picogk

name

picogk

description

Use when working on the silken_net Code-as-CAD surface — PicoGK voxel/SDF geometry generation in tools/cad/ (the Ti-coin in-vitro coupon, the Zone-1 gyroid anode, the CartesianGyroid SDF, future PEEK Radome) and the CEM (Computational Engineering Model) manifests that drive them. Knows the non-obvious gotchas — render lattices via `new Voxels(IImplicit, BBox3)` + BoolIntersect NOT voxBounding.voxIntersectImplicit (uncatchable native OpenVDB abort at fine voxel on thin bored parts), headless `Library.Go(voxel, task, bEndAppWithTask:true)` not the stale v1.6 `new Library()`, ImplicitRadialGyroid axis-singularity → cartesian gyroid, dimensionless gyroid wallParam + voxel-dependent porosity (MEASURE it), the voxel-resolution floor, ImplicitUsings-ENABLED for vendored LEAP source — and the local-verify discipline (`dotnet run -- verify` → metrics.json, exit 0/1). Routes to 01_02 §6 (PicoGK stack + Noyron methodology) + 01_01 §5 (anchor geometry) + tools/cad/README, does not restate. Examples: "generate the gyroid anchor", "add a CEM part / per-species SKU", "change anchor Ø / porosity / pore period", "why does the anchor crash with 'outside value 0'", "add the radial pore gradient", "build the Ti-coin STL", "set up the .NET CAD project".

PicoGK Code-as-CAD (tools/cad)

Navigation aid + non-obvious gotchas. The SSOT is the docs + code + tools/cad/README.md below — this skill points, it does not restate (so it can't drift). Verify a fact at its home before trusting a summary. Methodology (per LEAP 71 Noyron): a CEM is a deterministic algorithm, not generative ML — an agent writes the generator, the generator computes the geometry.

SSOT Documents — Read These First

Document	What it covers
tools/cad/README.md	Operational home: layout, local-verify recipe, the FULL gotcha list, license
docs/01_02_Ti_6Al_4V_Metallurgy_and_DMLS.md §6	PicoGK stack (.NET 9, PicoGK 2.2 NuGet + ShapeKernel/LatticeLibrary submodules), Noyron methodology, real API, honest caveats
docs/01_01_Coaxial_Gyroid_Topology_and_PEEK.md §5/§6	Anchor geometry: gyroid 65% porosity, pore gradient 300→100µm, Gibson-Ashby isoelasticity, Ti-coin Stage-2 coupon (A=2cm²+eyelet)
docs/00_07_Action_Plan_Tracker.md HW.1 / HW.33	Build state + the anchor geometry audit (founder decisions: radial gyroid (б), Ø11; open gaps: PEEK/hole chain, FEA)
docs/00_08_Beyond_TRL9_Planetary_Roadmap.md §1.3	Cross-biome 5-SKU (pine/oak/broadleaf/mangrove/tropical)
docs/00_02_AI_Native_Engineering_and_TRL.md §4a	Code-as-CAD vs generative-AI; In-Silico for the Hardware stream
tools/cad/docs/drawings_program.md	Engineering-drawing program: CEM-native DXF (netDxf) + SVG, the ASME-Y14.5≠projection fix, lattice-as-inspection-card, phased §7 rollout
extern/.../README_ImplicitLibrary.md	LEAP's own implicit/TPMS guide (splitting logic, modular workflow for the graded v2)

