Skip to main content

​
API Reference

VoiceGateway exposes three distinct API surfaces, each designed for a different integration point.

​
Python SDK

The voicegateway.inference module is the public Python surface. It is a drop-in mirror of livekit.agents.inference: change the import line in an existing LiveKit Cloud Inference agent and the rest of the code keeps working. Cost tracking, latency monitoring, and session correlation happen transparently in the middleware. 
from livekit.agents import AgentSession

from voicegateway.inference import STT, LLM, TTS

session = AgentSession(
    stt=STT("deepgram/nova-3"),
    llm=LLM("openai/gpt-4.1-mini"),
    tts=TTS("cartesia/sonic-3"),
)
Best for: application code, scripts, custom integrations. Full Python SDK reference

​
HTTP API

The REST API runs on port 8080 (default) via voicegw serve. It provides CRUD operations for providers, models, and projects, plus read-only endpoints for costs, latency, logs, and Prometheus-format metrics. The dashboard frontend consumes this API, and external monitoring tools can scrape /v1/metrics. 
curl http://localhost:8080/v1/status
curl http://localhost:8080/v1/costs?period=week&project=my-app
Best for: dashboards, monitoring, CI/CD pipelines, multi-language teams. Full HTTP API reference

​
Dashboard API

The dashboard API is mounted by the daemon (voicegw serve) under the /api/ prefix on the same port as the HTTP API and the React SPA. It exposes a smaller set of read-only /api/* endpoints optimised for the dashboard UI. These endpoints aggregate data slightly differently from the HTTP API (for example, /api/overview combines multiple queries into a single response). 
curl http://localhost:8080/api/overview
curl http://localhost:8080/api/costs?period=today
Best for: the built-in web dashboard (consumed automatically). Full Dashboard API reference

​
Choosing the Right API

Use caseAPI surface
Route voice AI requests in PythonPython SDK
Manage providers/models/projects remotelyHTTP API
Build a custom dashboard or integrate with monitoringHTTP API
Use the built-in web UIDashboard API (automatic, served by the daemon)
Integrate with AI coding agentsMCP server