Commit Graph

  • 6b06fadb47 DECISION: Preserve both hand workspaces in the requested project snapshot main Jerry 2026-11-17 13:41:10 +08:00
  • bc75337a81 DECISION: Decode canonical placement ticks before exporting MJCF frames Jerry 2026-11-17 13:41:02 +08:00
  • d932e26b80 DECISION: Keep both hand menus on the verified photo-refined revision Jerry 2026-11-17 10:38:37 +08:00
  • 8d3415602f DECISION: Refine the complete five-finger exterior from physical reference photos Jerry 2026-11-17 10:38:37 +08:00
  • a86e9b7fc9 DECISION: Replace the user's existing ChatGPT launcher entry with R0.3 Jerry 2026-11-17 10:11:54 +08:00
  • 88d2267e63 DECISION: Make the exterior revision the Codex gallery entry Jerry 2026-11-16 20:40:42 +08:00
  • 371c8c76c4 DECISION: Rebuild LH20 exterior proportions before refitting actuators Jerry 2026-11-16 20:40:42 +08:00
  • ff92b3aad9 DECISION: Keep both hand editors independent behind an author-labelled menu Jerry 2026-11-16 18:46:08 +08:00
  • b99e1250d7 DECISION: Gate LH20 delivery on reproducible geometry and export checks Jerry 2026-11-16 18:34:04 +08:00
  • 9c751b0cb0 DECISION: Bound compliant-control requests by posture torque budgets Jerry 2026-11-16 18:33:08 +08:00
  • 3d07e3c1ae DECISION: Use independent geared joint modules for LH20 Jerry 2026-11-16 18:33:00 +08:00
  • 451d7fd3c1 Merge pull request #85 from PhySpace/dev Ni Jingzhe 2026-11-12 14:49:40 +08:00
  • 72a029976b Merge pull request #93 from PhySpace/release/2.3.0 Ni Jingzhe 2026-11-12 14:48:35 +08:00
  • afb63fc375 release: 2.3.0 (includes the never-published 2.2.9) Ni Jingzhe 2026-11-12 14:47:57 +08:00
  • 8c48236610 Merge pull request #86 from PhySpace/fix/windows-uv-sync Ni Jingzhe 2026-11-12 11:49:01 +08:00
  • c1868f0edd Merge pull request #92 from PhySpace/fix/boolean-clean-and-fillet-check Ni Jingzhe 2026-11-12 11:18:20 +08:00
  • ddafe56379 Merge pull request #91 from PhySpace/fix/readonly-source-dir Ni Jingzhe 2026-11-12 11:17:28 +08:00
  • e86b9a38d3 Merge pull request #90 from PhySpace/fix/fingerprint-quantization-boundary Ni Jingzhe 2026-11-12 11:16:45 +08:00
  • ee5668c31a test(cpp): follow the clean-up in the goldens, guard the diagnosis builders, format (CI of #88) Ni Jingzhe 2026-11-12 04:06:38 +08:00
  • af696badd8 docs: regenerate the export API pages (the exporters now take a build result, ProductInput) Ni Jingzhe 2026-11-12 03:32:23 +08:00
  • 35a4e5efa9 fix(kernel): cut and intersect unify what they leave split, and a blend never returns an invalid solid (#88) Ni Jingzhe 2026-11-12 03:31:21 +08:00
  • 46635ce386 fix(runtime, exporter): a script can model and export STEP in a read-only, unattended run (#89) Ni Jingzhe 2026-11-12 03:00:59 +08:00
  • 342aae581d fix(artifacts): fingerprint quantization no longer splits a value that sits on a boundary (#87) Ni Jingzhe 2026-11-12 02:55:46 +08:00
  • 453bad341f build: a plain uv sync builds the C++ module on Windows (toolchain and OCCT detection, rebuild only when src/cpp changes) Ni Jingzhe 2026-11-10 17:34:22 +08:00
  • 51c60a67f1 Merge pull request #84 from PhySpace/release/2.2.9 Ni Jingzhe 2026-11-10 05:46:27 +08:00
  • d88289e0ef release: 2.2.9 (replaces the withdrawn 2.2.8) Ni Jingzhe 2026-11-10 05:27:04 +08:00
  • e01016c02e Merge pull request #83 from PhySpace/fix/part-holds-one-solid Ni Jingzhe 2026-11-10 05:00:52 +08:00
  • 5efd5bb31a tests: do not depend on the order of the names a single cell defines (Python 3.12 CI failed on it) Ni Jingzhe 2026-11-10 04:37:09 +08:00
  • 695a03a04b cpp: format part.hpp and bind_product_values.cpp with clang-format 18.1.8 (CI's clang-format job failed on them) Ni Jingzhe 2026-11-10 04:37:09 +08:00
  • 07e0ca2bed Merge pull request #82 from PhySpace/fix/editor-pid-probe Ni Jingzhe 2026-11-10 04:20:16 +08:00
