Next Gen STEP AP242 Native-ish BREP Geometric Modeling Kernel
AETHERIS-SHEETMETAL-M4 evidence
M4 reaches meaningful bounded production capability, not full commercial parity. The shared line/arc contour kernel, exact profile composition, stable semantic paths, template generators, and exact single-corner relief cases work through real formed/flat/STEP paths. CTC-03 remains NeedsReview, and simultaneous multi-corner relief plus nested-chain arrangement is the isolated next blocker.
Delivered
- Shared
PlanarContour2outer/inner loop IR with stable IDs, plane frame, native line/arc/circle geometry, and provenance. - Public bounded intersection, split, trim, validation, explicit-side offset, and known-topology arrangement operations.
- Exact flat region, cut, relief, and (when admitted) composed blank contours; analytic arcs survive SVG and flat STEP.
- Typed
Open,Mitered,RectangularRelief, andRoundReliefcorner records. - Canonical base/flange/formed/flat Concept Paths and
aetheris sheetmetal pathsinspection. - Capability-aware invalid-member diagnostics.
- Module-owned
LBracket,UChannel, andFourWallTraygenerators that emit ordinary Sheet Metal semantics. - PSU enclosure authoring fixture and semantic-only DFM repair fixture.
- Exact-blank/relief and minimum-flange DFM rules with deterministic suggestions.
Evidence map
- Contour audit and API
- Corner and relief behavior
- Concept Paths and authoring
- Templates and dogfood
- CTC-03 formed comparison
- CTC-03 flat comparison
- DFM
- Validation, performance, determinism
Retained generated artifacts:
psu-enclosure-formed.step,psu-enclosure-flat.step,psu-enclosure-flat.svgctc03-m4-formed.step,ctc03-m4-flat.step,ctc03-m4-flat.svg
Architecture verdict
Does Aetheris now have a sufficiently exact 2D contour/corner kernel? Yes for bounded common line/arc profiles, open/miter topology, individual rectangular/round corner reliefs, cuts, and ordinary exact blank composition. No for the full M4 ambition: multiple interacting relief removals on four-wall/nested hostile graphs can still be rejected by arrangement angular ordering, and curved formed relief walls are not yet emitted.
Are Concept/Concept Path ergonomics good enough? Yes for the current authored family. Main.Front, FrontWall.Outer, FrontWall.Bend, and their Flat.* counterparts are stable and inspectable; wrong members and capabilities produce useful diagnostics. The same resolved Profile/contour substrate is used outside Sheet Metal.
Do templates demonstrate parameter-first authoring? Yes. L brackets, U channels, and four-wall trays lower through ordinary semantics and preserve path shape across specializations. They are a C# module API today, not yet first-class generic declarations in Firmament source.
Largest blocker: robust multi-corner seam/relief arrangement and corresponding exact formed trim-wall materialization, especially when several corner removals meet nested flanges or ambiguous historical contours. This is narrower than M3’s generic “exact 2D kernel” blocker and is backed by typed sheetmetal-exact-blank-contour diagnostics on the M3 four-relief tray and CTC-03.