Nexus

Nexus is Simudyne's AI-powered model builder. It provides a five-stage pipeline that takes you from an initial research question to a validated, deployed simulation model — without needing to write the model from scratch.

How it works

Discover → Template-select → Refine → Generate → Validate

1. Discover — Nexus searches a knowledge base of 446 landmark agent-based models (ABM Wiki) and 252 system dynamics models (SD Wiki), all structured as Composable Model Schema (CMS) entries. You describe your problem; Nexus surfaces the closest matches.
2. Template-select — You pick the best-fit model from the ranked candidates. Nexus loads its CMS specification as the starting point.
3. Refine — You work with the Nexus architect to add mechanisms, adjust parameters, and set constraints. All refinements stay within the CMS schema.
4. Generate — Nexus auto-generates Python SDK-compliant simulation code from the finalised CMS specification.
5. Validate — Generated models are QA'd against empirical targets and stochasticity budgets defined in the CMS.

Built on

Nexus runs on the Simudyne Python SDK (abm-lab v0.7.0). Models generated by Nexus are valid abm-lab simulations — you can run, extend, and deploy them using the full Python SDK toolchain.

In this section

• Installation — install Nexus and abm-lab from Simudyne's private package index
• Deployment — run Nexus in Docker, Docker Compose, or Kubernetes
• Chat API — query simulation results in natural language via the AI chat endpoint
• Model Library — the CMS format, ABM Wiki, and SD Wiki reference