  • 0a961d14ec AGENT.md: write down how commits are organised (one decision per commit, with the reasons from the discussion) Ni Jingzhe 2026-11-10 04:19:36 +08:00
  • f60f69846d tests: the editor's API index test expects Part to take body and no bodies Ni Jingzhe 2026-11-10 04:17:28 +08:00
  • b56634ee6e tests: the cell-cache test of restored Pieces makes its Part from one piece Ni Jingzhe 2026-11-10 04:14:05 +08:00
  • a0923a65f4 docstrings: the source of the API pages say a Part holds one solid (and the pages are regenerated from them) Ni Jingzhe 2026-11-10 04:10:51 +08:00
  • d0040a7a7b changelog + SDK_DEFECTS: record that a Part holds one solid again (and what the multi-body notes now mean) Ni Jingzhe 2026-11-10 04:08:51 +08:00
  • 01852ed1b9 docs/skill: scadpkg-format says a part definition holds one solid (no solid_cache.bodies, no profile 3) Ni Jingzhe 2026-11-10 04:07:54 +08:00
  • fb50a97b07 docs/skill: domain, discipline, workflow and guide pages follow the one-solid Part (and the open-solids rule) Ni Jingzhe 2026-11-10 04:07:37 +08:00
  • 6441107ed3 docs/api: the API pages say a Part holds exactly one solid (drop bodies=, multi-body wording and inspect_body_overlaps_rlist) Ni Jingzhe 2026-11-10 04:06:25 +08:00
  • ba19f430ad cleanup: remove the package's per-body blob path check, and comments that still said "multi-body part" Ni Jingzhe 2026-11-10 04:09:08 +08:00
  • 7e2e99682e cpp: comments no longer describe multi-body Parts Ni Jingzhe 2026-11-10 04:08:07 +08:00
  • 0307aaa40f tests: replace the multi-body Part tests by tests of the one-solid rule (and of what the audit found broken) Ni Jingzhe 2026-11-10 04:07:57 +08:00
  • 910df75260 AGENT.md: record how to rebuild the C++ module here, and why the obvious commands fail Ni Jingzhe 2026-11-10 04:07:16 +08:00
  • f016a14984 editor-web: the "model shown" log says what the header says (open solids are not parts) Ni Jingzhe 2026-11-10 04:04:14 +08:00
  • d415647e96 editor: several open solids have no product until the author picks one (use_as_product), and say so Ni Jingzhe 2026-11-10 04:01:35 +08:00
  • 84aab45d96 api: remove the helpers that only existed for a Part of several bodies Ni Jingzhe 2026-11-10 03:55:34 +08:00
  • 2029beedea scene + exporters: a Part is drawn, exported and measured from its one body (no per-body branches) Ni Jingzhe 2026-11-10 03:53:50 +08:00
  • 90fe84ad38 runtime + preview: several open solids are shown side by side, but are no longer a Part Ni Jingzhe 2026-11-10 03:51:52 +08:00
  • 6c49c88ba7 design-docs: record the decision that a Part holds exactly one solid (and mark multi-solid-design.md superseded) Ni Jingzhe 2026-11-10 03:51:18 +08:00
  • 9fb9401fac artifacts: a part definition holds one solid (drop BodyRef, solid_cache.bodies and profile 3) Ni Jingzhe 2026-11-10 03:48:51 +08:00
  • 2c998266b6 python glue + replay: follow the one-solid Part (pickle through Body, refuse a recorded multi-solid make_part_rpart) Ni Jingzhe 2026-11-10 03:48:51 +08:00
  • 654a6d3e33 part: a Part holds exactly one solid (the constructor refuses several; make_part_rpart and Part lose bodies=) Ni Jingzhe 2026-11-10 03:48:34 +08:00
  • 292045d3fa editor: probe the editor pid without signalling it (os.kill(pid, 0) is a POSIX liveness probe; on Windows os.kill maps signal 0 to TerminateProcess, so every sca cell call killed the editor it was talking to and then failed with WinError 87) Ni Jingzhe 2026-11-10 00:52:40 +08:00
  • d70357f5e2 Merge pull request #81 from PhySpace/dev Ni Jingzhe 2026-11-10 00:18:46 +08:00
  • 87bb8894f0 tools: build_occt.py picks Ninja on Windows when a MinGW g++ is the compiler (with no cl on the path the default generator fell through to NMake, which a source build without MSVC does not have) Ni Jingzhe 2026-11-10 00:16:23 +08:00
