Run Local and Remote AI Models in One Agent

What This Video Covers

One agent, two models. A small local model triages every request, and a router sends each one where it belongs: greetings and short factual questions to a local model, coding and multi-step reasoning to a hosted model registered as an OpenAI-compatible API. The reasons are the ones you'd expect — keep sensitive requests on your machine, and stop paying API prices for easy work.

It is built from the Hybrid Reasoner sample. The triage model answers with a minified JSON route, a transform node combines that route with the original question, and the router follows it to the local or the remote branch. Two test runs show both paths, and the trace shows which model answered each time.

Try It Yourself

• Remote models
• The router node
• Sample agents