2026-06-20 20:57:08 -07:00
Shay
a145b7c3d6
feat(kernel): implement ProblemFrame proportional-change closure (#846)
* chore(kernel): update adequacy report, fix lint warnings, and record failure proposal

* chore(ci): refresh PR checks

* chore(ci): remove PR check refresh marker
2026-06-20 19:19:07 -07:00
Shay
a266324eea
refactor(kernel): fence legacy assessment-backed proposal adapter (#844)
* test(kernel): reject migrated proposal adapter paths

* refactor(kernel): fence legacy assessment-backed proposal adapter
2026-06-20 16:53:21 -07:00