Files
make-cad/docs/updates/2.0.4b2.md
T
Ni Jingzhe c454a3af10 feat(skill): per-harness skillbuild toolchain replaces skill-pack (#18)
* feat(skill): replace skill-pack with per-harness conditional compiler

skillbuild compiles the skill tree into per-target harness builds:
- <!-- skill:if omp,!claude --> blocks include text only for matching
  targets; name-list conditions only, no else/nesting, fail-loud on
  unknown targets and unpaired/nested directives
- fence-aware: directive-looking lines inside code fences are literal
- layout is the filesystem; tree root configured in skillproj.toml
- post-build assertion: zero directive residue, SKILL.md frontmatter
  name check

Migration to the new model:
- docs/skill/ becomes the self-contained source root (SKILL.md,
  references/, generated api/stdlib/core/guides now under
  references/docs/; SDK_OVERVIEW/LICENSE/inspect are ordinary tree
  files)
- auto_docs_gen output redirected into the tree (docs/api ->
  docs/skill/references/docs/api, likewise stdlib)
- skills/simplecadapi/ is now the compiled omp target output;
  dist/generic/simplecadapi the generic target
- dropped skill-pack entry point (plus dangling evolve/make-export
  entries and their orphaned test), release workflow builds via
  skillbuild + tar; README updated

First real conditional block: harness reading paths in SKILL.md
(skill:// mount on omp vs disk-relative on generic).

Verified: 59 tests pass (skillbuild 31, auto_docs_gen 19, persistence
regression 9); both targets build; outputs residue-clean; full suite
collects (1376 tests).

* refactor(tools): auto_docs_gen moves out of the installed package

tools/ scripts are repo-checkout dev scripts, not packaged CLIs:
- auto_docs_gen.py now lives in tools/ beside the other generators;
  auto_tools/ package removed entirely
- path resolution simplifies: repo root = parent of tools/, fail-loud
  if pyproject.toml missing; site-packages/cwd fallback branches were
  dead code once the script stopped shipping in the wheel
- auto-docs-gen console script dropped from [project.scripts];
  release.yml and README invoke python tools/auto_docs_gen.py
- MANIFEST.in stale src/auto_tools include removed; CadQuery-free
  runtime scan no longer needs the auto_tools filter
- run_examples.py retired (user removal): README example-runner text
  updated to direct examples/ scripts; dangling runner-contract test
  (runpy.run_path into the deleted script) removed

Verified: 53 tests pass (auto_docs_gen + skill_build + ocp runtime
scan) and test_example_model_contract passes; regeneration from
tools/ is byte-identical (zero docs drift).

* refactor(skill): strip harness and build-tool content from the skill tree

The skill is plain harness-neutral content again:
- SKILL.md 'Reading skill files' section removed (the omp skill://
  mount block and its !omp complement were the only conditional
  blocks; the tree compiles identically for every target now)
- references/README.md: build-tool meta removed ('per-harness
  conditional compilation', skillbuild packaging/validation claims —
  link validation was never implemented, so the claim was false)
- live test now asserts the tree stays harness-neutral (zero blocks)
- skillproj.toml targets aligned with the user's output table
  (omp -> skills/simplecadapi-omp, opencode -> skills/simplecadapi-opencode;
  leftover 'generic' entry had no output and failed the build)
- compiled outputs are regenerable artifacts: skills/simplecadapi/
  untracked from git, skills/simplecadapi-*/ gitignored, release.yml
  and README build via skillbuild and tar the omp output directly
  (bsdtar has no --transform; inner dir keeps its suffixed name)

Verified: 49 tests pass; both targets build 372 files with zero
conditional blocks; no omp/conditional/directive residue in the tree
or outputs.

* chore(docs): drop stale architecture review and design notes
2026-08-22 20:20:49 +08:00

