* Ribbon: an option to fold it to its tab row (#10112)
Preferences > Layout > "Show only the ribbon's tabs" (off by default), the
Ribbon Tabs Only command (view.ribbon_tabs_only), and its settings-profile key.
Rebuilt on main's redesigned Layout page.
* Ribbon Tabs Only: switching tabs with a panel menu open no longer keeps the overlay open (#10112)
A tab click now forgets the open panel drop-down of the previous tab, so the pointer leaving or Escape still closes the overlay.
* Ribbon Tabs Only: count its settings-profile key in the profile membership test (#10112)
* MCP: compact inspect_design and tool JSON, off by default (--compact / SOLIDCRAFT_MCP_COMPACT)
Split out of #2901 per review: the smaller inspect_design changed the default
for everyone, so it is now an MCP server option that is off unless asked for.
- solidcraft-cli mcp --compact, or SOLIDCRAFT_MCP_COMPACT=1 (true/yes/on;
0/false/no/off/empty leave it off; anything else is an error at start-up).
solidcraft_mcp::Options::resolve is the one place that decides, so an app
Preferences setting can feed it later.
- With it on, inspect_design summarises sketches (id, name, plane, counts of
points, curves, constraints and profiles, dof, status), returns full contents
only for the sketches named in `sketches` (names/ids, true, false or
"summary"), takes a `bodies` filter and measures only those; unknown names
list the existing ones. Tool JSON is compact rather than pretty-printed.
- Off, tools/list, argument checks, inspect_design and tool JSON are exactly
as before.
- document.inspect gains an opt-in `sketch_counts` (plane, points,
constraints); without it its reply is unchanged.
Tests: both modes in the MCP suite, Options parsing, and the CLI flag and
variable end to end.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* MCP compact output: a Preferences setting too (AI › Agents through MCP)
The desktop app's Preferences › AI gets "Compact output" under a new "Agents
through MCP" section, off by default and shown with the chat off too. The app
reports it on the control channel as `mcp.options` ({"compact": bool}); the
agent chat lets its own MCP server read it.
A bridged MCP server (`solidcraft-cli mcp --connect`) asks the app when an
agent first lists or calls a tool and keeps the answer for the session, so the
tools it listed keep their shape. An app not running yet, or busy past 2 s, is
asked again on the next call; an app from before the setting counts as off.
`--compact` and SOLIDCRAFT_MCP_COMPACT=1 still turn it on whatever the app says.
Tests: the server follows the app's answer (waits for it, keeps it, the flag
still wins); the checkbox reaches `mcp.options` and Restore Defaults clears it;
the chat gate allows `mcp.options`. docs/mcp.md and docs/control-protocol.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* MCP compact output: review fixes (mcp.options never blocks, headless test, locales, prefs doc)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* MCP compact output: the setting is a section of the AI page; the bridged test answers mcp.options
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* MCP compact output: the preference test sets it through ui.prefs; rustfmt
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* MCP compact output: look for a newly started app before asking its options; counts from sketch.inspect for an older app
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #4332
* Two-distance and distance-angle chamfers of free-form rims: a test on the SMLib kernel (#4332)
A spline, Bézier or text outline's rim used to be refused by the two-distance and
distance-angle chamfers. The SMLib kernel (#11351) chamfers it along the rim; this test pins
that down on a four-Bézier oval, picked as its tangent chain the way a pick takes it: the
volume taken off is the analytic one for two distances (both ways round) and for a distance
and an angle.
* Sketch: the geometry of a pierce, a point where another sketch's curve crosses this plane (#3599)
* Sketch: the Pierce constraint, a point on a curve of another sketch (#3599)
* Engine: draw a Pierce constraint at the point it holds (#3599)
* Engine: the pierce tests wait for the command that is not written yet
* style: rustfmt, the imports of the pierce geometry in the order rustfmt wants
The rebase onto the current main resolved the rename of Plane, Vec2 and Vec3
from solidcraft_geom to solidcore_record and left the two use lines in the
order the resolution wrote them. rustfmt wants them the other way round, and
the fmt phase of the gate reads the whole tree.
