Hey folks, I’m adding a new project to my Learn OpenHands Harness course focused on model routing. The lesson has students run the same set of coding tasks three ways: 1. Single model baseline 2. Static routing with RouterLLM 3. Cascading routing with OpenHands profiles + SwitchLLMTool I’d love feedback on whether this is the right way to present the OpenHands mechanisms, especially the split between RouterLLM for static harness-controlled routing and SwitchLLMTool for model changes during an active conversation. https://github.com/rajshah4/learn-openhands-harness/tree/main/projects/p09-model-routing-benchmark

left a small PR mentioning the OpenHands index https://github.com/rajshah4/learn-openhands-harness/pull/7/ will look at the rest of the tutorial this week

That's a great idea. Thanks.

@average-processor-75195 this looks good! On parallel: • I've set up a meta harness to optimize a prompt that classifies models. see https://github.com/OpenHands/research/blob/main/juan/intelligent-model-selection/optimization_200_4cat/classifier_prompt_iter5.txt • I'm testing that prompt on the benchmarks https://openhands-eval-monitor.vercel.app/?date=2026-06-12&days=15&text=router+classified

Very cool! Lots of people are excited about doing this.

Yes, I’m aiming for user modifiable with a good predefined example (s)