docs(sdk): add Route Task to Model guide - #824
juanmichelini wants to merge 2 commits into
Conversation
Document the route_task_to_model tool (ClassifyAndSwitchLLMTool) and its meta-profile shapes (structured classes + direct prompt), referencing the new examples/01_standalone_sdk/59_route_task_to_model.py in software-agent-sdk PR #4287. Co-authored-by: openhands <openhands@all-hands.dev>
|
Preview deployment for your docs. Learn more about Mintlify Previews.
💡 Tip: Enable Automations to automatically generate PRs for you. |
|
@OpenHands please look at https://github.com/OpenHands/openhands and notice that we are using names Router Pro and Router Flash instead of Pareto update docs accordingly |
|
I'm on it! juanmichelini can track my progress at all-hands.dev |
The OpenHands app brands the two built-in direct-prompt router meta-profiles as OpenHands Router Pro and OpenHands Router Flash (renamed from the max-score/min-cost Pareto templates in OpenHands#17782. Update the guide prose and the synced example comment accordingly. Co-authored-by: openhands <openhands@all-hands.dev>
|
Addressed in 9ec3c7c. Updated the guide to reference the OpenHands app's renamed built-in routers, OpenHands Router Pro and OpenHands Router Flash, instead of the old Pareto prompt meta-profile names — both in the prose and in the synced example comment. (This PR was created by an AI agent (OpenHands) on behalf of @juanmichelini.) |
|
Done. Here's a summary of what I did to address the review comment on PR #824: What the comment asked@juanmichelini pointed out that the OpenHands app now uses the names Router Pro and Router Flash (the two built-in router meta-profiles, formerly known as "max-score Pareto"/"min-cost Pareto" — renamed in Changes made (
|
Summary
Adds a new SDK guide, Route Task to Model, documenting the
route_task_to_modeltool (ClassifyAndSwitchLLMTool) introduced in software-agent-sdk PR OpenHands/software-agent-sdk#4287.The guide covers both meta-profile shapes the tool supports:
{description, model}rows; classifier returns a class index.prompt_template+model_table; classifier returns the model name as JSON.It references the new ready-to-run example
examples/01_standalone_sdk/59_route_task_to_model.pyand is registered in the "LLM Features" group indocs.json(after "Model Routing").This unblocks the
check-examplesCI check on the agent-sdk PR (the check requires each example to be referenced from a docs.mdxfile).AI disclosure: This PR was created by an AI agent (OpenHands) on behalf of @juanmichelini.
Checklist
.mdxguide added undersdk/guides/docs.jsonllm-routing.mdx)