No token of the file changed: the same 2983 tokens before and after, only the
order of the two use lines moved.
---------
Co-authored-by: bherbruck <bherbruck@gmail.com>
Co-authored-by: vdirienzo <vdirienzo@users.noreply.github.com>
* claim: #10745
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Control channel: ui.menu refuses missing targets, reports the History item menu, and tells the joint menu items apart (#10745)
- ui.menu looks a target up and refuses one the design doesn't have ("no body `NoSuchBody`").
- The History bar's feature menu is the shared context menu (Target::Timeline), so ui.menu
reports it, ui.menuPick runs its items and {"type":"timeline","feature":id} opens it.
Its Show All toggle moves to TreeState so the menu can list Delete after the marker.
- Edit Joint Limits is ui.editJointLimits and opens Edit Joint with Limits expanded.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Icon audit: open the section and canvas menus on a real section and canvas; a hidden section has a menu too
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* ui.menu: check a group's or folder's scope and folder name; Delete after the marker gets its own id
A group or folder target is refused unless its scope is a row of its component (`c<id>` or a
placement of it) and the folder is bodies, sketches or construction. The History item menu's
delete of everything after the marker is ui.deleteAfterMarker, so ui.menuPick can tell it from
Delete (review on #10746).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* ui.menu: a component's own c<id> scope is the design root's only; test that Delete after the marker deletes
The browser draws a part's folders and groups under its placements (o:<path>), so a c<id> scope
for any component but the root names a row that isn't there (review on #10746). The History test
now picks ui.deleteAfterMarker and checks the features are gone.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* ui.menu: a placement scope must be spelt as the browser keys it
o:012 parses to placement 12 but isn't the row's key (o:12), so Show/Hide would store a key no
row reads (review on #10746).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* History item menu: no Delete after the marker when nothing shown is after it
A timeline target for a feature the bar's component filter leaves out listed no features, so the
item did nothing (review on #10746).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* ui.menu: a plane target must be construction geometry
The plane menu's Show/Hide keys the feature's name as an origin item, so any other feature got a
menu whose actions don't fit it (review on #10746).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* ui.menu: accept the view navigator's and selection filter's menus, which main added
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* AGENTS.md and the contributing skill: say how to work as an agent
Two rules that the loop's own agents follow but the repository never
wrote down, so every agent that reads it had to rediscover them:
* The code, the comments, the commit messages and the docs are in English,
always. A comment in another language is a defect: the people who
maintain the code cannot review it. A user-visible string is not inline
in the code either; it goes to the message catalogs, and the UI text
skill says how to name and translate it.
* Parallel work runs in parallel, and dependent work stays sequential.
One subagent per independent part, with a self-contained brief, and no
two subagents writing the same file. A fix and the test that proves it,
a refactor and the callers it breaks: those stay in order.
AGENTS.md is the router, so it gets the two rules in one line each, and
skills/solidcraft-contributing/SKILL.md gets the "Working as an agent"
section with the detail.
* AGENTS.md: fit the two agent rules under the router cap
`cargo xtask skills` reads the router as a router: one line per rule,
with the detail in the skill the line points at. The two new lines
broke that contract twice over. They carried their own detail (a
comment in another language is a defect; one subagent per part, no two
writing the same file) instead of pointing at the "Working as an
agent" section that already holds it, and the file grew to 5361
characters, over the 5000 a router may have.
So the two lines shrink to what a router line is, and five clauses
that repeated a rule the pointed-at skill already states in full give
up their place:
* "Act when the code, docs and rules settle a question" is in the
contributing skill, in more detail.
* The unsafe allowlist is the crash-safety skill's own list.
* "No second PR for an issue a lower-numbered one closes" and "never
force-push main, no pushes just to catch up with main" are the
contributing skill's claim and push rules.
* "Do not switch branches while the gate runs" is in the contributing
skill too.
No rule is lost: each one still has exactly one place that states it,
and that place is a skill. The router is 4970 characters.
Open GitHub issue form filled the issue through the link only up to
2,000 characters, with every space and line break encoded as three, so
most reports opened GitHub with the title alone and their text on the
clipboard, and many arrived empty (#10556, #10654).
