Download
This repository is a customer service agent demo built using the OpenAI Agents SDK to showcase how to build a production-style conversational assistant for use cases like airline customer support. It consists of two major parts: a Python backend that orchestrates agent logic (tool calls, handoffs, memory, routing) and a Next.js UI for chat interaction and visualizing agent state. The demo covers tasks you’d expect in customer service: changing flights, checking status, answering FAQs, etc. It shows how multiple subagents can be coordinated under a triage agent that decides which specialized agent should handle a given request. The UI includes visualization of which agent is active, routing logic, and conversation tracking. It also demonstrates guardrails to validate or constrain responses, memory usage to maintain context, and tracing to help debugging of workflows.
Features
- Backend logic for orchestrating multiple agents in a customer service flow
- Frontend (Next.js) UI for chat, state visualization, and agent monitoring
- Triage agent that routes requests among subagents specialized by task
- Guardrails and validation to prevent hallucinations or unsafe outputs
- Conversation memory and context management across turns
- Tracing and logging of agent decisions and tool usage for debugging
Project Samples
