- simplecad-cache/simplecad-export entry points are gone (hard cut):
the same commands live under the unified dispatcher
(simplecadapi.cli) as 'sca cache ...' and 'sca export ...', each
group registered at parser level so per-group --help works
- sca init now wires the shell by default (--no-shell skips): a
~/.sca/bin/sca shim (Windows sca.cmd) execs the installing venv's
console script — the venv bin itself never goes on PATH — and the
shim dir lands on PATH via a marked, removable block in ~/.zshenv
(read by non-interactive shells, so agents see the command),
~/.bashrc, fish conf.d, or the Windows user PATH registry value
with a change broadcast; rc files are created only for the login
shell or when they already exist; re-running init refreshes the
shim (target discovery is symlink-safe, no resolve() out of venv)
- addon/cache/export cli modules expose configure(subparsers) +
_execute(args); module run() keeps the CLI-shaped argv
- docs: sca namespace across evergreen docs, shell-integration
section, 'never writes shell profiles' contract line retired
- tests: +17 (dispatcher groups, shell integration, entry-point
absence); cache/export suites moved to group-token argv
Add the drawing inspection namespace and its parser, calibration, section and
independent acceptance dependencies on the current dev architecture.
Protect input BREP during material fusion and sectioning. Bind dimensions,
annotations and visual reviews to source identities and measured geometry.
Connect the workflow and generated API docs to all seven skill build targets.
Make standalone verifier tests work with the pytest console entry point.
Validation: 186 drawing, documentation, skill-target and public API tests;
16 non-render BREP query tests; compileall and skillbuild --check.
- reverse-engineering-studio.md: free-form composer (entity/op chips,
[label](token) serialization, IME-safe Enter), operation tips as a
markdown registry (viewer/server/operations/, mtime-hot, errors named),
submission schema incl. operations + operation_context, self-contained
artifact rule (transcription = literals, extraction harnesses beside
the case), kernel-ordering facts (bores before blend fillets; fitter
to_dict unique_knots contract), artifact names the UI polls
- step-reconstruction.md: route to the studio variant when a human
operator is available
- SKILL.md: workflow count corrected to six; viewer/README.md updated
for the composer + tips registry
Interactive re-mode (re.html) plus a local case server implementing the
ppt-master handshake: the human selects faces / draws regions / labels intent
on the original STEP, the browser composes a submission with server-side
entity context, and a waiting agent reconstructs per the existing
reconstruction contract and writes artifacts the UI polls back.
- viewer/server (stdlib-only): STEP -> synthesized Scene 2.0 projection
(GLB + entity sidecars, same shape as package scenes), region resolve
(centroid projection + front-facing filter), entity descriptors,
atomic submissions, ndjson side recording, --daemon/--wait/--wait-only/
--shutdown CLI with lockfile and 590s budget
- viewer/src: extract shared cad-three/entity-facts modules (main.ts now
consumes them), new SceneView class (assembly/picking/highlight) reused
by both studio viewports
- re-mode UI: annotation canvas (circle/lasso), intent palette, context
preview, submit flow, rebuilt v3 package loading, comparison image,
feature tree, evaluation tab
- governance: only the browser writes submission.json; UI never edits
geometry — code stays the single source of truth
- tests: 6 server tests (scene contract, region math, submission context
+ named failures, wait handshake exit codes, HTTP endpoints)
The drawing declarations were exploratory scaffolding for the dxf engine
and outlived their purpose once the engine contracts landed; example
keeps only the 3D modeling chain (u_link/shell/assembly/export/verify).
Generated artifacts (dxf/png/diag/action-log) were already removed from
out/ — only 3D exports remain. Workflow doc now points at driving the
loop directly via SheetPlan (solve/evaluate/apply/render) with the
diag_png auto-diagnostic instead of the deleted driver script.
The FTC convention dissolved the sketch-feature-modeling discriminator
(planar profiles are sketch tier inside every part source), and
standard-part usage belongs to the assembly route. Workflow set: 6 -> 4.
