The kernel failure is reported as such since #381; the failure itself came
from split_by_plane's oversized boxes, which #384 replaced with body-sized
ones. The regression test now requires the cut (x = 110, 148, reversed
normal, the diagonal plane) from commands, with the cells' volumes adding up.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
The projection layout gave every fillet a blend layer along the sides it runs between, so a
screw hole close to a rounded corner sat inside the layer and the plan fell back to frame cuts
and swept cells, which fan into the hole. A corner fillet that lies inside the square from the
hole to the two sides is now taken in by the hole's O-grid ring: the frame reaches both sides,
its outer corner projects onto the middle of the fillet, and no layer is needed. An O-grid's
frame is centred on its hole (a near side is the frame, the far side as far), and the plan
reports the layout's frames and drops the frame-cut notes when it does not cut.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
--section x|y|z[=VALUE] hides the elements on the camera's side of a plane
(at the mesh's middle by default) so the element faces inside show;
--field contours a nodal field of the solved step (results.contour). The
MCP server's headless screenshot draws with the viewport's pipeline on the
shared device and its reply, like snapshot's, names the renderer.
Refs #422
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
fieldcraft-cli snapshot script.json --out shot.png [--view] [--size WxH]
draws the engine's framed scene (engine::view::frame, shared with the CPU
render_png) with the viewport's GPU pipelines offscreen: hardware GPU, then
a software one, then the CPU rasterizer (FIELDCRAFT_GPU=auto|hardware|
software|cpu). CI installs lavapipe so the GPU path runs there.
Refs #422
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
Topology::split cuts a block at given fractions and every block chained to it
through parallel edges, orienting the fractions through the chain, so the
topology stays conforming. A certified-invalid element is reported with its
block, cell, centre and the block vertex and blocks meeting there; the mesh is
never emitted.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
The section comes from a set (a 2D mesh or element faces), the mesh already
on the source faces (tagged, or a neighbouring body's boundary lying on
them, whose nodes are shared so the meshes conform), or a fresh quad mesh
of one planar face. Target and side faces project through the tet mesher's
B-rep projection. Replaces the body's mesh like the other meshers, tags the
body's own source, target and side faces, and is on the MCP mesh tool as
hex_extrude.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
Every tied interface between cells of one design body carries `why`: the tet cell (and why
auto fell back), the reason the body's cells did not sweep together, or the cell meshed alone
by its method; so an agent knows which cut or method to change.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
A partition intersects the body with a box on each side of the plane as
large as the body itself. SolidCraft's split_by_plane uses boxes several
times larger, and its boolean then put the cut up to 1.8 um off the plane
on the enclosure's air, so the cut faces were not recognized and the cells
did not close. A plane along a face of a body it cuts, or a cell whose
boundary does not tessellate into one a mesher takes (a cut through a
hole's axis), is a warning naming the partition, the body and where.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
Deleting, suppressing or editing a partition (or its datum), or cutting a
meshed body, removes the meshes of the bodies that are gone, reported as
replaced, in the same undo step. model.check lists meshes of partition
cells the open design no longer has (a model saved before the cut
changed), and mesh.clear removes them, or the meshes of named bodies.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
A node on the secondary surfaces of two ties (a chip's underside and its
sides, a cut face and an outer face) got one constraint per tie. When the
main nodes of one follow the other's main surface, the second constraint is
nearly the first again, and eliminating it divided by round-off: the Pi 5
heat model put 21.9 W more in than its 8 W. Such a node now follows the
main faces of all its ties together (balanced to 5e-11 of the power), a
node following a node that follows it back (a tie each way on one
interface) is an error naming both ties, and a tie row that still nearly
repeats the constraints before it is an error naming them.
mesh.cells ties one way by a fixed order of the cells (hex main to tet)
and makes one tie per tied cell, tie_<cell>, replacing the ties an earlier
run made, so its own ties never hold a node twice.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
hex_mesh_swept_cells lays out one cross-section for several cells: every
cell's walls (its cut faces included) are imprinted together, every cell's
levels are levels of all, one interval assignment and one set of layers,
and each region of each level interval is filled by the cell it lies in.
Cells meet node for node across every cut, T-junctions included.
