Workflow for AI-Assisted Model Generation
Five steps: prepare the AI assistant → describe the problem in natural language → review the metafiles → generate the QuickField problem → open and continue working.
QuickField uses AI to help prepare a simulation model, while the model's creation, verification, field computation and subsequent editing all remain transparent and under human control.
Prepare the AI Assistant
Give your AI assistant the QuickField Skills Archive, and ask it to use this material when generating QuickField Builder metafiles.
The Skills Archive contains documentation, examples, tests and implementation details describing the three JSON metafile types: .qfp, .qfs and .qfd.
Describe the Problem
Describe the simulation model you need in natural language. For example:
“Build a magnetostatic analysis model of a hollow solenoid, outer diameter 20 mm, inner diameter 10 mm, length 100 mm, current 100 A.”
The AI assistant then generates the corresponding JSON metafiles: .qfp for the problem definition, .qfs for the geometric model, and .qfd for materials, field sources and boundary conditions.
(Optional) Review the Generated Metafiles
Before running the computation, you can review the generated JSON files to check that the geometry, materials, field sources, boundary conditions and analysis settings match the intended problem statement.
Generate the QuickField Problem
Place the related .qfp, .qfs and .qfd metafiles in the same directory, then run QuickField Problem Builder — open the .qfp file with the standard Windows file dialog, or pass the .qfp file as a command line parameter.
Problem Builder converts the JSON metafiles into native QuickField files through the QuickField API.
problem_builder.exe solenoid.qfpOpen and Continue Working
Depending on the settings in the .qfp file, Problem Builder can also start solving automatically.
You can then open and analyze the generated QuickField problem, continue working from it, or adjust it as needed using the regular QuickField user interface or any automation method.
Human Oversight and Intervention
Human oversight and intervention are possible at every stage:
- When defining the task in natural language
- When reviewing or editing the generated JSON metafiles
- When checking or adjusting the generated QuickField model
This lets AI help prepare the model while keeping the model-generation process transparent and under control. Once generated, the model is a regular QuickField problem that can be modified, solved, post-processed or automated through the standard workflow.