Source Files

File	Role
tools/cad/cem/*.json	CEM manifests — the Git-SSOT parameter inputs (kind discriminator: ti_coin, anchor_zone1, mechanical_lock, cathode_flange, radome, zone2_sleeve, anchor_assembly, anchor_axial_stack)
tools/cad/src/SilkenCad/Program.cs	CLI dispatch (smoke/build/verify/scan/draw/render/section) + RunHeadless (the Library.Go wrapper); draw is pure-managed (no Library.Go), render/section drive the native viewer
tools/cad/src/SilkenCad/Cem.cs	CEM records + JSON parse (snake_case). Engineering-drawing PMI lives here: optional ToleranceSpec (fits / Lamé-µm / GD&T datums) + NotesSpec (material/process/post-process/surface/coating-restriction/lattice-spec/inspection) on each part record — Noyron-native SSOT, fed to draw
tools/cad/src/SilkenCad/Drawing.cs	CEM-native engineering drawings (draw <cem>): SVG (human) + DXF via netDxf (CAD-native factory deliverable, opens in AutoCAD/Fusion) — no PDF (never built; drawings_program §3). Pure-managed string/entity build, no Library.Go. Consumes the CEM ToleranceSpec/NotesSpec (zero hard-coded eng-text); DrawingStandard param (ISO 1st-angle default / ASME). Shipped kinds = ti_coin + cathode_flange (§7 rest deferred — docs/drawings_program.md). ⚠️ gotcha #11
tools/cad/src/SilkenCad/TiCoin.cs	Ti-coin coupon — BaseCylinder disc + BaseRing eyelet, BoolAdd
tools/cad/src/SilkenCad/Zone1Anode.cs	Zone-1 anode + CartesianGyroid:IImplicit (the from-scratch SDF) + Anode() render path
tools/cad/src/SilkenCad/Validation.cs	golden-metrics via Voxels.CalculateProperties (porosity needs an envelope ref) + reuses LEAP Measure.fGetSurfaceArea
tools/cad/src/SilkenCad/Connectivity.cs	ARCH.25 two-phase topological audit — SDF-sample + 6-conn flood-fill (open/closed-pore, percolation, solid-island, specific-surface); pure-managed, display-less xUnit
tools/cad/src/SilkenCad/MechanicalLock.cs	§4.3 mechanical lock — MechanicalLockShank asymmetric ratchet-barb + DIN-471 groove SDF on the shank (Zone-1 solid monolithic / Zone-3 channelled); render = solid voxConstruct + thin barb-ridge BoolAdd; self-support metric (Noyron manufacturing-awareness)
tools/cad/src/SilkenCad/CathodeFlange.cs	Деталь 3 — Zone-3 cathode flange (Ø25): reuses the §4.3 shank/barbs via ShankCem + radial bayonet lugs + bus channel + O-ring groove
tools/cad/src/SilkenCad/Radome.cs	Деталь 4 — PEEK radome v2c (Ø25): hollow dome + shield bell + bayonet socket + PCB cavity + O-ring groove (gotcha #9 INVERTED — hollow is intended)
tools/cad/src/SilkenCad/Assembly.cs	Capsule-end mate-audit (Деталь 3↔4, 02_02 §4.4) — bayonet datum via voxApplyTransformation lift + Z/MATE-Ø/RF mismatch + skirt/inboard candidates; pure mate-math (xUnit) + render interference (verify)
tools/cad/src/SilkenCad/Zone2Sleeve.cs	Деталь 2 — Zone-2 PEEK sleeve (bore Ø11 / OD Ø15 / 50 mm): plain hollow tube via BasePipe (smooth bore; hex + flange-shoulder deferred, bench-gated)
tools/cad/src/SilkenCad/AxialStack.cs	Full axial stack mate-audit (Зони 1↔2↔3↔4, 02_02 §4.5) — press-fit interference + insertion budget + span; reuses Assembly.Build + Zone-1 envelope; pure mate-math (xUnit) + render (verify)
tools/cad/src/SilkenCad.Leap/	vendored LEAP source compiled in (ImplicitUsings ON, warnings relaxed — not ours)
tools/cad/extern/LEAP71_{ShapeKernel,LatticeLibrary}	git submodules (source-only; not on NuGet)

Voxel/triangle counts + porosity drift with voxel size — don't hardcode them ([[feedback_no_volatile_counts]]); MEASURE via verify.

