Conversational Voice Agent with Tool Calling

A conversational voice agent that listens to speech, transcribes it in real-time, processes it through an OpenAI agent with tool-calling support, and speaks the response back. Supports both single-turn demo mode and interactive streaming for continuous conversation.

No demo yetView source Fork template

The numbers

latency—

cost / min—

frameworkCustom

The stack

telephonyWeb Only

speech-to-textDeepgram Nova-3

llmGPT-4.1

text-to-speechDeepgram Aura-2

System prompt

No prompt published.

Config

config.json

{
  "llm_model": "gpt-4.1-mini",
  "stt_model": "nova-3",
  "tts_model": "aura-2-asteria-en",
  "smart_format": true,
  "interim_results": true,
  "streaming_supported": true,
  "tool_calling_supported": true
}

Voice AI recipes, picks, and analysis.

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