* Delete: the check in a rolled-back history builds the history once when nothing breaks (#9917)
With the history rolled back, the check before a delete (which features
the delete would break) built the whole history twice, before and after
the change, with the marker removed. On a large design each build ran
every boolean after the marker: deleting twelve splines took half a
minute on the window thread. Only a feature that fails after the change
can be newly broken, so the history after it is built first, and when
nothing fails there the history before it isn't built at all.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Delete: test that the check in a rolled-back history still reports what breaks (#9917)
Deleting a hole in the plate changes the plate and every cut after the
marker and breaks none of them: the check reports nothing. Deleting a
side of the outline breaks the plate, and the check still says so, with
a cache too small to keep a state.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #8341
Co-authored-by: Cursor <cursoragent@cursor.com>
* Graphics: keep a preferred card when surface marks it incompatible (#8341)
Prefer the saved adapter by exact then case-insensitive name, including when only listed as surface-incompatible, and explain in Preferences when another card is still in use.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Format and clippy
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* feat(engine,ui): allow restricting point-to-point align to chosen axes (#9946)
* style: apply cargo fmt across crates
* fix(clippy): use filter then map instead of bool::then in filter_map
* claim: History: build the row plan once per revision and draw only the visible rows
* History: build the row plan once per change and draw only the rows in view
The vertical History worked out its rows from the design on every frame: for each row it
searched every folder's last feature (cubic in features and folders), cloned every name, and
looked up each row's result, sketch and selection by a linear search. It also drew and
hit-tested every row, including the ones scrolled out of view.
- The row plan (features, folders' depth, extent, header places, contents and joints, the rows
with listed sketches, which sketches have patterns) is built once per design change, keyed
by session, revision, marker, counts and the "Sketches: under their feature" choice, and kept
in the egui context; the visible order is kept beside it until a folder or fold opens.
- A feature row outside the scroll area's clip takes its place and publishes its handles, but
is not drawn or made interactive; one being found, dragged or focused, or under an open
menu, is drawn as before.
- Selection and results are collected once per frame; folder rows read their contents from
the plan.
- The horizontal strip collects the picked features (selection and dialog items) once per
frame and writes a chip's tooltip only while it shows.
Tests compare the plan with the per-frame searches it replaced on tangled folders in every
fold state, the kept plan with a fresh one after each kind of change, and the History drawn
with every row against the culled one (same places, same text in view, fewer rows drawn).
An ignored test times the panel on a large design.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Drop a needless borrow
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
* claim: #8749
Co-authored-by: Cursor <cursoragent@cursor.com>
* Recover egui chrome after a long pause between frames.
Sleep/wake can leave the window black while the GPU device stays up: re-upload
the font atlas, clear cached icon textures, and repaint when the frame clock
jumps so the ribbon and panels are not stuck blank (#8749).
Co-authored-by: Cursor <cursoragent@cursor.com>
* Format
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Sketch: a driven dimension nothing fixes any more takes a value again (#10018)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Sketch: a driven dimension nothing fixes any more takes a value again (#10018)
A dimension placed while a constraint fixed what it measures goes in
driven, and stayed driven once that constraint was deleted: a
double-click only said it has no value to edit. The double-click now
tries making it driving on a scratch copy; when nothing else fixes it,
the value box opens at the measured value, and Enter makes it driving
(sketch.toggle_driven) at the typed value (parameters.change) in one
undo step. A refused value goes back to a checkpoint, so it stays
driven. While something still fixes it, it stays read-only as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
* claim: #9945
* Sketch: Project gives its new curves the selected line type (#9945)
The sketch palette gives the drawing line type (construction, centerline, infinite) to the
curves a SKETCH › CREATE command reports under `curves`. Project reports what it made per link,
`links[].curves`, so its curves were never found and stayed normal lines whatever the palette
said. `apply_mode` now also reads the links' curves, so a projected line becomes construction,
centerline or infinite and a projected circle construction, in the projection's own undo step.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EMbqRqYimvY26QzcvrojSC
---------
