ModelScript is a completely web-native, polyglot incremental compiler supporting multiple engineering domains — from requirements and system architecture to simulation, CAD, CAE, and manufacturing. By natively supporting Modelica, SysML v2, STEP, OWL2, and OpenFOAM (CFD), ModelScript enables seamless cross-domain simulation-based requirements verification.

Designed for modern workflows, it features a robust engine for incremental compilation and flattening, seamless FMU import/export and integration, surrogate ROM generation from FMUs, as well as advanced model optimization and parameter calibration. All of this is accessible directly in the browser or via the command line.

• Salsa-Based Incremental Compiler: Powered by a query-based incremental compilation architecture (inspired by Rust's Salsa). ModelScript caches intermediate representations and only recompiles what has changed, enabling instantaneous feedback in the IDE.
• Arena Data-Oriented Simulator: The simulator utilizes a zero-garbage, Data-Oriented Design (DoD) architecture (ArenaDAEBuilder). By moving from legacy recursive AST visitors to direct flat-buffer indexing, it provides high-performance equation sorting, Pantelides index reduction, and bipartite matching with zero allocation overhead.
• Polyglot & Cross-Domain: Unified AST querying and cross-language interoperability between Modelica (Simulation), SysML v2 (Architecture/Requirements), and STEP (CAD/Manufacturing).

This project is a monorepo managed with Lerna, Nx, and npm workspaces. The structure reflects our transition to a fully modular, high-performance compiler and simulator architecture.

• Polyglot Incremental Compiler — completely web-native, query-based (Salsa) architecture supporting Modelica, SysML v2, and STEP for cross-domain engineering.
• Data-Oriented Simulation Engine — arena-native ODE/DAE solver featuring zero-garbage buffer allocation, Pantelides index reduction, and BLT ordering.
• Cross-Domain Verification — continuous simulation-based requirements verification spanning architecture, simulation, CAD, CAE, and manufacturing.
• Incremental Flattening — state-of-the-art incremental compilation and unrolling of hierarchical models into flat Differential Algebraic Equations (DAE).
• FMU & ROM — seamless FMU import/export, integration, and high-performance surrogate ROM generation from FMUs.
• Optimization & Calibration — direct collocation solvers for optimal control problems, model optimization, and parameter calibration.
• Accurate Parsing — custom Tree-sitter grammars for efficient, incremental parsing of all supported domains.
• Semantic Analysis — full scope and name resolution for complex multi-domain hierarchies.
• Diagram Rendering — interactive SVG diagrams and X6-based visual layouts with auto-placement.
• Language Server — rich completions, hover, diagnostics, formatting, and color provider.
• Linting & i18n — comprehensive lint rules and string extraction support.

• Node.js ≥ 24 (see .nvmrc)
• emsdk (required for building the Tree-sitter WASM parser):

  git clone https://github.com/emscripten-core/emsdk.git
  cd emsdk
  ./emsdk install latest
  ./emsdk activate latest
  source ./emsdk_env.sh

git clone https://github.com/modelscript/modelscript.git
cd modelscript
npm install

Build all packages (in dependency order via Nx):

npm run build

Start all services concurrently:

npm run dev

This launches:

The IDE includes a browser-local AI assistant powered by WebLLM (Qwen3-0.6B). To enable it, download the model weights (~350 MB, one-time):

npm run download-model --workspace=@modelscript/ide

Once downloaded, restart npm run dev and open the ModelScript AI panel in the IDE sidebar.

After building, the CLI is available as msc:

# Flatten a model to DAE incrementally
npx msc flatten Modelica.Electrical.Analog.Examples.CauerLowPassAnalog path/to/MSL

# Simulate a model (outputs CSV by default)
npx msc simulate BouncingBall model.mo --stop-time 5

# Simulate with JSON output
npx msc simulate BouncingBall model.mo --format json

# Solve an optimal control problem
npx msc optimize MyModel model.mo \
  --objective "u^2" --controls "u" --control-bounds "u:-1:1" --stop-time 10

# Lint Modelica files
npx msc lint model.mo

# Render a diagram to SVG (outputs to stdout)
npx msc render MyModel model.mo > diagram.svg

Run the test suite across all packages (including the new compiler tests):

npm test

npm run lint
npm run format

Pre-built images are published to the GitHub Container Registry on every push to main.

Run the latest images without building:

docker compose pull    # Pull latest images from ghcr.io/modelscript/*
docker compose up -d   # Start containers
docker compose down    # Stop containers
docker compose logs -f # Tail logs

To build from source instead:

npm run docker:build   # Build images locally
npm run docker:up      # Start containers

ModelScript is licensed under the AGPL-3.0-or-later. See COPYING for more details.