- sketch-feature-modeling merged into single-part-modeling: new standing
discipline 'Sketch tier is the default part form' carries the sketch
gates (solve status + remaining DOF before promotion, promotion
preserves measured geometry, stable entity ids, intentional free DOF
recorded); capability detail stays in domains/sketch-and-features.md
- single-part-modeling gains the hard rule: if the requirement can be
satisfied by a standard part or based on one, scad.std.* MUST be the
basis; single std part stays here via domains/standard-parts.md,
mechanisms mating several route to assembly
- standard-part-assembly merged into assembly-product-build: scope now
includes stdlib mechanisms; new step 3 selects std components first
and computes the kinematics explicitly (center distance from tooth
counts and module, backlash, ratios, measured bearing/gear seats);
mesh-consistency gates added; steps renumbered 1-10
- routing table and references/README workflow table reduced to four
goal rows; step-reconstruction's citation of the removed workflow
replaced with the sketch/geometry tier pointer; mirrors rebuilt
Codify the FTC coding standard as references/discipline/
feature-tree-convention.md:
- the block paradigm sketch -> basic body op -> bool -> modifier as
explicit dataflow (body = op(body, tools)), one application call
per block, no accumulator state
- mandatory boundary comments '# ---- feature: <slug> (<role>) ----'
with closed role vocabulary (build|add|subtract|intersect|modify|
pattern|annotate) and tier annotations profile=/path=; slugs are
stable identities, never sequence numbers
- tier rules: closed planar profiles and planar sweep paths are sketch
API only (tangency is a constraint, not hand-computed coordinates;
open-chain promotion is legal, closedness is a consumer contract);
geometry tier is restricted to non-planar 3D curves, transcribed
geometry with unknown constraints, and tool bodies whose shape is
completely contained in a primitive form — primitives are never a
shortcut for a profiled feature
- modify blocks select via QL predicates + exactly(n) cardinality;
parameter guards stay in source, geometric verification stays
external; block ordering defers to feature-ordering.md
Wire it as required reading into every workflow on this branch (all
six): single-part (standing discipline + Role 4 reading), sketch-
feature (required_domains + build step), assembly-product (per-part
authoring step), step-reconstruction (build step, geometry-tier
honesty), standard-part-assembly (custom parts step), export-and-
translation (source edits while making a package current). SKILL.md
global rule 3 points at the convention for part-source form. Mirrors
rebuilt via tools/skillbuild.py.
Mirror the single-part-modeling role architecture: Role 1 Drawing
Requirement Confirmer (blocking ask round -> DRAWING.md), Role 2 Sheet
Planner (view roster + R5 coverage ledger), Role 3 Layout Arbiter, Role 4
Visual Reviewer (delegated PNG verdict, repair: loop), Role 5 Drawing
Exporter — with the Role 3/4 alternation running per sheet.
Role 3 now carries the engine's core layout method as an explicit
validate-and-rollback cycle: viz colored review image (ok/unsolved/warn/
accepted + free-movement corridors) -> plan.evaluate() digested
violations/corridors -> action-log JSON (rad/dia/lin/lead/accept) ->
plan.apply() with reject-and-rollback residuals -> pixel diff of the two
replays -> repeat until warn=0. Record the division of labor (engine owns
deterministic truth, agent owns discrete arbitration) and the
action-log-as-decision-log invariant. Route evaluate/apply in the domain
API groups.
Add references/workflows/engineering-drawing.md (preliminary staged
workflow: preconditions -> declare views -> declare datums/dims -> solve
and audit -> render and visual review) and references/domains/
engineering-drawing.md (capability boundary, R1-R7 solver contract, API
groups, known engine limits, failure modes). Route the new goal in
SKILL.md.
Backfill docstrings for the dxf_engine public surface (declaration
dataclasses, SheetPlan, render_plan) so auto_docs_gen can produce their
API pages under references/docs/api/ with an Engineering Drawings index
section, following the inspect/brep sub-namespace pattern.
Teach the single-part modeling workflow how to design the Role 1 question
payload for opencode's native question tool (header-keyed ledger rows,
recommended-first ordering, free-text passthrough). Include the
u_link_motor_mount worked example: requirements/build plan, part and
assembly scripts, export pipeline, and the staged verify/probe scripts.
The DXF drawing scripts land separately with the dxf_engine branch.
Rewrite SKILL.md in the routed-authority register (mandatory numbered
load order, hard rules, routing-discipline table): deterministic route
selection, forbidden route-choice menus, one pre-scripted discriminator
question for ambiguous requests, authority precedence, and an
execution-discipline section that binds every gate to a named artifact.
Apply S1-S8 from design-docs/skill-solution.md plus five deeper
patterns from session-audit-01a02469 evidence:
- S1 Ask-or-Record gate + mounting/load brief fields; S6 companion line
- S2 selection-evidence gate before fillet/chamfer (card + cardinality
assertion + highlight render), anti-patterns named
- S3 edge-selection defaults: shared_boundary first, index getters only
for intentional picks; wired into the workflow step
- S4 check-definition protocol (posture vs path, verification vs design
geometry, feedback-to-predicate)
- S5 fastener envelopes precede geometry as brief inputs
- S6 visual acceptance minimum replaces 'Visual plausibility'
- S7 reading-an-API-page contract + external-dependency boundary
- S8 a run analysis ends in a verdict
Deeper audit patterns folded in: acceptance set before geometry,
own-the-methods/ask-the-requirements split, assumptions as brief fields
with derivation tracking, enumeration-is-not-repair, and
discriminating-evidence selection.
* 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