The link now encodes only what a query value can't carry (a space is
`+`) and is limited by what GitHub keeps: 6,000 characters, and 7,000
once its sign-in page encodes it again. A report too long for that still
goes through the clipboard, whole.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #7641
* Sketch on a thickened surface follows the face (#7641)
Thicken dropped the kernel's face history, so Create Sketch refused
every face of the solid. Carry that history through, the same way
Shell and Trim do.
The regression covers the offset face and an end wall, and checks
that a boss on the offset face follows a thickness change and reopens.
* claim: #9623
Signed-off-by: Maurizio Monticelli <maurizio@whatwedo.ch>
* Sheet metal and plastic rules: Thickness in a rule's expressions is the rule's thickness
A rule's expressions read `Thickness` as the rule's thickness only while the
design had no parameter of that name; with one, the default gap, reliefs and
a `Thickness` bend radius took the parameter instead (a 2 mm rule beside
Thickness = 10 mm closed corners to a 10 mm gap). Now `Thickness` is always
the rule's, as documented; the rule's own thickness still reads the design's
parameters, so `thickness: Thickness` keeps naming the design parameter.
Renaming a parameter called Thickness, its users and what it drives follow
the same reading: only a rule's thickness field refers to it. Renaming a
parameter to Thickness is refused while another rule field uses it, since
that field would then read the rule's thickness instead.
Fixes#9623
Signed-off-by: Maurizio Monticelli <maurizio@whatwedo.ch>
---------
Signed-off-by: Maurizio Monticelli <maurizio@whatwedo.ch>
* Claim #9736
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Assembly: a move not captured yet is an unsaved change (#9736)
A component moved without capturing its position left the design reading
as saved: file.save wrote the file without the move and said nothing, and
closing the tab didn't ask. Now:
- Session::is_dirty is true while a move is pending, so closing a tab or
quitting asks "save changes?".
- file.save and file.save_as refuse while a move is pending, naming the
component and the capture and revert commands.
- In the app, Save As (and Save As JSON, Save a copy, Download design)
asks the Capture Position question first, as Save already did; a tab
closing after "Save" waits for that answer, and Cancel keeps it open.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Close: the save-changes card waits behind the Capture Position question (#9736)
Saving a tab that closes asks the Capture Position question first; the
"save changes?" card stayed drawn over it and took a second answer. It
now hides and ignores answers until the question is answered. The design
container doc says a save stops for a move not captured.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Assembly: an exploded view shown is not a move to save (#9736)
explode.show displays a view through the pending moves, as an uncaptured
move does. Those placements last only while the view is shown, so they
don't make the design unsaved and don't stop a save
(Session::moved_uncaptured). A move on top of a shown view still does.
The relink test saved with a move pending; it captures it first now.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Assembly: the pending-move save test reads the world state as main now gives it
world_state() returns a Result since the kernel merge; the test unwraps it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
* Docs: add BoxInsertDesigner to the showcase
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Docs: attribution rows for the BoxInsertDesigner showcase recording
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The joint origin input only accepted faces, edges and vertices, so the world origin and sketch points could not be picked to place a joint. Extend the origin input's snap mask with ORIGIN and POINTS and resolve a sketch point (and the empty-body world-origin vertex) to a world snap in snap_of, so a joint can be placed on the design origin or a hole centre like Fusion.
* Sketch Project: a face or edge of a part's second placement projects where that placement is (#9957)
Projecting the face of `Block (Part:2)` failed with "face of Block: the
face no longer exists". The pick's body name was turned into the
definition's own name, and the model a sketch reads holds only each
component's first placement, so the face was looked for on the first
placement, somewhere else.
Sketch picks now keep the name the browser shows, and the further
placements a sketch's links name are placed into the model it reads
(`with_placed_copies`), when the sketch is evaluated and when a link is
added or refreshed. Only the named ones are placed. A link to a body
the model lacks also makes the sketch follow every placement, so moving
the copy moves what was projected from it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Sketch Project: a projection from a second placement follows renames of its part and body (#9957)
The link keeps the copy by the name the browser shows (`Block (Part:2)`),
which a rename of the component or of the body changes. component.rename
and body.rename now rename it in the sketches' links as well.
