Skip to main content

Google Gemini

Supported

  • LLM: Gemini 3.5 Flash (current recommended), 3.1 Flash Lite, 2.5 Flash/Pro; multimodal
  • Realtime speech-to-speech: gemini-3.1-flash-live-preview via the Gemini Live API (experimental relay mode)
  • Streaming supported

Setup

import { GeminiLLMProvider } from '@llmrtc/llmrtc-provider-google';

const llm = new GeminiLLMProvider({
  apiKey: process.env.GOOGLE_API_KEY,
  model: 'gemini-3.5-flash' // current recommended; provider default is gemini-2.5-flash
});

Env vars

  • GOOGLE_API_KEY
  • Optional: GOOGLE_MODEL

Notes

  • Strong on vision tasks; Flash is cost-effective for real-time voice.

Realtime speech-to-speech (experimental)​

GeminiLiveSpeechProvider connects sessions to the Gemini Live API for native voice-to-voice conversations, used with the server's realtime relay mode. The adapter handles Gemini's ~10-minute socket lifetimes internally (session resumption with input buffering, so reconnects don't clip user speech):

import { GeminiLiveSpeechProvider } from '@llmrtc/llmrtc-provider-google';

const provider = new GeminiLiveSpeechProvider({
  apiKey: process.env.GOOGLE_API_KEY,
  model: 'gemini-3.1-flash-live-preview' // default
});

Note: the Gemini Live API is a Google preview; this adapter is experimental and conformance-tested against the documented wire format.