Configure a Pipeline

The Pipeline is what your agent hears and speaks through. Pass the components you need (STT, LLM, TTS, plus optional VAD and turn detection) and the pipeline auto-detects its mode from what you give it.

Pipeline flow from user audio through input processing to cascading, realtime, or hybrid mode, then audio output

from zrt import Pipeline
from zrt.plugins import DeepgramSTT, GoogleLLM, CartesiaTTS, SileroVAD, TurnDetector
from zrt.inference import AICousticsDenoise

pipeline = Pipeline(
    stt=DeepgramSTT(),
    llm=GoogleLLM(),
    tts=CartesiaTTS(),
    vad=SileroVAD(),
    turn_detector=TurnDetector(model="echo-large"),
    denoise=AICousticsDenoise(model_id="quail-vf-2.2-l-16khz"),
)

Pass pipeline to your Agent, then run it.

Configure it further

Modes

Cascade, Realtime, and Hybrid, chosen automatically from the components you pass.

Hooks

Tap into the pipeline at runtime to inspect or transform each stage.

Fallback Adapter

Fail over between providers to keep sessions resilient.

What’s Next

Run Your Agent

Tools and Capabilities

Give the agent function tools and external services.

References

Python

Examples

Cascade Basic

Minimal STT-LLM-TTS cascade agent you can run.

SDK Reference

pipeline.py

pipeline.py in the Python API reference.

Agent Invoke

Get started

Modalities

Configure Pipeline

Tools and Capabilities

Turn Detection and Interruptions

Telephony Agents

Context Management

Observability and Monitoring

Configure it further

Modes

Hooks

Fallback Adapter

What’s Next

Run Your Agent

Tools and Capabilities

References

Examples

Cascade Basic

SDK Reference

pipeline.py

​Configure it further

Modes

Hooks

Fallback Adapter

​What’s Next

Run Your Agent

Tools and Capabilities

​References

​Examples

Cascade Basic

​SDK Reference

pipeline.py

Configure it further

What’s Next

References

Examples

SDK Reference