-
Notifications
You must be signed in to change notification settings - Fork 0
Home
Massimiliano Lupo Pasini edited this page May 2, 2026
·
2 revisions
Multi-agent AI framework for atomistic materials simulation and discovery.
matsim-agents orchestrates large language models (LLMs), machine-learned interatomic potentials (MLIPs), and ASE-based atomistic workflows into a single agentic loop. The user states a research objective in natural language; agents plan, run HydraGNN-driven simulations, score chemical and dynamical stability, and report the findings — with optional human review at every gate.
| Page | Description |
|---|---|
| CI-CD | GitHub Actions workflows, branch protection, how to run tests locally |
| Contributing | PR process, required checks, code style, review rules |
┌──────────────────────────────────────────────┐
│ USER │
│ natural-language objective / chat dialogue │
└───────────────────────┬──────────────────────┘
│
┌───────────────────────▼──────────────────────┐
│ LangGraph workflow │
│ │
│ planner ───► executor ──┐ │
│ ▲ │ │
│ └────────┤ while pending │
│ ▼ │
│ analyst ──► END │
└───────────────────────┬──────────────────────┘
│ tool calls
┌───────────────────────▼──────────────────────┐
│ Discovery wrapper │
│ composition parsing → phase enumeration │
│ → relaxation (HydraGNN+ASE) → stability │
└───────────────────────┬──────────────────────┘
│
┌───────────────────────▼──────────────────────┐
│ Atomistic backends │
│ HydraGNN (fused MLFF + BranchWeightMLP) │
│ ASE (FIRE / BFGS / BFGSLineSearch) │
│ pymatgen (optional prototypes) │
└──────────────────────────────────────────────┘
Maintained by the ORNL Multi-Agentic AI for Materials team.