⚠️ Status (2026-05-19): This is a public snapshot of EngiAI accompanying the IDETC 2026 paper. The codebase is actively evolving in a private repository and will be made fully public as soon as the journal extension is published. For the smoothest experience, follow the demo setup in README.demo.md. The full version works but the RAG (MMORE) integration currently requires manual setup — official Docker images for RAG will be released soon.
EngiAI provides an AI-powered assistant for mechanical engineering design, integrating topology optimization, research retrieval, HPC simulation, and 3D printer control in a single conversational interface.
📄 Paper (arXiv): https://arxiv.org/abs/2605.19743
🎥 Demo video: https://www.youtube.com/watch?v=QbQVZFCq3X0
New here? Follow the lightweight demo build in README.demo.md — chatbot + Postgres only, no external dependencies, works on Windows / macOS / Linux.
Full stack (Docker):
git clone https://github.com/gioelemo/EngiAI.git
cd EngiAI
cp .env.example .env # add OPENAI_API_KEY / GOOGLE_API_KEY / TAVILY_API_KEY
docker compose up -dThen open http://localhost:8501. The RAG (MMORE) service requires separate setup — see docs/source/installation.md.
The supervisor routes requests to specialized agents:
- Engineering — topology optimization with EngiBench (beams2d, ThermoElastic2D, Photonics2D)
- Search — web research via Tavily
- RAG — document Q&A with MMORE multimodal RAG
- ArXiv — scientific paper search and analysis
- Prusa — 3D printer control via MCP (optional)
- HPC — SLURM job management over SSH
- CLI — local shell with human-in-the-loop confirmation
Details and examples: docs/source/usage/agents.md.
Full documentation lives in docs/. A hosted documentation site is planned (in the style of https://engibench.ethz.ch).
| Topic | Link |
|---|---|
| Installation (Docker + conda) | installation.md |
| Quick start | quickstart.md |
| Configuration (env vars, LLM models, tracing) | configuration.md |
| Architecture | architecture.md |
| Using agents | usage/agents.md |
| HPC integration (SLURM / SSH) | usage/hpc.md |
| 3D printer (Prusa) | usage/prusa.md |
| Web UI | usage/ui.md |
| Benchmarking with Weave | weave_integration.md |
| Deployment | deployment.md |
| Database setup | database_setup.md |
| Paper import (RAG) | paper_import_guide.md |
| Troubleshooting | troubleshooting.md |
| Contributing | contributing.md |
To build the docs locally: make docs (or make docs-watch for live preview).
src/ # agents, tools, UI, models
benchmarks/ # evaluation framework and per-problem prompts
services/ # standalone services (host_service, Prusa MCP)
docs/ # full Sphinx documentation
scripts/ # utility scripts
tests/ # unit and integration tests
- Troubleshooting guide
- GitHub Issues
- Container logs:
docker compose logs -f