Python Voice Agent Examples

A collection of production-ready voice AI agent examples built with Plivo. Each example demonstrates a different combination of AI models and frameworks for building real-time phone-based voice agents on the Plivo voice AI platform.

How It Works

All examples follow the same general pattern:

┌─────────┐ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐
│ Phone │────▶│ Plivo │────▶│ Server │────▶│ AI Agent │
│ Call │◀────│ (Voice AI) │◀────│ (FastAPI) │◀────│ │
└─────────┘ └─────────────┘ └─────────────┘ └─────────────┘

1. A phone call comes in (or is initiated) through Plivo
2. Plivo hits a webhook on your FastAPI server
3. The server establishes a bidirectional WebSocket for audio streaming
4. The AI agent processes speech and generates responses in real-time

Examples

Each example directory is self-contained with its own dependencies, environment configuration, and documentation. Directory names follow the convention {llm}-{stt}-{tts}-{framework} (see CONTRIBUTING.md for details).

Speech-to-Speech (S2S)

These examples use models that handle both speech input and output natively — the simplest architecture with the fewest moving parts.

Example	Model	Framework	Highlights
gemini2.5-live-native	Gemini 2.5 Live	None	Direct API integration, function calling, auto-webhook config
gemini2.5-live-pipecat	Gemini 2.5 Live	Pipecat	Modular pipeline, built-in VAD, less code
gptrealtime1.5-native	GPT Realtime 1.5	None	Silero VAD, barge-in support, function calling
grok3-voice-native	Grok 3 Voice	None	Silero VAD, barge-in support, function calling

STT + LLM + TTS Pipeline

These examples wire up separate providers for speech-to-text, language model, and text-to-speech — offering more flexibility to mix and match.

Example	STT	LLM	TTS	Framework
gemini2-deepgramnova2-cartesiasonic2-native	Deepgram Nova 2	Gemini 2	Cartesia Sonic 2	None
gemini2-deepgramnova2-elevenflashv2.5-native	Deepgram Nova 2	Gemini 2	ElevenLabs Flash v2.5	None
gpt4o-deepgramnova3-openaitts4o-pipecat	Deepgram Nova 3	GPT-4o	OpenAI TTS 4o	Pipecat
gpt5.4-assemblyaiu3-cartesiasonic3-native	AssemblyAI U3	GPT 5.4	Cartesia Sonic 3	None
daily-plivo	Deepgram	OpenAI	Cartesia	Pipecat + Daily

Prerequisites

• Python 3.10+ (3.12 recommended)
• uv package manager (recommended) or pip
• ngrok for local development
• A Plivo account with a phone number
• API keys for the AI services used by your chosen example

Quick Start

1. Choose an example from the tables above and navigate to its directory:

   cd gemini2.5-live-native # or any other example

2. Install dependencies:

   uv sync
   # or: uv pip install -r requirements.txt

3. Configure environment variables:

   cp .env.example .env

   Edit .env with your API keys and Plivo credentials.

4. Start ngrok (in a separate terminal):

   ngrok http 8000 # port varies by example

5. Update PUBLIC_URL in .env with your ngrok HTTPS URL.

6. Run the server:

   uv run python server.py # entry point varies by example

7. Call your Plivo phone number to talk to the agent.

See each example's README for detailed setup and configuration.

Contributing

Want to add a new voice agent example? See CONTRIBUTING.md for the project naming convention, required structure, and submission guidelines.

License

MIT