Gotchas Not Obvious From Docs

1. Render lattices via new Voxels(IImplicit, BBox3) + BoolIntersect — NOT voxBounding.voxIntersectImplicit(impl). The convenience method yields a malformed (background-0) OpenVDB level set at fine voxel on a thin bored part → an uncatchable native abort (libc++abi … ValueError: expected grid A outside value > 0, got 0 — a process kill, not a .NET exception). The ctor path runs at 0.1mm; voxIntersect's ceiling was ~0.4mm. (Zone1Anode.Anode.)
2. ImplicitRadialGyroid is degenerate near r=0 (cylindrical singularity) — fine for large annular parts, empty grid for a small rod near its own axis. Use a cartesian gyroid (uniform everywhere, still bicontinuous → honors founder decision (б), HW.33).
3. Headless = Library.Go(voxel, task, bEndAppWithTask:true), not new Library() (the v1.6 headless pattern → "relies on Library::Go" abort in v2.2). It briefly inits a Metal/GL viewer then closes with the task — so CI needs a display (macOS runner / xvfb).
4. Gyroid wallParam is DIMENSIONLESS (gyroid eq ∈ [-1.5, 1.5]), not mm: |eq| < 0.5*wallParam ⇒ solid; a clean wall needs wallParam ≪ amplitude. Porosity is voxel-dependent → MEASURE it (a coarse voxel under-resolves voids → falsely high porosity; at 0.1mm the Ø11 anode reads 67.6% ≈ the 65% target, vs 21–28% at 0.4–0.5mm).
5. Voxel-resolution floor + gradient distortion — sub-100µm pores need voxel ~0.03mm → ~10⁹-voxel grids. The Ø11 anode renders cleanly at 0.1mm (pores ~2.5mm); realistic 300→100µm pores are the HW.33 ceiling (and un-printable at 65%: SLM wall ~200µm → min pore ~1.2mm). Separately, a continuous radial gradient distorts above ~0.8× period ratio (non-Eikonal |∇eq|∝f; the ∇f·coord term collapses porosity 67→42% at 2.5→1.3mm) → keep continuous gentle, or use ZonedGyroid (stepped) for strong contrast. Per-shell porosity = cumulative-diff (thin rings under-count metal on distorted geometry).
6. ImplicitUsings ENABLED for src/SilkenCad.Leap — vendored LEAP source relies on implicit using System / System.Collections.Generic. Disabling → 200+ CS0246. Strict knobs (warnings-as-errors, nullable) are ON for OUR code (src/SilkenCad, tests/), OFF for the vendored project (mirrors the firmware/extern vs firmware/common split).
7. DOTNET_ROOT=$HOME/.dotnet if the apphost binary is run directly (native runtime is in the non-standard ~/.dotnet install → libhostfxr.dylib not found otherwise); dotnet run is unaffected.
8. Connectivity needs WALL-resolution, not pore-resolution (Connectivity.cs, ARCH.25) — the gyroid wall is only ~period/10 thick; sampling the SDF coarser than ~wall/2 fragments thin walls into false solid "islands" (measured 24–61% disconnected at 0.3mm → ~0% at period/16). Pore-phase metrics (open/percolation) are fine at any step → SampleAnchor ties the step to the finest period. And a sheet gyroid is tricontinuous → PoreClusterCount==2 is a topology FACT, not a defect (don't gate on it). NB Ex_ImplicitGyroidGenus is misleading (renders a gyroid on a genus-torus shape; computes no genus) — LEAP exposes no connectivity, but Measure.fGetSurfaceArea (surface) + fGetVolume exist and are reused, not re-implemented.
9. A FILLED (solid) body must come from ShapeKernel voxConstruct, NOT new Voxels(IImplicit, BBox3). The SDF ctor builds a narrow-band field (voxels near the surface only), so a solid core falls outside the band and renders as a hollow shell (MechanicalLock.cs: a Ø11 shank measured ~17 mm³ vs ~1700 expected). The gyroid dodges this only because it is thin-walled everywhere. Pattern: solid = BaseCylinder().voxConstruct(); thin features (barb ridges) = SDF BoolAdd (the Ti-coin split). The verify solidity gate (volume > 0.8·annulus) guards the regression.
10. render/section (presentation) need a display + the screenshot is TGA. The PicoGK native viewer (Library.oViewer()…RequestScreenShot inside Library.Go) renders only with a display (macOS desktop OK; headless CI = Library.Go SIGSEGV/139). The frame is TGA regardless of a .png name → convert TGA→PNG via sips (mac). The managed Viewer exposes qOrientation + view-cube presets (instance, not static) but NOT SetViewAngles/RequestClose (drop explicit camera → auto-frame; bEndAppWithTask exits). ColorFloat alpha does NOT show a rod through a dense gyroid → use section (cutaway) + a gold material.
11. 🔴 The drawing tract SILENTLY invents and silently drops — and the reviewer sees LESS than the factory (deep-dig 2026-07-16; the topology: "sheet" default below is ONE MEMBER of this class, not a one-off).
    • Silent drop: Drawing.NotesLines → void Add(label, v) { if (!string.IsNullOrWhiteSpace(v)) … } — a null CEM field removes the whole line. There is no empty Post-process: ___ for the shop to query; the drawing looks COMPLETE. Same for ToleranceLines: Feature renders only if Plus or Minus exists (cathode_flange.json's shank_dia is absent from BOTH svg and dxf today), and InterferenceMin/Max need BOTH or both vanish.
    • Silent invention (worse): the fallbacks are unmarked defaults, so a missing field becomes a FABRICATED factory instruction — ?? "Ti-6Al-4V" / ?? "SLM/DMLS" (title-block) stamp the 4V baseline onto a Ta/Au/7Nb/CP-Ti coupon; the Notes == null branch prints "Ti-6Al-4V (Grade 5)" outright; N(t.PlusMm ?? 0) renders "bore: 0.1/0 mm" — a zero minus-tolerance the CEM never stated, in PMI, in the DXF. CAD_REV ?? "local" stamps rev local → zero git traceability.
    • Reviewer < factory (inverted risk): the SVG title-block truncates (CathodeFlange, PROCESS → 22 chars) and overflows its canvas (TiCoin title-block writes past x=820 → PROCESS clipped mid-word; the SSOT line sits at y=566 on a 560-tall frame = invisible); the DXF truncates nothing. So a bad note rides to the shop precisely because self-review can't see it.
    • Tests protect the bug: every xUnit CEM is an INLINE literal — not one test reads a shipped cem/*.json — and DrawingTests asserts Contains("Ti-6Al-4V", svg) on a Notes == null coin, i.e. it green-lights the fallback that IS the defect. CI never runs draw on a real CEM (cad_smoke.yml runs verify only). The one test worth writing: read the real cem/ti_coin.*.json and assert every non-empty field appears VERBATIM in the DXF. Reflex when touching Drawing.cs/CEM: ask "what does a NULL here print on a factory drawing?" — and prefer a loud NOT SPECIFIED IN CEM over a plausible default.

