Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Foam-Agent: An end-to-end, composable multi-agent framework for automating CFD simulations in OpenFOAM. NeurIPS 2025 Machine Learning and the Physical Sciences Workshop.
| Date | Stars |
|---|---|
| 2026-07-25 | 280 |
| 2026-07-28 | 280 |
| 2026-07-30 | 280 |
| 2026-07-31 | 284 |
| 2026-08-06 | 291 |
Today
+7 stars today
This week
+11 stars this week
This month
— stars this month
Momentum
29.0
growth rate 3.93%/day
# Foam-Agent <a href="https://arxiv.org/abs/2505.04997"><img src="https://img.shields.io/badge/arXiv-2505.04997-b31b1b.svg" alt="Paper"></a>
<p align="center">
<img src="overview.png" alt="Foam-Agent System Architecture" width="800">
</p>
<p align="center">
<em>An End-to-End Composable Multi-Agent Framework for Automating CFD Simulation in OpenFOAM</em>
</p>
**Foam-Agent** automates the entire **OpenFOAM**-based CFD simulation workflow from a single natural language prompt. It manages meshing, case setup, execution, error correction, and post-processing — dramatically lowering the expertise barrier for Computational Fluid Dynamics. Evaluated on [FoamBench](https://arxiv.org/abs/2509.20374) with 110 simulation tasks, our framework achieves an **100% success rate** with Claude Opus 4.6.
Visit [deepwiki.com/csml-rpi/Foam-Agent](https://deepwiki.com/csml-rpi/Foam-Agent) for a comprehensive introduction and to ask questions interactively.
## Key Features
- **End-to-End Automation**: From meshing (including external Gmsh `.msh` files) to HPC job submission to ParaView/PyVista visualization — one prompt does it all.
- **Multi-Agent Workflow**: Architect, Input Writer, Runner, and Reviewer agents collaborate through a LangGraph pipeline with automatic error correction (up to 25 iterations).
- **RAG-Enhanced Generation**: Hierarchical FAISS indices built from OpenFOAM tutorials provide context-specific retrieval for accurate configuration file generation.
- **Composable Service Architecture**: Core functions are exposed as MCP tools, enabling integration with Claude Code, Cursor, and other agentic systems.
## Quick Start
### 1. Pull and run the Docker image
```bash
docker run -it \
-e OPENAI_API_KEY=your-key-here \
-p 7860:7860 \
--name foamagent \
leoyue123/foamagent
```
The container comes with OpenFOAM v10, Conda, and all dependencies pre-installed.
> For a specific release: `docker pull leoyue123/foamagent:v2.0.0`
### 2. Write your prompt
Edit `user_requirement.txt` inside the container:
```text
do a Reynolds-Averaged Simulation (RAS) pitzdaily simulation. Use PIMPLE algorithm.
The domain is a 2D millimeter-scale channel geometry. Boundary conditions specify a
fixed velocity of 10m/s at the inlet (left), zero gradient pressure at the outlet
(right), and no-slip conditions for walls. Use timestep of 0.0001 and output every
0.01. Finaltime is 0.3. use nu value of 1e-5.
```
### 3. Run
```bash
python foambench_main.py --output ./output --prompt_path ./user_requirement.txt
```
That's it. Foam-Agent will plan the case, generate all OpenFOAM files, run the simulation, and fix errors automatically.
## Configuration
All settings live in `src/config.py` with sensible defaults. Every setting can be overridden via environment variables — no need to edit files, especially useful for Docker and CI.
### LLM Provider and Model
| Environment Variable | Purpose | Allowed Values |
|---|---|---|
| `FOAMAGENT_MODEL_PROVIDER` | LLM backend | `openai`, `openai-codex`, `anthropic`, `bedrock`, `ollama` |
| `FOAMAGENT_MODEL_VERSION` | Model identifier | e.g., `gpt-5-mini`, `gpt-5.3-codex`, `claude-opus-4-6` |
Example:
```bash
docker run -it \
-e FOAMAGENT_MODEL_PROVIDER=anthropic \
-e ANTHROPIC_API_KEY=your-key-here \
-e FOAMAGENT_MODEL_VERSION=claude-opus-4-6 \
-p 7860:7860 \
leoyue123/foamagent
```
### Embedding Provider and Model
| Environment Variable | Purpose | Allowed Values |
|---|---|---|
| `FOAMAGENT_EMBEDDING_PROVIDER` | Embedding backend | `openai`, `huggingface`, `ollama` |
| `FOAMAGENT_EMBEDDING_MODEL` | Embedding model | e.g., `Qwen/Qwen3-Embedding-0.6B`, `text-embedding-3-small` |
Defaults to `huggingface` with `Qwen/Qwen3-Embedding-0.6B` (runs locally, no API key needed).
### API Keys
| Variable | When needed |
|---|---|
| `OPENAI_API_KEY` | Using `openai` provider |
| `ANTHROPIC_API_KEY` | Using `anthropic` provider |
| AWS credentials | Using `bedrock` provider |
### Input Writer GenerExcerpt of 12,375 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:e185be64ee28f691, topic:multi-agent-systems, desc:agent framework, readme:agent framework