  • 218fc23805 cpp: define _USE_MATH_DEFINES for MinGW (OCCT headers gp_Dir.hxx & co. use M_PI; MinGW's math.h hides it under -std=c++20 / strict ANSI, and until now only MSVC got the define in ScadCompile) Ni Jingzhe 2026-11-10 00:16:23 +08:00
  • d4cb97fadb Merge pull request #79 from PhySpace/dev Ni Jingzhe 2026-11-09 22:01:59 +08:00
  • 9d915c4d7c Merge pull request #78 from PhySpace/fix/macos-runner Ni Jingzhe 2026-11-09 22:01:18 +08:00
  • fc8b50178a release: build the macOS wheels on macos-15 (GitHub no longer serves macos-14: 'The configuration macos-14-arm64-us-default is not supported' cancelled the macOS job of both v2.2.8 release runs); the wheels still target macOS 14.0 (MACOSX_DEPLOYMENT_TARGET in pyproject.toml) NiJingzhe 2026-11-09 22:00:41 +08:00
  • 9c9e71eb30 Merge pull request #77 from PhySpace/dev Ni Jingzhe 2026-11-09 15:48:30 +08:00
  • 920822527b Merge pull request #76 from PhySpace/fix/gcc14-algorithm Ni Jingzhe 2026-11-09 15:47:38 +08:00
  • dba952ced8 cpp: operators_promote.cpp includes <algorithm> (GCC 14, the manylinux release toolchain, no longer gets it through another header; the 2.2.8 release build failed on std::find) NiJingzhe 2026-11-09 15:47:08 +08:00
  • bd16ada188 Merge pull request #75 from PhySpace/dev Ni Jingzhe 2026-11-09 14:07:21 +08:00
  • 64e3ef1600 Merge pull request #74 from PhySpace/release/2.2.8 Ni Jingzhe 2026-11-09 14:04:01 +08:00
  • 19962588da release: 2.2.8 NiJingzhe 2026-11-09 14:02:31 +08:00
  • b8aca99e14 Merge pull request #73 from PhySpace/release/editor-viewer Ni Jingzhe 2026-11-09 04:04:07 +08:00
  • 3ad47b60fc README: the idea, the evidence, the skill, the plugins NiJingzhe 2026-11-09 03:02:05 +08:00
  • 8d8532c273 tests: goldens follow the edge-as-wire change (f8a13de): an edge given where a wire was refused is taken as its wire and fails on being open or disconnected (face_from_wire, face_from_wires, wire_from_edges: only the error fields that differ); the type stubs regenerated (import order) NiJingzhe 2026-11-08 21:51:46 +08:00
  • 41ae6d5a0e test: untrack the seven pre-port test files again (swept back in by 60fa2a8 and f4cb667; dev removed them with the C++ port, they import names it does not have) NiJingzhe 2026-11-08 21:27:16 +08:00
  • 4eedc7e89c cpp: sketch/edit.cpp returns an explicit CurveHandle (GCC does not convert handle<Geom2d_Line> / handle<Geom2d_Circle> to handle<Geom2d_Curve> implicitly: the Linux builds failed); clang-format 18 on the five files it listed NiJingzhe 2026-11-08 21:04:54 +08:00
  • 2ee7114adf editor: code mode only for the release: manual modelling is switched off (MANUAL_MODE in lib/manual-mode.ts: no mode switch, command bar, manual keys or sketch/curve pick kind), its code stays; the viewport's renderer, level of detail, instancing and BVH picking stay NiJingzhe 2026-11-08 20:43:47 +08:00
  • f3c4d1640a editor: the 3D curve drawer starts with a free name (path, path_2 ...) instead of path every time NiJingzhe 2026-11-07 02:21:45 +08:00
  • 80fb891a1f editor i18n: the assembly panel (connectors, components, constraints, exploded view, its checks and kind names), the selection hint under the command bar, the part count, and the log's own lines (added, created, exported, a failed or slow cell, re-evaluating); the coverage test also checks the context actions' words, the constraint kinds and the radius fields NiJingzhe 2026-11-07 02:19:36 +08:00
  • 08050b1fd8 editor: surfaces from curves and a solid from surfaces NiJingzhe 2026-11-06 19:04:16 +08:00
  • 6813528ab2 sdk: sew_faces_rshell takes faces and shells (a shell is sewn as its faces; recorded with its pieces, replayed the same), a TypeError for anything else NiJingzhe 2026-11-06 19:04:16 +08:00
