Skip to main content
A plugin handles one stage of the pipeline. You import it from zrt.plugins (or the matching path in your SDK) and pass it to the pipeline. Each plugin page below shows setup and usage in Python, JavaScript, and Go. Plugins are grouped by stage, because that’s how you choose them: pick the best speech-to-text, LLM, and text-to-speech for your agent.

Plugins by provider

A “Supported” cell works today but is not documented yet; ”-” means that provider doesn’t serve that stage. Availability can vary by SDK, so check each plugin page. Turn detection (Echo), voice activity detection (Silero), and noise cancellation (RNNoise, Sanas, ai-coustics) are listed below.

Speech-to-text

LLM

Text-to-speech

Realtime models

Speech-to-speech models that run in the pipeline’s llm slot (no separate STT/TTS). See Realtime mode.

Turn detection

Voice activity detection

Noise cancellation

Plugins marked Supported work today and are added the same way; their pages will follow. Availability can vary by SDK, so check your SDK’s plugin list.