A walkthrough of a multi-agent pipeline for deep research tasks. A triage agent evaluates incoming queries for specificity, routing vague ones to a clarifying agent and specific ones to a planner agent. The planner delegates to a search agent that runs concurrent searches, then passes results to a writer agent that synthesizes everything into a final report. A brief code pattern shows how to define individual agents with a name, instructions, and model, executed via a runner.
•1m watch time
Sort: