Apps Overview¶
AtomGPT.org hosts 50+ web applications for materials science. Each app provides a web interface, REST API endpoints, and Python examples for agentic workflows.
Categories¶
| Category | Apps | Description |
|---|---|---|
| Explore | 22 | Browse databases, search materials, visualize properties |
| Build | 4 | Construct structures, design workflows, write code |
| Predict | 6 | ML predictions, force fields, quantum algorithms |
| Characterize | 3 | Match spectra, analyze images, identify structures |
| Apply | 5 | Screen materials for specific applications |
| Validate | 2 | Verify AI outputs, check references |
Architecture¶
Every app follows the same pattern:
┌─────────────┐ ┌──────────────┐ ┌─────────────┐
│ HTML/JS │────▶│ FastAPI │────▶│ JARVIS │
│ Frontend │◀────│ Backend │◀────│ Data/ML │
└─────────────┘ └──────────────┘ └─────────────┘
│ │
▼ ▼
Web Browser REST API
(for agents/scripts)
- Frontend: Single HTML file with embedded CSS/JS, dark theme
- Backend: FastAPI route with authentication, served from
custom_routes/ - Data: JARVIS-DFT (76K materials), ALIGNN models, Wannier TBH databases