Commit graph

3 commits

Author SHA1 Message Date
Shay
fa2712ebd7 feat(realizer): extend to all 13 English v1 constructions
Engineer the deterministic realizer to handle negation, conjunction,
disjunction, embedded clauses, relative clauses, quantification, tense,
and aspect — covering all 13 grammatical-coverage v1 constructions.

- generate/morphology.py: rule-based English inflection (past, participle,
  base form) for seed vocabulary predicates
- generate/templates.py: match-case inflection dispatch for tense/aspect/negation
- generate/graph_planner.py: add CONJUNCTION, DISJUNCTION, COMPLEMENT, RELATIVE
  relations; add grammatical feature fields to ArticulationStep
- generate/realizer.py: compound construction handling via graph edge traversal

grammatical-coverage eval: dev=100%, public v1=100% (from baseline of 24%/19%).
2026-05-16 05:55:49 -07:00
Shay
eb30c75810 feat: Full Proof — surface realizer join, Rust diffusion parity, benchmark harness
Surface realizer join: pulse output_versor → vault recall → ground_graph fills
<pending> obj slots with recalled words → realize_semantic produces deterministic
sentences. PulseResult replaces bare word list. Every intent type surfaces.

Rust backend parity: unitize_f32 (exponential-map with boost/rotation blade
distinction) and graph_diffusion_step now in core-rs. Python dispatches through
algebra.backend, falls back transparently. 37x speedup on 200-step diffusion.

Benchmark harness (core bench): determinism (100% trace stability), latency
(~150ms median), backend speedup, versor closure audit (0 violations across all
intermediate states), convergence proof (41/45 exact, 4 bounded oscillation),
realizer coverage (8/8 intent types).

Proof property tests (31 tests): Rust/Python parity, pulse determinism across
prompts, V3 convergence for 10+ topologies, coupled V4 output validity, realizer
coverage per intent, versor closure at every intermediate step.

CLI: core pulse, core bench, core test --suite pulse, core test --suite proof.
Fix test_correction_pulls_toward_target (diffuse first, then correct).
2026-05-15 17:39:14 -07:00
Shay
8dcc26581a feat: add intent-proposition graph comprehension layer
Implements the dialogue understanding pipeline:
  prompt -> dialogue intent -> proposition graph -> articulation target

New modules:
  - generate/intent.py: rule-based classifier (7 intent tags + UNKNOWN)
  - generate/graph_planner.py: immutable PropositionGraph DAG, topological
    walk to ArticulationTarget with rhetorical moves

Tests cover definition, cause, comparison, correction with prior-turn
linking, and deterministic serialization.
2026-05-14 19:52:57 -07:00