SpeechStack
Submit a template
← All templatestemplates / multi-agent-voice-framework-for-customer-service-and-chat-supervision
Customupdated May 26, 2025 · other · support

Multi Agent Voice Framework for Customer Service and Chat Supervision

A demonstration framework for building advanced voice AI agents using OpenAI's Realtime API and Agents SDK. Features two core patterns: chat-supervisor (where a voice agent delegates complex tasks to a text-based supervisor model) and sequential handoffs (where specialized agents transfer users between them). Includes examples for customer service, authentication, returns processing, and sales with state machine prompting and output guardrails.

No demo yetView sourceFork template
The numbers
latency
cost / min
frameworkCustom
The stack
telephonyWeb Only
speech-to-textOpenAI Whisper
llmGPT-4o
text-to-speechOpenAI TTS-1
System prompt
No prompt published.
Config
config.json
{
  "agents": [
    "authentication",
    "returns",
    "sales",
    "simulatedHuman",
    "greeter",
    "haikuWriter"
  ],
  "features": [
    "multi_agent_orchestration",
    "tool_calling",
    "state_machine_prompting",
    "output_guardrails",
    "agent_handoffs"
  ],
  "patterns": [
    "chat-supervisor",
    "sequential-handoffs"
  ],
  "chat_model": "gpt-4o-realtime-mini",
  "push_to_talk": true,
  "escalation_model": "o4-mini",
  "supervisor_model": "gpt-4.1",
  "voice_activity_detection": true
}
Tags
multi-agentcustomer-serviceagent-handoffschat-supervisorstate-machineguardrailstypescriptnextjsopenai-agents-sdkrealtime-api
Voice Notes

Voice AI recipes, picks, and analysis.

Get the useful new templates plus the occasional teardown of what’s working in production voice AI.

contributed by @openai · MIT · source: github discoverylanguages: en-US