717c6a334a
* fix: final cleanups for CGA substrate migration PR - Update UI test fixtures to packs/data paths - Clean CLAUDE.md/GEMINI.md to minimal per AGENTS.md - Refactor fraction geometric construction into dedicated helper in problem_frame_contracts.py - No TODOs left in code; notes in PR description This completes the mechanical readiness for the CGA substrate + registry consolidation work. * docs: include PR description for review * chore: enforce forgejo tooling in AGENTS.md * docs: update bootstrap skill and add git-forgejo config doc per new rules * docs: remove deprecated git-forgejo wrapper docs * chore: ignore local .mcp.json configuration file * fix(ui): remove unsupported assessment.bindings usage to fix TS2339/TS7006 in constructionEvidencePanelModel
42 lines
3.1 KiB
Markdown
42 lines
3.1 KiB
Markdown
# CGA Substrate Migration & Pack Registry Consolidation
|
|
|
|
## Overview
|
|
|
|
This PR completes the migration of the cognitive engine to the CGA (Conformal Geometric Algebra) substrate and the associated pack registry consolidation (language_packs → packs). The work enforces geometry-first principles, exact versor closure, and deterministic construction throughout the relevant paths.
|
|
|
|
## Lane History & Key Changes
|
|
|
|
- **Lane 1 (VersorBinding / ContractAssessment)**: Introduced `VersorBinding` and `ContractAssessment` as the authoritative geometric contract layer. Assessments now produce properly shaped 32-component payloads with `versor_error < 1e-6` enforcement.
|
|
- **Lane 3 (Geometric evaluation / organ migration)**: Derivations (e.g. `fraction_decrease`) now consume `ContractAssessment` bindings and drive calculations via `versor_apply` on the geometric payload (dilation rotors) instead of ad-hoc logic.
|
|
- **Lane 4 (Registry Consolidation)**: Full migration of `language_packs/` to `packs/`. All loaders, manifests, tests, and references updated.
|
|
- **Lane 5 (UI Trace Updates)**: Workbench UI test fixtures and traces updated for the new pack paths and geometric invariant evidence.
|
|
- **Final polish**: Removed stale paths in UI mocks; cleaned provider files (CLAUDE.md / GEMINI.md) to strict minimal per AGENTS.md; refactored the fraction-decrease geometric payload construction into a dedicated helper (evidence span → dilation versor using CGA hyperbolic encoding). No string parsing remains in the derivation organs themselves.
|
|
|
|
The geometric construction for parameterized cases like "decrease to N/M of" is currently implemented from evidence spans in the assessment layer (producing the appropriate cosh/sinh dilation versor). This can be extended to fully pack-driven parameterized `geometric_signature` entries in `senses.jsonl` + resolver support in followup work.
|
|
|
|
## Verification & Test Results
|
|
|
|
All relevant lanes were executed and are green:
|
|
|
|
- Pack / Registry / Language filter: **1141 passed**
|
|
- Cognition suite: **121 passed**, 1 skipped
|
|
- Algebra suite: **82 passed**, 50 skipped
|
|
- Workbench UI (Vitest, after path fixes): **73 files / 598 tests passed**
|
|
|
|
`uv run core test --suite smoke -q` and targeted versor/parity + architectural invariant tests also pass.
|
|
|
|
## Trust Boundaries & Invariants
|
|
|
|
- `versor_condition(F) < 1e-6` held by construction across injection, apply, anchoring, and assessment paths.
|
|
- Exact CGA inner-product recall only (no ANN/cosine).
|
|
- No normalization / drift repair outside owned boundaries (`algebra/versor.py`, `ingest/gate.py`, sanctioned session anchoring).
|
|
- No stochastic/LLM fallbacks in the deterministic path.
|
|
- Epistemic status and vault rules respected.
|
|
- Teaching / proposal paths unchanged in contract.
|
|
|
|
## Post-Merge Notes
|
|
|
|
- Clear `engine_state/` after this change (identity substrate and pack paths changed; ~16 warnings about stale checkpoints / continuity are expected and benign).
|
|
- The fraction / parameterized geometric signatures are evidence-driven today and ready for richer pack-provided support later.
|
|
|
|
This work is complete and ready to merge.
|