Show HN: SIMD Agent – AI that runs OpenFOAM simulations from natural language (github.com)

A new AI tool called SIMD Agent has been introduced, enabling users to run OpenFOAM simulations through simple natural language prompts. By describing the desired physics—such as flow and heat transfer—along with a mesh of the physical domain, the AI autonomously selects the appropriate governing equations, turbulence models, and sets up necessary parameters. This significant advancement allows what used to take a week for a computational fluid dynamics (CFD) specialist to resolve into a prompt and a geometry file, drastically streamlining the simulation process and enhancing design-decision support. The SIMD Agent features a self-healing loop that diagnoses and corrects solver failures by employing a smaller language model to focus on errors, and it streams live data, including residuals and flow fields, via WebSocket. The tool supports various turbulence models and is open-source under the AGPL v3 license, making it accessible for a broad range of applications. It can operate in Docker mode or bare-metal mode, and integrates multiple LLMs like Gemini and Vertex AI, promoting flexibility in its deployment and use. This innovation marks a notable leap in making complex simulations more user-friendly and efficient for engineers and researchers in the AI/ML community.