  • 6f1c2485b7 editor: several curves picked join into a wire (the wire feature: make_wire_from_edges_rwire of the picked curves, in any order), and every form slot that asks for a wire takes a curve (an edge) too NiJingzhe 2026-11-06 06:12:46 +08:00
  • f8a13de6e4 sdk: an edge is taken wherever a wire is (a sweep path, a loft section, a face boundary, a binormal guide): the wire of that one edge, made by the recorded make_wire_from_edges_rwire so a graph sees where it came from (as_wire, edges_as_wires; Wire(edge) and an implicit Edge -> Wire conversion in Python). make_wire_from_edges_rwire takes its edges in any order (they are chained by their ends first, then added in turn so each keeps its identity) and an edge that touches none of the others is an error instead of being left out. loft_rshell of open sections (the surface between two curves) no longer fails looking for end wires: only a closed end section is a boundary loop of its own NiJingzhe 2026-11-06 06:12:46 +08:00
  • 53de920e81 editor: a 3D curve is one edge: a path drawn in one go becomes one curve feature per line, arc or spline (path, path_2 ...), each starting at the end of the one before (a point end_of another curve: path.get_end_vertex().get_coordinates(), so it follows that curve), a path closed on its start ends at the first curve's start, and a point snapped to an end of another curve is kept at it; all of them are added or none. An older curve cell of several segments is still one wire. The feature list takes each value's kind from the kernel for every feature (a curve is an edge, a sketch a face or a wire) NiJingzhe 2026-11-06 05:45:24 +08:00
  • 3e6ea0b661 editor: 3D curves are picked with the sketches (the "sketch / curve" selection mode): hovered and picked on screen (within a few pixels of the drawn line), drawn wider and in the selection colour, shown in the inspector (kind, edges, length); a picked curve offers Sweep (op.sweep_rsolid with the curve as its path; a sketch picked with it is the profile, and with both given it starts at once), the selection bar hangs at the curve's middle NiJingzhe 2026-11-06 05:17:23 +08:00
  • 1505ff1f42 editor: shell as an operation in progress: picked faces, Shell, and the cup is there at once with a 1 mm wall; an arrow on the opened face drags the thickness into the body (at least 0.1 mm), the panel lists the openings and takes one out, a click on a face of the model (the shelled body itself) opens it too; Esc undoes the feature with every opening added or taken out meanwhile. Fillet and chamfer take more edges the same way (any operation whose one selection takes several picks) NiJingzhe 2026-11-06 04:47:31 +08:00
  • 5c2acac1d6 runtime: a notebook with no Part, Assembly or solid yet is no fault: find_product raises NoProductYetError (a ProductNotFoundError, still raised where a product is needed: export, a headless run with a product) and the editor previews it as an empty model ("no solid yet" in the log) instead of reporting ProductNotFoundError; the editor server's own empty-notebook case is gone (the kernel's empty preview covers it), and a notebook's first model is framed when it has one NiJingzhe 2026-11-06 04:29:36 +08:00
  • d3eb5a6794 editor i18n: sketch mode in Chinese: the toolbar (tools, toggles, constraints), the drawer (status, constraint list, diagnostics, profile), the plane chooser, the bar (name, finish, face/wire), typed entry, value and corner boxes, project, the hints and edit notes of the sketch editor, the sketch rows of the inspector and the name check; a test checks that the keys held in the sketch tables (tool names, constraint labels and shorts, typed fields, solver states) have their Chinese entries NiJingzhe 2026-11-06 04:23:12 +08:00
  • 5aca0dc782 editor: the viewport's ground and view controls, smooth folds and the operations' icons in the feature tree NiJingzhe 2026-11-06 04:10:22 +08:00