137 lines
6.2 KiB
Markdown

# SimpleCADAPI 2.0.4b2 Update Notes
[Chinese](2.0.4b2.zh-CN.md)
> **Beta release notice:** These changes are being validated before the final
> 2.0.4 release. Review reconstructed geometry and generated CAD documents
> before using them in production.
SimpleCADAPI 2.0.4b2 adds an Agent-oriented STEP/BREP reconstruction workflow.
The release turns the existing inspection and strict-comparison utilities into
a focused, framework-neutral tool surface for iterative reconstruction. It also
adds replayable interpolated B-spline profiles for freeform and periodic input.
## 1. Indexed STEP/BREP Inspection
`simplecadapi.inverse_engineer.brep` now builds a bounded, cached `BRepModel`
for each STEP input. Bodies, Faces, Edges, and Vertices receive deterministic
zero-based identifiers such as `body:0`, `face:12`, and `edge:4`, allowing an
Agent to carry local topology references across inspection calls.
The model summary reports validity, material-body counts, bounds, volume,
surface area, centroid, topology counts, and analytic surface/curve statistics.
Optional bounded groups expose repeated carrier parameters, canonical axes, and
adjacency signatures as evidence for a hypothesis; these groups do not infer a
pattern by themselves.
Entity inspection reports geometry parameters, bounds, adjacency, endpoint
differentials, and optional exact B-spline or Bezier definitions. Surface
definitions remain explicitly separated from trim-loop evidence so an
untrimmed carrier cannot be mistaken for a complete Face.
## 2. Framework-Neutral Agent Tools
The new `agent_tool_schemas()` registry exposes 17 JSON-Schema-validated tools,
and `call_agent_tool(...)` dispatches them without depending on a particular
Agent framework:
```text
get_model_summary inspect_entity
get_topology_neighborhood measure_relation
make_section extract_face_boundaries
probe_point render_region
compare_global_properties compare_boundary_distance
compute_material_difference compare_sections
build_difference_regions find_nearby_entities
compare_entities evaluate_result
compare_brep_strict
```
The `simplecad-brep tools` command prints the schemas. `simplecad-brep tool`
invokes one tool with an inline JSON object or an `--arguments-file`, so the same
contract is available to shell-driven Agents and external orchestrators.
## 3. Targeted Diagnosis and Acceptance Gates
Local query tools can expand a bounded topology neighborhood, measure supported
parallel/coplanar/coaxial/concentric/perpendicular/touching/tangent relations,
cut physical sections, extract ordered Face loops, and find the nearest indexed
entities to a point. Compact section and boundary modes preserve structural
evidence while omitting large sampled arrays.
Candidate diagnostics now include:
- global body, bounds, volume, area, centroid, and topology deltas;
- bidirectional sampled-boundary-to-exact-boundary distances;
- strict missing/excess material components from directional Boolean cuts;
- target/candidate section comparison on the same physical plane;
- ranked material or boundary difference regions and nearby entity lookup;
- local entity parameter and adjacency comparison.
`evaluate_result(...)` applies replay, BREP validity, material, bounds, volume,
area, and boundary gates in one report. Strict geometry-labelled incidence
comparison remains opt-in. Fuzzy Boolean results are diagnostic only and cannot
claim strict material equality.
## 4. Region Rendering Without Geometry Side Effects
`render_region(...)` renders selected stable entities and an optional topology
neighborhood over a translucent context model. Existing full-model
`render_step_views(...)` output remains available for synchronized isometric and
orthographic inspection.
Rendering now meshes a detached BREP copy. OpenCascade triangulation therefore
cannot alter the cached model's measured bounds or influence a later Agent
evaluation in the same process.
## 5. Replayable Interpolated B-Splines
The public modeling API adds:
```text
make_interpolated_spline_redge(...)
make_interpolated_spline_rwire(...)
make_periodic_spline_rwire(...)
```
Open and periodic curves interpolate their input points through the OCP kernel,
validate minimum point counts and duplicate-point tolerances, and retain their
construction data in graph metadata. Canonical model JSON records and strictly
replays interpolated edges, including expression-valued points. Periodic
profiles can be promoted into replayable Faces and Loft workflows.
## 6. Reconstruction Agent Workflow
The new reconstruction specification defines three explicit operating modes:
- **inspection-only reconstruction** restricts access to exact control arrays;
- **report-assisted reconstruction** permits targeted exact curve/surface data;
- **exact BREP transcription** permits complete carrier, trim, and topology
evidence while requiring the result to be labelled as transcription.
The workflow requires isolated inputs and outputs, a written feature hypothesis,
Sketch-first planar profiles unless a documented exception applies, fresh-process
model replay, valid BREP output, and measured final classification. Expensive
Boolean, dense boundary, and strict-topology checks are delayed until cheaper
evidence shows that a candidate is close enough to justify them.
## 7. Verification
Regression coverage includes stable entity identifiers, bounded schemas,
compact sections and boundary definitions, cross-model relation semantics,
material-difference strictness, difference-region reuse, evaluation gates,
region-rendering side effects, interpolated-curve replay, expression retention,
and periodic-profile Loft replay.
A real STEP round trip was also exercised with a replayable drilled solid. The
target and replayed STEP passed strict material, boundary, validity, and
geometry-labelled incidence gates after both full-model and highlighted-region
rendering.
## Related Documentation
- [Reconstruction Agent test specification](../skill/references/docs/guides/reconstruction-agent-test-prompt.md)
- [STEP BREP reverse-engineering guide](../skill/references/inspect/brep-reverse-engineering.md)
- [SimpleCADAPI Agent Skill BREP reference](../../skills/simplecadapi/references/inverse_engineer/brep-reverse-engineering.md)
- [Previous 2.0.4b1 update notes](2.0.4b1.md)