Co-authored-by: sascha2101 <221609842+sascha2101@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #9489
* Sketch: the Line tool can switch to a tangent arc and back without changing tool (#9489)
With Preferences > Sketch > "Rest on a line's last point to switch to a
tangent arc" on (off by default), leaving the chain's last point and resting
back on it for 0.35 s makes the next click a tangent arc, by the existing
sketch.arc.tangent command, and the same gesture goes back to lines. A label
at the cursor says which. sketch.arc.tangent's end can now name an existing
point, so an outline drawn this way closes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Sketch: the cursor label names the command in words (#9489)
The label asked the catalog for a message id that doesn't exist, so it showed
the id. It now takes the command's translated label, and the test checks the
words it paints. Moved clear of the pointer's own badge.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Show/Hide put every sketch it showed in `shown_sketches`, the list of
finished sketches kept shown although a feature uses them. Hiding and
showing a sketch nothing used yet put it on that list, so once an
Extrude used it, it stayed shown instead of hiding itself as a sketch
never hidden does. Show now marks only a sketch a feature already uses;
one nothing uses is shown anyway and leaves the list alone.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: view corner Cancel cancels an unfinished tool first (#6709)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* View corner Cancel: cancel an unfinished tool before the sketch (#6709)
With OK and Cancel in the view's corner on, the cross took the whole sketch
even with a line half drawn. It now takes the steps one Esc takes first (a
drag, Select Other, the tool's points, then the tool) and cancels the sketch
only when nothing is unfinished. Esc and the cross share cancel_unfinished.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #9333
* Equation curves: the variable can carry its range's unit (#9333)
A curve can set variable_unit "range" (or tick the new box in its window): the
variable is then a length, so a ^ 2 / (4 * t) is a length when a is one. Start
and end must both be lengths, or one a length and the other a plain number in
the design's unit; anything else is refused. Saved only when set, as one pending
format change; without it curves behave as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: Engine: lower the per-command cost of evaluating large designs
* Engine: lower the per-command cost of evaluating large designs
- The fingerprint texts of features are kept in one store shared by every
worker, not per worker: a command evaluated ahead runs on a new worker each
time, where the per-worker store was always empty and every feature was
written out again. Same texts, same hashes.
- Pinning body names after a command keeps, per feature, the names of the
bodies it made with the state they were read from, so it no longer searches
every result's bodies (features x bodies per command), and finds results by
their index before searching.
- Pinning fingerprints the timeline again only from the first renamed feature
(or a pattern or mirror of features replaying one) when the model proves its
fingerprints are the design's; the fingerprints are those of the full pass.
- The evaluation cache's byte count looks bodies and sketches up instead of
searching them per body (bodies squared per feature). Same counts.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #9176
* Preferences: turn the view to isometric when Extrude opens, and back afterwards (#9176)
Two new preferences on the Design page, next to "Extrude to a picked target", both off by
default:
- `extrude_view_isometric`: each time Extrude opens, the view turns to the isometric view, so
the depth can be seen right after a sketch seen face-on.
- `extrude_view_return` (only with the first): when Extrude closes, by OK, Cancel or Esc, the
view goes back to where it was before it turned, keeping the projection as the end of a sketch
does (#7038). The check is per frame, so every way the dialog closes is covered.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EMbqRqYimvY26QzcvrojSC
---------
Co-authored-by: sascha2101 <221609842+sascha2101@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* MCP: batch and execute take expect {volume, bodies, bbox} and flag a mismatch (refs #3380)
An optional `expect` argument on the `batch` and `execute` tools: the volume
(with a relative `tol`, default 0.002), body count and bounding box (0.05 mm)
the agent computed before building. Once the call has gone through the server
measures the design as the `measure` tool does and the reply gains
`expectation: {ok, measured, expected, mismatch}`. A mismatch flags the tool
result `isError` with the mismatch text first; the design is kept as built.
A malformed `expect` is refused before anything runs. Without `expect`
nothing changes, and a batch step carrying one is a step with a stray key.
In 29 logged agent sessions every agent computed these numbers beforehand
and then spent a `measure` call plus a comparison by hand after every stage;
one missed a 0.25 mm skin until it compared. Volume and body count are
kernel-independent checks.