The copies a sketch reads are found through their component's bodies,
not by going through every body for every placement.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The Scale dialog gets a Type row, Uniform or Per axis. Per axis shows
X, Y and Z rows and sends factors, which solid.scale already takes. A
Scale made with factors opens per axis with its three values; before,
they rode along as an extra and the dialog's factor did nothing.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #4644
* Add the Fair Surface and Match Surface dialogs (#4644)
Neither command had a dialog, so the toolbar and command search ran them
with no parameters and they failed with "`body` is required".
Fair Surface picks the surface or some of its faces and sets the
strength, the iterations, whether the edges are kept and an optional
tolerance (blank: 0.001 of the body's size). Match Surface picks the
surface or face to rebuild, the target and an edge chain of the target,
and the continuity (G1, G2 or G3, default G2). Both preview, and Edit
Feature restores the saved picks and settings. New labels go through
the catalogs with English, Japanese, Chinese, German and Korean text.
* Keep surface-dialog strings pending with complete translations (#4644)
* Fair tolerance is a length; Fair and Match picks reopen on their saved bodies
The Fair dialog's tolerance now goes through the same design-unit length
conversion as the other surface lengths, on OK and when an edit is refilled.
Editing a Fair or Match feature looks its face and edge picks up on the body
they were saved on, so a coincident face or edge of another body cannot take
their place.
* Surface dialogs: strings stay in the pending catalog only (#4644)
* Surface dialogs: edge picks use the session world, as main's edge_sel now does (#4644)
* Surface dialogs: refuse to refill when a saved pick no longer resolves, and pass body names by reference (#4644)
* Surface dialogs: Fair and Match now open dialogs, so the refused-command tab test no longer lists them (#4644)
* Stderr: a closed stderr no longer stops the app, the CLI or the MCP server (#10111)
`eprintln!` panics with "failed printing to stderr" when stderr is a
pipe its reader has closed. A process another program starts writes
into such a pipe: `solidcraft mcp` under an agent, and the app that
one starts, which inherits it. #10111 is the app stopping that way
while an agent (Codex) connected. `solidcraft_geom::note!` (re-exported
as `solidcraft_engine::note!`) writes the same line and ignores a
failed write, as the CLI's logger already does; the app, the CLI, the
MCP server, the crash hook and the crates they run use it instead of
`eprintln!` outside tests.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Stderr: test that a closed stderr leaves the CLI's exit code as it is (#10111)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Stderr: the old name's alias and the fuzz command don't stop on a closed stderr either (#10111)
`solvecraft-cli` writes its deprecation notice before it starts
`solidcraft-cli`, so an agent still configured with the old name got the
panic before the MCP server started (found in review). It and the fuzz
command now write through `note!` too, and the test covers the alias.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Browser: the component menu keeps its everyday actions and groups the rest in flyouts
A component's or placement's right-click menu had grown to about thirty entries in seven
blocks. Activate, Edit in Place, Open Linked Design, Move/Copy, Show/Hide, Isolate, the
clipboard, Delete, Rename, Appearance, Physical Material and Properties stay on the menu;
Ground, Capture and Revert Position, Unlock Contents and Turn Up Axis move to a Position
flyout, Relink Source or Extract to File and the exports to File, New Component, New Group
and Paste New to New, and Show All Bodies, Cycle Component Color, Expand All and Collapse
All to Display. context_menu::entries lists a menu with its flyouts, as ui.menuPick reads it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Browser: the component menu nests its flyouts by topic, and flyouts no longer scroll
Position (Ground, Capture and Revert Position, Unlock Contents, Turn Up Axis), Component
(New Component, Rename, Appearance, Physical Material, Cycle Component Color, Relink Source
or Extract to File, Properties; Design on the design's row), Export (Export, Save As Mesh,
Send to Slicer) and Browser (New Group, Expand All, Collapse All) each open a flyout; Paste
New stays with the clipboard and Show All Bodies with Show/Hide. Flyouts open inside flyouts,
so Appearance keeps its quick choices inside Component; the keys walk in and out of them.