  • 0e90fc152b editor: the left column slides like the mode switch (both modes' panels side by side on one strip, no gap; to code it moves left, to manual right); the command bar sits at the bottom only, sliding up into manual mode and down out of it (docking to the top removed) NiJingzhe 2026-11-06 03:51:32 +08:00
  • a68f5d5558 editor: one layout for both modes: the left column holds the panels (the mode switch on top of it, run all there in code mode; only its content slides on a mode change), the inspector and the selection kinds float over the viewport in both modes, the header keeps the notebook, status, theme and language (undo and redo are the feature tree's) NiJingzhe 2026-11-06 03:33:29 +08:00
  • 69ac64025a press_pull_rsolid: one signed distance pulls a planar face out (its prism joined) or presses it in (cut), recorded as its extrude and boolean steps; the editor's Press / pull on a picked flat face replaces Pull face / Push in on the selection bar: an arrow on the face along its normal drags the distance (0.5 mm snap, Shift free) while the panel types it, across zero; the face's frame comes from its triangles on screen, at once NiJingzhe 2026-11-06 03:23:58 +08:00
  • 8e01e1908c inset_face_rsolid: a planar face's loops moved inside it are imprinted (the face splits into its inner face and ring, roles inset.inner / inset.ring, tracked as fragments, replayable); the editor's Inset on a picked flat face, tuned in place like every operation, so inset then pull builds steps and bosses without a sketch; the operation panel's Enter handler no longer assumes its target is an element NiJingzhe 2026-11-05 23:13:51 +08:00
  • c7c3c3ceeb editor: an operation in progress (lib/operation-session.ts) instead of a dialog when the selection is all it needs: it is added at once with small defaults (fillet and chamfer 0.5 mm), its numbers float in a panel by the selection and every change rebuilds the model (a failure shows in the panel, the log stays shut), Enter keeps it (the numbers are written into the notebook), Esc takes it back (undone; an edited feature gets its numbers back), the dialog has every option; a feature of the tree is tuned in place the same way. The catalog and SDK index are revalidated (no-cache with an ETag) so an updated editor never serves an old catalog NiJingzhe 2026-11-05 22:43:40 +08:00
  • 8e48abd140 editor i18n: English and Chinese (i18next; the English text is the key, zh-CN.json maps it; the browser's language unless chosen in the header, remembered); the chrome, notices, feature dialog, command bar and the operation catalog (names, groups, fields, options, descriptions) translated, a Chinese query finds operations by their Chinese names; a test fails when a literal key has no Chinese entry. Not yet: the assembly panel, sketch mode and the selection hints NiJingzhe 2026-11-05 22:17:51 +08:00
  • 37a3fe97b4 editor manual mode: the viewport takes the whole window, the panels float over it (browser top left and collapsible, stepping aside for a task's own panel; inspector under the axis gizmo, collapsible; selection kinds top centre; the command bar stays the floating search bar and docks under them); code mode keeps its columns and the viewport is never remounted; rows of buttons and tabs scroll sideways (scroll-row, wheel too) instead of wrapping NiJingzhe 2026-11-05 21:57:35 +08:00
  • d35c92e951 a new part notebook is empty (sca new and the editor's new; the parametric cylinder is --kind sample); an empty notebook is where it starts, told as info, not a ProductNotFoundError; an operation or drawing started with no notebook open asks for one first and creates it in place NiJingzhe 2026-11-05 21:42:09 +08:00