Refs #3380
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* MCP expect: filter then map for the bbox coordinates that are off (clippy)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* MCP: expect's coordinate text helper no longer shares a name with the JSON reader
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Co-authored-by: bherbruck <tenbee91@gmail.com>
* claim: Save: hard-link the backup copy and check for unsaved changes once per revision
* Save: keep the backup as a hard link and check for unsaved changes once per revision
The backup of a save is the old file under a second name (a hard link renamed over the
old .bak) instead of a copy written out again; where links are refused (FAT, shares,
cloud folders) or the old file cannot be flushed, it is copied as before. Every point a
save can stop at leaves the design file whole and the backup the version before it.
Session::is_dirty compared the whole document with the saved one each time it was asked,
which the window does every frame for every tab. The answer is kept per revision and pair
of documents; the cache holds both, so neither can change in place while it stands, and
doc_mut lets it go first so edits stay in place.
On the rover in two unsaved tabs an idle frame takes 15.0 ms instead of 19.3 ms; a save
with its backup of the 11.8 MB file writes in 9.6 ms instead of 18.9 ms.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #9811
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Agent chat: the system note reaches agents beyond the Claude adapter
The note only went out as `_meta.systemPrompt.append`, an extension of the
Claude adapter, so Codex and Antigravity never saw it. It now also goes in
the standard ACP `systemPrompt` field for agents that offer it, and in
`_meta.antigravity.systemPrompt.append` for Antigravity.
Fixes#9811
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
- 247 help texts (help.*) and 95 UI messages: theme options, viewport status, quick measure, personal sheet-metal rules, analysis details, sketch report
- Terms aligned with fr.json; no-break space before : ; ? ! as in the catalog
- Pending file, folded on main after merge
Co-authored-by: P'tite Tête <asso@ptitetete.com>
* claim: #9919
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Extrude: with every body hidden, a new extrude stays New Body instead of turning to Cut (#9919)
Smart Extrude passed the shown bodies as `participants` to keep hidden
ones out (#6901), but with none shown that list was empty, and the
engine reads an empty list as the whole scope: a profile on a hidden
body chose Cut, which has no shown body to change, so nothing was
previewed and OK did nothing. With no body shown it now chooses New
Body.
The operation it latches for a side is also decided again when a body
is shown or hidden while the window is open, as a new start already
does (#467): showing the body an extrude goes into turns it to Cut,
hiding it again back to New Body.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Claim #9916
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Report a bug: the window can be dragged by its title (#9916)
The window was pinned at the top middle of the screen (ui_kit::Window
anchor), over the model, in a sketch or not. It now starts there and can
be moved (`placed`), as the file-load card can.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
The `initialize` instructions were one paragraph of command shapes. Replace
them with version-neutral guidance: read `skill_reference solidcraft-modeling`
first, plan the numbers, build in a few `batch` calls, the primitive routes
and the fillet-shell-inside-holes order, what follows a parameter, `measure`
after each stage with a coalesced parameter check, and `timeline.edit` /
`parameters.change` / `edit.checkpoint` to fix instead of rebuilding. The
Assistant's system note gets the same pointer; docs/mcp.md says so.
`instructions_name_only_real_tools` now checks the command ids as well as the
tools, and `instructions_recipe_builds_a_parametric_shelled_box` builds a part
the way the instructions describe and checks its volume, axes and parameters.
Refs #3380
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
* MCP: measure summary and body_topology faces: cylinders for small verification replies (refs #3380)
`measure {"summary": true}` replies one short object per body (name,
volume_mm3, bbox, faces, closed) with body_count and total.volume_mm3,
as a tool and as a batch step. `body_topology {"faces": "cylinders"}`
replies only the cylindrical faces, the kernel's split halves merged
by radius and axis line into {radius, origin, axis, extent, names},
and no edges. Both defaults are unchanged. The merge uses only what
model.faces and model.edges already reply.
Measured on a 13-feature enclosure, body_topology is 74 KB and measure
1.1 KB per body; agents filter them with scripts, and seven of 29
benchmark runs asked for exactly a cylinders-only listing.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* MCP: keep body_topology's cylinder merge next to check_parameters' without a name clash
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Co-authored-by: bherbruck <tenbee91@gmail.com>