Import CAD, build or edit geometry, generate a mesh, assign materials and boundary conditions, and run structural, thermal, and modal analysis from one workspace. Every UI action becomes editable Python, while Sesame can automate the same visible setup, solve, and post-processing steps. Advanced workflows extend the same model state to CFD, FSI, contact, nonlinear dynamics, flexible multibody dynamics, PINNs, and remote TorsoHPC solves.
OCCT handles geometry; Gmsh, structured, and octree paths handle meshing; DOLFINx, Kratos, MFEM, and Torso's native assembly share a common model contract. TorsoCAE provides one browser interface over these proven numerical backends without hiding which backend or solver is executing.
Browser-native CSG modeling. Boolean operations on STEP bodies with stable entity IDs for simulation-ready geometry.
Gmsh-powered meshing with hex-dominant, transfinite, and Delaunay algorithms. Automatic refinement based on curvature.
DOLFINx, Kratos, and MFEM backends for structural, thermal, CFD, FSI, and contact simulations — swap backends without changing your setup. Linear and nonlinear, serial or MPI-parallel.
Differentiable surrogate modeling powered by PhysicsNemo. Fast, predictive analysis for design space exploration.
Click your way through, replay a Python journal, or ask the AI. All three paths produce identical state.
Every action you take is also written to a journal entry — so exploration is automatically a recording. No "save script" step, no separate macro recorder mode.
Saved as plain Python with stable body_0 / body_1 references — readable, diffable, parametrisable. session.csg_transform("body_2", "rotate", {...}).
Ask Sesame to create CSG bodies, group faces, assign materials, mesh with HXT, apply loads, solve, select result fields, or run a report. Its commands appear in the same replayable journal, where you can inspect and edit them.
The browser is the engineering workspace. Heavy geometry, meshing, and numerical work runs on explicit server and remote-compute layers.
Viewport, model tree, selections, materials, boundary conditions, solver controls, live progress, and post-processing.
CAD translation and healing, meshing, physics assembly, journal execution, project storage, and result streaming.
Dense, sparse, distributed DCSR, direct, iterative, and eigensolver paths receive numerical operators from TorsoCAE.
CAD, mesh identity, physics setup, journals, results, and reconnect state remain part of the TorsoCAE project contract.
TorsoCAE owns geometry, mesh, physics, and post-processing. Selecting a TorsoHPC solver routes the assembled numerical problem to the remote solver framework and returns the solution to the same model and result pipeline.
Explore TorsoHPCClick faces in the viewport. Group the ones that should share a boundary condition. Assign a fixed support, a traction, a pressure, or a temperature. Face tags survive across CAD edits and mesh rebuilds.
TorsoCAE exposes the numerical choices and preserves the artifacts needed to reproduce or audit a run.
Backend, algorithm, preconditioner, device, core count, tolerances, and time-integration controls remain visible and journaled.
Live status, available residual histories, convergence failures, and solver warnings are surfaced instead of silently replaced by another method.
Save the project and Python journal, download numerical results as NPZ, capture viewport images, and run field, reaction, contact, line, or surface reports.
Compute credits for only what you use.
Current metering is execution-based: one credit per successful server execution second, with a 0.01-credit minimum for a charged operation. Browsing, documentation, and browser-only interactions do not consume credits; failed server operations are not charged.