  • 3bacb931c6 editor viewport: level of detail (a part seen up close is tessellated again by the kernel, POST /api/preview/refine, until its chords are under half a pixel; same keys and entities, highlights rebuilt), parts several nodes draw are instanced (faces per colour, edges batched; the nodes stay for picks and visibility), picks go through BVHs (three-mesh-bvh, indirect: the face and edge ranges keep their order), the axis gizmo composited by the output node from its own transparent target NiJingzhe 2026-11-05 16:08:28 +08:00
  • 6e745f3407 editor: opening another notebook drops what the last one showed (workplanes, curves, tag splits, a partial preview, a placer or drawing in progress, the active workplane, sketch picks) NiJingzhe 2026-11-05 15:52:23 +08:00
  • 84974c7b2f editor: the viewport renders with three's WebGPURenderer (WebGL 2 where WebGPU is missing, or with ?webgl): one multisampled scene pass, GTAO on a single-sampled pass of the opaque faces, edges and overlays sorted after the faces; edge depth bias moved along the view ray in the vertex stage; sized points are instanced sprites with their own bounds; the orthographic sketch projection in the renderer's clip space; the axis gizmo cleared each frame; lib/sketch-face.ts (face-boundary snapping), which a69d1ef imported but did not add NiJingzhe 2026-11-05 15:46:14 +08:00
  • 3701e43fcc editor: connectors placed on a face with move/turn handles (the same placer as workplanes), the workspace tab in manual mode too, notebooks deleted from the workspace tab (moved to .sca-editor/trash) NiJingzhe 2026-11-05 04:41:21 +08:00
  • a69d1ef431 3D curves drawn in the viewport (W: lines, arcs, splines on workplanes, snapped to vertices, edge midpoints, workplane origins; a point on a workplane follows it, a snapped vertex is a scad.ref), wires and edges drawn from kernel samples; a failing step previews the model before it; independent spline handles, anchors added while drawing, Shift+A anchor mode; sketch regions (make_faces_from_sketch_rfacelist: outer loops with their holes); a tag on several entities names a set and keeps its pieces; exact surface normals in preview meshes (smooth shading, shared vertices); SimpleWorkplane.point; scad.ref.of a vertex NiJingzhe 2026-11-05 04:30:26 +08:00
  • fabbbe251b editor interaction: workplanes placed with move/turn handles (make_workplane_offset_rworkplane) and chained on the active one; a hole placed with the pointer; pick-intent keys 1-6; sketches drawn only when they are what a click picks; no shape-specific shortcut features, same-name ops labelled with what they make; sketch splines as anchors with two handles (insert/remove/convert anchor SDK ops, pen tool); a sketch can be made into a wire NiJingzhe 2026-11-05 02:26:34 +08:00
  • 7d37131e71 make_workplane_offset_rworkplane: a workplane moved and turned relative to a workplane or a named plane (what the editor's handles write) NiJingzhe 2026-11-05 01:29:48 +08:00
  • 98e24b6aad tags follow their entity as references do: a split names the pieces <tag>.split<i> (numbered when recorded), reported by TagSplitWarning, sca run and the editor's highlight; solids have history (boolean solid witnesses, union merge, one-solid continuation) and scad.ref.solid; references resolve over the pieces of a split body; an ambiguous anchor is told apart by its history NiJingzhe 2026-11-04 20:53:29 +08:00
  • cc2ee95de5 the pieces of a split body flow on: one-shape operations apply to every piece (scad.Pieces), selectors checked over all pieces; the cell cache keeps them as pieces NiJingzhe 2026-11-04 15:10:24 +08:00
  • c3368e6ace fillet and chamfer take the pieces of a split body; bigger snap circle and pick tolerance in the sketch; skill page split NiJingzhe 2026-11-04 14:38:32 +08:00