OpenAI Agents SDK Deep Research Demo

This demo extends the Deep Research workflow from OpenAI Agents Python SDK Demos by adding a browser-based UI, making it easier to experience the full multi-agent research pipeline in action. It's designed to be run side-by-side with the Temporal UI to show durable execution in real time.

When given a research query, a Triage Agent can route the workflow to a Clarifying Agent for follow-up questions, or if the query is already specific enough, let research continue without them. An Instruction Agent processes the responses, then hands off to a Planner Agent that creates a web search plan. Multiple Search Agents execute queries in parallel, and a Writer Agent compiles the findings into a comprehensive markdown report.

Every agent invocation and tool call is backed by Temporal, meaning the workflow is fault-tolerant, observable, and resumable — even if it takes several minutes to complete. This makes it a practical reference for teams building production-grade research agents with the OpenAI Agents SDK.