Common Tasks

• Add a part / per-species SKU: write cem/<name>.json (with a kind) + a generator in src/SilkenCad/ + wire the build/verify switch in Program.cs; dotnet run -- verify.
• Change anchor geometry: edit cem/anchor_zone1.*.json (Ø, bore, period, wallParam). Geometry numbers are owned in 01_01 §5 + founder decisions in 00_07 HW.33; MEASURE porosity after (gotcha #4). Render via Zone1Anode.Anode (the ctor route, gotcha #1). 🔴 topology defaults to "sheet" SILENTLY (Cem.cs; a member of gotcha #11's class) and 6/7 anchor_zone1.* omit the key — so every SKU renders sheet, while canon 01_01 §5.5 says "дані схиляють до network". That choice is an un-made founder verdict (00_07 HW.33 ⚖️), not a default to inherit: a factory STL cut today would ship the disfavored branch. Do NOT quietly pick a side when touching these manifests. (stepped is a separate, already-decided THIRD branch — implemented ZonedGyroid, orthogonal to the open sheet-vs-network verdict; Cem.cs now documents all three.)
• Monolithic bus rod (01_01 §1.4, HW.34, SHIPPED): bus_rod_diameter_mm > 0 ⇒ a SOLID central rod core. Zone1Anode.BuildMonolithic = Anode (gyroid, ctor) + BoolAdd(BusRod.voxConstruct()) (solid via voxConstruct, gotcha #9 — NOT the SDF ctor). 🔑 Porosity stays a property of the gyroid (Anode + the annulus envelope, InnerRadiusMm = rod surface) — the rod is SDF-invisible, so connectivity /porosity gates are untouched; verify separately MEASURES the fused rod volume (ReportAnchor, ≳π(r)²·L). Cathode keeps its channel (mechanical_lock.zone3/flange bore); anode shank bore→0 (solid). F3 audit = AxialStack.BusRodClears (rod + 2·liner ≤ channel).
• Graded gyroid (v2, SHIPPED): own SDF, NOT LEAP ImplicitModular (FunctionalScaleTrafo is a hard-coded Z-demo, not radial; LatticeLibrary submodule ~1 yr stale). Three CEM-driven axes in Zone1Anode: GradedCartesianGyroid (continuous period+wall taper) + ZonedGyroid (stepped zones). Pick by goal — continuous-gentle (smooth, ≤~0.8× period ratio), GyroidWallParamRim porosity gradient, or topology: stepped for a STRONG ~2× pore contrast. MEASURE porosity.
• Mechanical-lock barbs (MechanicalLock.cs, SHIPPED): asymmetric ratchet R(z) on the Ti shank — solid BaseCylinder().voxConstruct() + thin barb-ridge BoolAdd + groove-ring BoolSubtract + a central bore (0 ⇒ SOLID monolithic anode shank, 01_01 §1.4; Ø1.3 ⇒ the cathode channel the bus rod threads) (gotcha #9 — never the SDF ctor for the solid). Tooth is over-specified in 01_01 §4.3 → keep α/β + h, DERIVE base = h·(cotα+cotβ), MEASURE in verify. Self-support is orientation-conditional (print leading-ramp-down, 01_02 §1.6); DIN-471 groove = real shaft dims (not the off-spec canon 0.8×0.6).
• Connectivity / validation (ARCH.25): Connectivity.cs samples the CEM SDF → 3-phase grid → 6-conn flood-fill → open-pore (Archimedes) / percolation (EAAE flow-through) / solid-island (AM + electrical) / closed-pore (trapped-powder) / specific-surface. Pure-managed → fast display-less xUnit. Two-phase resolution split: pore OK at the coarse step, solid needs ~period/16 (gotcha #8). The verify gate adds open≥95% · solid-disc≤2% · percolate axial+radial. Feeds HW.33 sheet-vs-network (topology-agnostic).
• Capsule-end assembly (Assembly.cs, SHIPPED): brings Деталь 3 ↔ Деталь 4 into one frame at the bayonet datum (radome lock-groove ↔ flange lugs) via MeshUtility.voxApplyTransformation (lift; the per-part Builds stay untouched) and MEASURES the residual mismatch (radial / bayonet-Z / RF) + models the skirt/inboard MATE-Ø candidates. An AUDIT table — verify exits on a broken render only; the mismatch numbers are asserted by pure xUnit. Canon 02_02 §4.4; reconcile is 👤 bench (HW.17/HW.8).
• Full axial stack (AxialStack.cs, SHIPPED): the SECOND integration artifact — brings ALL FOUR zones (anode → Zone-2 sleeve → flange → radome) into one axis and MEASURES the press-fit interfaces the capsule-end never touched: Zone1↔2 line-to-line (real +interference = H7/s6 band on bench) · Zone2↔3 = −1.0 mm = the Ø9-in-Ø11 clearance = F1, shank Ø placeholder → HW.8.9 · insertion budget · span. AUDIT table; render uses the Zone-1 envelope (solid Ø11 — a press-fit cares about OD, not porosity; also keeps the 0.2 mm voxel safe). Reuses Assembly.Build + voxApplyTransformation. Canon 02_02 §4.5. 🔑 BasePipe/BaseCylinder Z-origin = [0, L] from the frame (grows along +localZ; verified in LEAP Frames.cs — NOT centred), so stack lifts are absolute; the render overlap sleeve∩capsule is the flange SHOULDER on the sleeve top face, not the shank (the Ø9 floats in the bore).
• Generate an engineering drawing (Drawing.cs / draw, SHIPPED Phase 0+1): draw <cem> emits SVG (human) + DXF via netDxf (factory, opens in AutoCAD/Fusion) — no PDF (scoped in research, never built); ⚠️ read gotcha #11 BEFORE touching notes/tolerances — pure-managed, no Library.Go, consuming the CEM ToleranceSpec/NotesSpec (zero hard-coded eng-text; DrawingStandard ISO/ASME param). Drawing carries fits (Lamé-µm, NOT a blind ISO-286 metal H7/s6 on a PEEK bore), GD&T datums, post-process + coating-restriction notes, lattice-spec. Shipped kinds = ti_coin (Phase 1) + cathode_flange (Phase 2, rode the Ø25 freeze) — flange has NO xUnit yet while Ti-coin carries five. Still deferred: Zone-2 sleeve, radome, draw anchor_zone1 (gyroid inspection-card — does NOT exist, so the Zone-1 coating-restriction map has no carrier yet), assemblies. Home: docs/drawings_program.md.
• Render / section for presentation (render/section, SHIPPED): render <cem> = a PicoGK native-viewer screenshot (gold Ti-metallic material); section <cem> = a −X cutaway (shows the bus rod through a dense gyroid where ColorFloat alpha can't). Display-gated (gotcha #10); output → out/*.png (native TGA → sips). The presentation gallery docs/images/cad/ is NOT SSOT (scripts/render_gallery.sh rebuilds it).
• Local-verify: dotnet build SilkenCad.sln (0W/0E) → dotnet run --project src/SilkenCad -- verify cem/<x>.json (metrics.json + exit 0/1) → dotnet run -- draw cem/<x>.json (SVG+DXF → out/) → dotnet test. CI = enterprise 2-job cad_smoke.yml (logic = Linux pure-xUnit hard [incl. draw/DXF]
  • render = macOS build-hard + verify best-effort, Library.Go 139 headless).
• Update a submodule: git -C tools/cad/extern/<repo> pull + re-pin the gitlink (vendored LEAP source, kept out of owned code).