Skip to content

AI Copilot

The Copilot can inspect your model, answer questions about it, and propose edits.

There is no bundled AI subscription. You supply a key for a provider you already use:

  • Anthropic (Claude)
  • OpenAI, and any OpenAI-compatible endpoint - which covers Grok, Gemini, local models through Ollama, and self-hosted deployments
  • Demo mode, offline, for trying the interface without a key

Your key is stored in your browser and used to call that provider directly. Sending a message needs a free account; the Copilot is available on every plan.

It has tools to read the model - element counts, properties, results, validation issues - so questions like “which conduits are surcharged and what do they have in common” are answered from your actual model rather than from generic knowledge.

When the Copilot suggests a change, it does not apply it. You get a diff preview showing exactly what would change, and you choose whether to apply.

Applied changes go through the normal undo stack, so Ctrl+Z reverses them like any other edit.

Good at: explaining what a result means, finding patterns across many elements, drafting repetitive edits, recalling SWMM parameter meanings and reasonable ranges.

Not good at: engineering judgement. It does not know your local standards, your client’s constraints, or which of two defensible approaches is right here. Treat it as a capable assistant whose work you check, not as an authority.