EngiAI Documentation

EngiAI Documentation#

A Jarvis-style multimodal AI assistant for closed-loop design-for-manufacturing correction.

EngiAI is a sophisticated multi-agent system that combines:

  • 🐳 Docker Deployment: Production-ready containerized deployment

  • 🤖 Multi-Agent System: Supervisor coordinates specialized agents

  • 🔧 Engineering Tools: EngiBench integration for structural optimization

  • 📚 RAG System: Document Q&A with MMORE multimodal RAG service

  • 🖨️ 3D Printer Integration: Prusa Connect integration via MCP

  • 🖥️ HPC Integration: SLURM job management for remote compute clusters

  • 🔍 Research Tools: ArXiv search and web research capabilities

  • 💬 Interactive UI: Streamlit web interface with chat and visualization

Features at a Glance#

Multi-Agent Architecture#

The system uses a supervisor-based architecture where specialized agents handle specific tasks:

  • Supervisor Agent: Routes requests to appropriate specialized agents

  • Engineering Agent: Structural optimization and topology design

  • RAG Agent: Knowledge base queries with retrieval-augmented generation

  • ArXiv Agent: Research paper search and analysis

  • Search Agent: Web research and information gathering

  • HPC Agent: High-performance computing job management

  • Prusa Agent: 3D printer control and monitoring

  • CLI Agent: Command-line tool execution

See System Architecture for detailed system design.

Key Capabilities#

Engineering & Optimization

  • Topology optimization with EngiBench (beams2d, photonics2d, thermoelastic2d)

  • STL export for 3D printing

  • Design simulation and constraints checking

  • Multi-physics optimization

Research & Knowledge

  • ArXiv paper search and download

  • RAG-based document Q&A with MMORE

  • Multimodal document processing

  • PDF processing and chunking

3D Printing

  • Prusa Connect integration

  • Print job submission and monitoring

  • File management

  • Printer status tracking

HPC & Remote Computing

  • SSH-based cluster access

  • SLURM job submission

  • Job monitoring and output retrieval

  • File transfer (SFTP)

Web Interface

  • Interactive chat with agents

  • Multi-chat session management

  • File uploads (PDFs, designs, data)

  • Real-time visualization

  • W&B report embedding

Technology Stack#

  • Frameworks: LangGraph, LangChain, Streamlit, Fabric

  • LLMs: OpenAI GPT-4o/GPT-4.1, Google Gemini-3-Flash, Ollama (Qwen3, Qwen3.5)

  • Storage: MMORE (RAG), PostgreSQL/SQLite (relational)

  • Integrations: EngiBench/EngiOpt, Prusa Connect, Weights & Biases, DuckDuckGo/Tavily

Deployment Options#

🐳 Docker (Recommended for production)

  • Isolated environment with all dependencies

  • Easy deployment with docker-compose

  • Optional Prusa MCP server integration

  • PostgreSQL database included

💻 Local Development (For active development)

  • Conda environment with Python 3.13+

  • Hot reload for code changes

  • SQLite database

  • Full IDE integration

See Installation for setup instructions.

Getting Help#

Indices and tables#