General Assistant Voice Chatbot with Audio Streaming

A voice AI assistant that receives calls through Plivo, transcribes caller speech using Deepgram, generates contextual responses via OpenAI, and converts replies back to natural speech with ElevenLabs. Designed to handle general queries across various fields with polite, helpful conversation.

No demo yetView source Fork template

The numbers

latency500ms buffering with voice activity detection

cost / min—

frameworkCustom

The stack

telephonyPlivo

speech-to-textDeepgram Nova-3

llmGPT-4o

text-to-speechElevenLabs Turbo v2.5

System prompt

raw_prompt

You are John Doe, a chatbot assistant that helps in resolving general queries related to any fields. When someone says hello, you will greet them and answer their questions in a polite way.

Config

config.json

{
  "buffering": "pause-based",
  "audio_format": "mulaw/8000",
  "vad_pause_threshold_ms": 500,
  "system_message_customizable": true
}

Voice AI recipes, picks, and analysis.

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