mesh.cells sweeps the cells of each design body together first and meshes
cell by cell only what does not sweep together (saying why); the reply
gives every cell's volume, scaled Jacobian and edge ratio.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
split_by_plane gives one body per side, which may hold several shells when
the side falls apart; every lump is now its own cell (B/a1, B/a2), at every
cut, as the partition docs promise.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
The drop-test and impact samples have no supports, by design, yet the browser marked Loads &
BCs as the next stage with an Add boundary condition row even after they were solved.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
results.export.csv's form listed only nodes and sets, so the parameter check refused the
probe's set, at, mode and frame and the reactions' about point that the command documents.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
Four field kinds in the schema, each checked and described in JSON Schema: labels (nodes or
elements, typed as 1-5, 8 with Use selection taking the viewport's picks), points (rows of
x, y, z, added and removed), multi (checkboxes: coupling DOFs, step output) and number or
auto (the deformed scale). Node and element lists, probe and Delaunay points and block
corners use them; the remaining JSON fields hold geometry-specific shapes.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
results.export.csv runs solve.probe or results.reactions and writes their reply as an RFC 4180
CSV: a row per node (node, x, y, z and the quantities the step has, columns it does not have
left out) or per set with the total and the applied load; numbers in shortest round-trip form,
so they read back bit for bit. A stale result is refused. The results panel has Export CSV…
beside Reaction sums and under the probed node: the save picker, or the form without one.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
results.contour takes range {min?, max?} (or "auto"), kept until the field changes: a
singular peak no longer squeezes the field into the bottom bands. Values above a clamped
maximum go to an extra band drawn in the over colour, split exactly at the clamp; the legend
shows that band up to the true maximum and says it is clamped. The results panel has Legend
max/min (Auto or a value), the MCP show_results tool a range argument, and headless pictures
follow the clamp too.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
diagnostics.crash_reports.delete deletes one report or all. The window lists each report
(when, what, how many times) with Report…, Copy Path and Delete, and Delete All; ui.prompt
answers it over the control channel.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
The README video's model as a sample: the corpus cantilever-i-beam design opened as geometry,
swept all-hex, clamped at F2:start, 1 kN on F2:top through a kinematic coupling; static step
Load. It has a Start-card thumbnail and a gallery capture; a test checks its tip deflection
against beam theory with shear.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
file.new replaces the session with an empty one: no design, mesh, definitions, results or
journal, nothing to save, the undo history empty. The File menu and Cmd/Ctrl+N run it, asking
first over unsaved changes. A sample name the app does not know (a typo, an old ?sample= link)
opens the samples to choose from, and the status names it.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
Persistent names grow with a design's split history; opening accepted them but mesh tags, sets
and loads capped geometry names at 256 characters. One cap (64 KiB, against hostile input)
now holds for geometry references, partitions and saved views.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
- DisplayMesh::from_model_hiding leaves out a hidden body's elements (and the
nodes only they use); Session::hidden_elements names them, so the viewport,
probing and headless pictures skip hidden bodies (#365).
- A solve summary's applied and reaction totals set the round-off of
cancelling sums to 0 (#319).
- results::principal for the probe's principal stresses (#329).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
- Quantity names the unit of every number the results show from the model's
unit system (mm, MPa, N, N·mm, mJ, mW, °C or K); Field::title puts it in the
legend title, and calls a mode shape normalised (#321).
- A solved step's result stays current through edits that leave what it reads
alone (Model::step_inputs: other steps and what only they use), and comes
back current on undo (#318).
- Session::step_glyphs: the symbols of what a solved step used, for drawing
over its results; no load arrows on a frequency step's modes (#327).
- Nonlinear static steps open deformed at true scale like explicit ones, and
their automatic scale is 1 (auto_scale, shared by the viewport and the
headless pictures) (#326).
- results.reactions with an empty sets list sums the step's boundary-condition
sets, as its form says (#328).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
- geometry.bodies lists each body's vertices; geometry.datum.point by
vertex takes the design's vertex names (labelled Vertex, not Mesh
vertex) (#346).
- geometry.centre: the point for what is picked: a vertex, nodes, a
hole's centre (cylindrical faces or circular edges), else a centroid
(#351).
- mesh.cells' form: the cells first, and a method drop-down per cell
(schema kind RefChoices) instead of JSON (#344).
- interaction.rigid_wall(.delete): rigid walls of an explicit step from
the Interactions tab, no longer coming soon (#350).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
SolidCraft names the pieces of a split body by stacking suffixes
(Bracket (1) (1)' (1)). On opening a design, the pieces of one body are
named <body>.1 … <body>.N, ordered by centroid, so the tree, tooltips,
forms and picks read as one part with numbered cells.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
The design file stores no measure to compare with, so the rebuild's own
failures are the guard: each failed feature lists the bodies it names (all of
them when it names none), and geometry.open replies with a warning per failure,
which the window shows with the open.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
- A solve or a mesher runs as a job with one progress mechanism: the overlay
shows the step, stage, count and elapsed time with a bar when bounded, OK
waits while it runs, and the status sums it up (elements, smallest angle,
time) (#322, #357).
- Over the control channel such a command is a job too: answered when it
ends however long it runs, or at once with wait: false, then job.status,
job.wait, job.cancel; ui.inspect gives job; MCP waits for it (#367).
- A command that cannot run yet lists what it misses in the context panel,
each with a button to the command that provides it (#323).
- A cancelled or refused solve stays in its workspace (#324, test).
- A mesher form starts from the parameters of the mesh it would replace (#343).
- Seed Edges and edge/vertex/face datum points switch the selection filter
while open and restore it (#341).
- Datum forms draw the datum live, the Partition form shows its datum, and
new datums and partitions get the next free name (#347).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
The setup checklist marks the mesh Outdated (a warm ⚠ in the browser, with
what changed and where to remesh) when seeds, partitions or datum planes
changed after the last mesher ran, and model.check warns (#342). Replacing
a mesh asks with the element type change it makes, such as tet10 → tet4 (#343).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
solve.heat's reply and the results report give heat_flows: each convection,
flux and source load's heat into the body, the prescribed temperatures'
total and their sum; the results panel lists them for a heat step.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01UhhZMBTviDQxZvwnyP89fo