A flyout no longer scrolls or sits away from its row: every flyout has its own area, measured
whole, where a short one used to squeeze every later one into its height. A component's only
body is no longer offered as a component of its own on a folded Bodies folder's entries.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Browser: the component menu starts with what a double-click does, and keeps the old order of its sections
A linked component's menu starts with Edit in Place and Open Linked Design when a double-click
edits it in place (double_click_components), and with Activate otherwise. Position, Component
and Export follow Move/Copy, then the clipboard, then Show/Hide and Isolate, with Browser last,
as the sections came before the flyouts.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Browser: Show All Bodies is on a component's menu only while a body is hidden
It is listed while a body is hidden itself, in a hidden placement, or in a hidden Bodies folder or
group of bodies, anywhere in the design, since that is what it shows.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Browser: Relink Source and Extract to File sit right below Rename in the Component flyout
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Browser: Placement and Organize flyouts, Turn Up Axis back as a flyout, and flyout fixes
Placement (Ground, Turn Up Axis, Capture and Revert Position, Unlock Contents) follows
Move/Copy above the first separator, Turn Up Axis opens its two turns in a flyout again, and
the Browser flyout is called Organize. The Component flyout has three blocks: New Component
with Relink Source or Extract to File, then Appearance, Physical Material and Cycle Component
Color, then Rename and Properties.
A flyout of a flyout that opened to the left opens to the left too instead of covering the
menu, Left back to the menu chooses the row that opened the flyout, ui.menuPick refuses a row
that only opens a flyout, and the icon audit checks every level. The guides and the control
protocol describe the new menu paths.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Browser: ui.menu gives a flyout's rows their placeholder flag too
ui.menu set placeholder only on the top-level rows, so a client reading
a flyout row's placeholder got nothing. The icon audit now reads the flag
of the flyout rows instead of resolving their icons itself.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #11505
* Selection filter: started from command search, it offers Faces, Bodies, Components (#11505)
Command search ran select.filter with no parameters, which could only fail
with "missing field `mode`". Started from the UI (command search, a
shortcut, ui.start), it now opens its three modes as a menu with the
current one marked, as View > Visual style and Navigation tool do (#8868).
Programmatic calls keep their contract and open nothing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: thocsmith <193096300+thocsmith@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #8201
* Sketch Mirror: use an origin axis that is not in the sketch plane (#8201)
A sketch on a part face does not hold the world X, Y or Z axis, so Mirror refused the pick. It now mirrors about the axis's projection into the sketch plane and keeps that line linked as fixed reference geometry. An axis perpendicular to the plane stays refused because it has no line in the plane. The sketch's own X and Y axes keep working as before.
* claim: #3523
* Linux 3D mouse: decode spacenavd's axes in its documented frame (#3523)
spacenavd reports x to the right, y up and z away from the user, with the
rotations about those axes. The decoder treated y as away and z as up, so
with a stock spacenavd push and lift traded places, as did sideways tilt
and twist. Map the daemon's axes to the raw HID frame as translation
[x, -z, -y] and rotation [rx, -rz, -ry].
The decode still runs before a device's calibration and the per-axis
motion choice, so both apply unchanged on top of the corrected reading.
The tests now move each daemon axis on its own, in both directions, so a
swap of two axes can no longer pass.
Adds DeepSeek and Qwen presets to both endpoint protocols: OpenAI-compatible (api.deepseek.com, dashscope-intl compatible-mode/v1) and Anthropic-compatible (api.deepseek.com/anthropic, dashscope-intl apps/anthropic).
