cfdpilot 0.1.4

AI agent for OpenFOAM — diagnose, fix, and relaunch cases from your terminal

CFDpilot

An AI agent for OpenFOAM — in your terminal.

CFDpilot reads your OpenFOAM case directly from disk, diagnoses why it diverges or gives wrong physics, proposes the exact fix as a colored diff, and — once you confirm — patches the files and reruns the solver. No ZIP upload, no browser. It runs where you already work: your laptop, your workstation, or an HPC node over SSH.

pip install cfdpilot
cfdpilot login                 # authenticate once
cd path/to/your/case && cfdpilot

Then just ask, in plain English:

"Why won't my rhoSimpleFoam case converge?"

The agent maps your case, reads fvSchemes, fvSolution, your boundary conditions and solver log, reasons from the actual numbers (Courant number, residuals, turbulence wall treatment, scheme/switch consistency), and reports the bugs ranked by impact — with the exact fix for each. Nothing is changed without showing you a diff first, and every patch makes a .bak backup.

What it checks

Boundary-condition consistency, turbulence wall treatment (low-Re vs high-Re), Courant number and time-step control, relaxation factors, discretisation schemes, solver-mode switches (e.g. transonic), and solver/turbulence-model compatibility — across incompressible, compressible, multiphase (VoF) and heat-transfer solvers.

Requirements

• Python 3.10+
• An OpenFOAM case directory (Foundation or ESI). OpenFOAM itself can run natively or in Docker; the agent will use it to run the solver when you ask.

Access

CFDpilot is in free early access for the first engineers. Run cfdpilot login to get started, or request access at cfdpilot.com.