Co-authored-by: Claude Code <noreply@anthropic.com>
* claim: #2217
* Dialogs: drag a value's label sideways to change it (#2217)
A value row's label (expr_row, hint_row) now scrubs a literal value: about a
hundredth of its size per point of drag, on that step's grid, Shift ten times
finer, Ctrl/Cmd ten times coarser. The unit and decimal mark stay as typed, a
size never turns negative, and expressions and parameters are left alone. The
text changes as if typed, so previews follow and OK still makes one command.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Label drag: measure from the press point, cap values, keep scrub kit-private
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* fix: Revolve axis origin takes expressions instead of silently zero
- AxisRef::Line origin is a PointExpr evaluated with parameter values in the design unit (format note in pending)
- An unreadable origin or dir is a BadParams error instead of (0, 0, 0)
- Edit Feature keeps the saved expressions of an untouched origin or dir, for solid and surface revolves, in moved components too
- Face role names stay stable with a parameter-driven axis
* chore: Re-run PR checks after the description fix
* fix: Edit Feature of a revolve maps nothing where nothing is placed
- unplaced designs skip world_state and the pick table, as before the merge
- the untouched-edit test checks that OK applied the edit (CodeRabbit)
---------
Co-authored-by: P'tite Tête <asso@ptitetete.com>
* claim: #8220
* Agent chat: Custom instructions in Preferences, added to the system note (#8220)
Preferences > AI gets a multi-line Custom instructions field, saved with
the other AI preferences (agent_instructions, kept by page resets like
other user-written text). The chat adds the text after its own system
note when a session starts, by the routes #9811 set up: ACP session/new
(the Claude adapter's _meta key, the standard systemPrompt field, the
Antigravity key) and the system message or field of every API protocol.
Connect and New chat set it, so an edit applies from the next chat and
never changes the one running. Blank text adds nothing; at most 4000
characters are kept, in the field, over ui.agent_chat and from a hand-
edited settings file. ui.agent_chat takes and reports `instructions`.
* Custom instructions: typing in the field stays in the field
The text area did not report itself to Preferences search as a key sink, so with
the field focused, search took the typed characters and moved focus away. The
field now notes the sink while it has focus, like the other preference fields,
and the #8220 test types into it.
---------
Co-authored-by: MateuszSychXpl <151755914+MateuszSychXpl@users.noreply.github.com>
The tool face of Split Body is now a face pick: it is named when it is
picked and found again by that name, as other face picks are, so a split
by a rail's face follows the rail when a parameter moves it.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #12066
* Hole: a hole on a face that another part touches drills its own part (#12066)
A hole placed by position with no direction drills against the outward
normal of the face at the point. That face was looked for on every body,
so where another part touched the face at the hole, the part made first
won. The hole then drilled away from the part `targets` named and cut only
the tool's run-in pocket into it, without an error.
A hole that names its bodies now goes by their faces wherever one of them
passes through its position, when it is made (hole_direction_at) and when
a design is evaluated (hole_dir_into_face). Saved holes whose direction
leaves their own body's face there are mirrored into it, as format 18 does
for a thin plate's far side; the pending format note records it. A hole
whose bodies have no face at its position (placed on another part's face
to drill the part behind it) goes by every body's faces, as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PVKpSUSGYjbrBhfZGjqZEE
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The value text is drawn level, in an upright box, and the line runs
through its middle. The layout stopped the line where the box's farthest
corner reaches along it (half.x*|u.x| + half.y*|u.y|), not where the line
leaves the box (min(half.x/|u.x|, half.y/|u.y|)). They agree only when the
line is level or upright: at 45 degrees a radius leader stopped about
12 px short of its text, and an aligned dimension's line broke too wide
around its value.
One helper, box_exit, gives where the line leaves the box, for the radial
leader, the linear break and the leader to a text pulled past the
extension lines. Whether the text fits between the arrows still goes by
the whole box. A text box that isn't finite counts as empty, and one
larger than a sketch can be is held to MAX_COORD.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #10428
Signed-off-by: Maurizio Monticelli <maurizio@whatwedo.ch>
* Primitives: a pick field waits for its pick and says what it is for
Clicking Face, Reference 1 or 2 in the Cylinder window, or Point in the
Box, Sphere and Torus windows, and then clicking something the field
does not take (empty space, the cylinder's own preview) gave the
keyboard to Diameter, so the field no longer looked active and nothing
said why the click picked nothing. A pick field clicked into now keeps
the keyboard until a click picks something, and a click it does not
take leaves a note under the form saying what the field is for.
These fields also get their own tooltips: Face is the face or vertex
the cylinder stands on, Reference 1 and 2 are straight edges of that
face the distances below are measured from, and Point is a vertex,
edge, face, sketch point or the origin whose point fills X, Y and Z.
Fixes#10428
Signed-off-by: Maurizio Monticelli <maurizio@whatwedo.ch>
---------
Signed-off-by: Maurizio Monticelli <maurizio@whatwedo.ch>
* claim: #6818
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* Hole: the handle drawn with the shared marker (#6818)
A Single hole's drag handle (Move holes by clicking or dragging) is drawn with the placement
marker from place_gizmo/look.rs instead of its own ring and cross, so a hole and a Rest on the
same model show one language for the same action. place_gizmo::marker is the one crate-visible
door: it runs the marker's hover and drag motions under the handle's id and draws the crosshair,
never dimmed. Hole keeps its snapping, the click that takes a hole out, its handle names and its
hover text.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* Hole: hole_handle removed (#6818)
The shared marker draws the handle now; nothing else used it.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
* Hole: the marker takes no colours it does not use (#6818)
place_gizmo::marker took a &Tokens only to hand it to crosshair, which reads its colours itself,
so the hole's show fetched tokens nothing read. The argument and that fetch go.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Hole: a test sees the handle drawn as the shared marker (#6818)
The hole tests found the handle by its published position, which the look does not change. This
one reads what the frame paints round it: the white dot and four dashed arms at rest, no ring
round the centre, arrowheads when pointed at, solid shafts while it drags. It fails with the old
ring and cross back on the call line.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Place handles: marker hover eases in 120 ms, drag in 220 ms (Refs #6818)
The marker answered the pointer in 220 ms and a drag in 360 ms, which felt slow on review. Both
stay in look.rs as the one place a later preference can read them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
* Evaluation: a construction axis moved past a feature makes it evaluate again (#9544)
The state digest in a feature's cache key left out the construction
geometry before it, and the context digest hashes construction features
without their order. A feature started from the same key with or without
the axis before it, so after a move it kept its old result, and a cached
state restored whole brought back the construction list it was made with.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Evaluation: say plainly what an empty construction list hashes (#9544)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
* claim: #9236
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* GPU path tracer: don't rerun a setup that crashed in the driver (#9236)
A driver that takes the process down while it compiles the path tracer
(RADV on an RX 7600) is beyond error scopes and unwind guards. A running
setup now keeps gpu-path-setup.json (its adapter key) in the settings
folder, removed when the setup answers or the app closes normally. A
launch that finds it for the same adapter leaves the setup failed with an
explanation and a Try again button; CPU ray tracing stays available, and
a driver update changes the key and so tries again.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* Appearance: a picture of your own behind the model
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Background picture: a label that fits its row
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* claim: #11949
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Contact sets: input and output schemas for the seven contact commands (#11949)
describe_command now gives the exact parameters and reply of
contact.enable_sets, enable_all, disable_all, create, edit, list and
check, and the warn and strict parameter checks use them. The handlers
read their parameters from the raw JSON, so the contracts in
schema/motion.rs read each key as its handler does, with its message:
`occurrences` (OccurrenceKeys) and `set` are required in the schemas
but left to the handlers' messages in the types, through a small
`require` transform. The replies are built from typed values instead of
json! literals, with the same JSON; the listed sets mirror the doc
crate's ContactSet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
A 2D TRIMMED_CURVE in a pcurve read its basis straight through read_curve,
past the depth check that only the cached entry points (curve_in, curve_2d,
surface_in) made. A trimmed curve whose basis is itself, or a loop of two,
recursed until the stack overflowed and the app closed.
The depth check now sits in read_curve and read_surface, which every curve
and surface read passes through, so any chain of references, a cycle
included, ends at MAX_GEOM_DEPTH with an error that names the entity:
"curve #9: definitions nested more than 12 deep, or referring to
themselves". The other nested entities (surface, seam, intersection and
bounded surface curves, pcurves, 3D offset curves, offset, trimmed,
revolution and extrusion surfaces) already went through the checked entry
points; they now share the one check.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>