gemini-rs

Full Rust SDK for the Gemini Multimodal Live API — wire protocol, agent runtime, and fluent DX in three layered crates.

┌─────────────────────────────────────────────────────┐
│  gemini-adk-fluent-rs (L2 — Fluent DX)                    │
│  AgentBuilder · Live · S·C·T·P·M·A · .govern/.on_enter │
├─────────────────────────────────────────────────────┤
│  gemini-adk-rs (L1 — Agent Runtime)                       │
│  Agent · Tools · State · Phases · TextAgent · LLM  │
│  Governed Agents: Flow · Extract · Resolver        │
├─────────────────────────────────────────────────────┤
│  gemini-genai-rs (L0 — Wire Protocol)                     │
│  Transport · Session · Protocol · VAD · Buffers    │
└─────────────────────────────────────────────────────┘

Governed Agents — Flow (control DAG), Extract (deterministic + async fact resolution), and Resolver (orchestration) are L1 runtime primitives, not a layer above the fluent API. L2 surfaces them ergonomically (Live::govern, .extract_record, .on_enter). They compose over one shared State spine — see the Governed Flows, Extraction, and Agent Orchestration chapters.

Quick Start

use gemini_adk_fluent_rs::prelude::*;

// `connect_from_env()` resolves Google AI vs Vertex AI from the environment —
// no auth ceremony. See "Authentication & Connecting" for the variables it reads.
let handle = Live::builder()
    .model(GeminiModel::Gemini2_0FlashLive)   // native-audio Live model
    .voice(Voice::Kore)
    .instruction("You are a helpful voice assistant.")
    .on_audio(|pcm| { /* play audio */ })
    .on_text(|text| print!("{text}"))
    .connect_from_env()
    .await?;

handle.send_text("What's the weather like?").await?;
handle.disconnect().await?;

New here? Start with Setup and Running and Authentication & Connecting, then browse the cookbook (Crawl → Walk → Run).

Guide Structure

This book is organized into six sections:

  • Getting Started — Local setup, architecture overview, migration guide, and best practices
  • Voice & Live Sessions — Building real-time voice agents with phases, state, and watchers
  • Tools & Extraction — Tool system, deterministic + LLM extraction, MCP
  • Composition & Patterns — Governed Flows, Agent Orchestration, text-agent combinators, S·C·T·P·M·A operators, middleware
  • Examples — 40 progressive cookbook examples (Crawl/Walk/Run/Governed) plus interactive gemini-adk-web-rs demos
  • ADK Web UI — Design system, dark/light mode, DevTools panels, and the cookbook browser

API Reference

For detailed type and method documentation, see the rustdoc API reference.

CrateLayerAPI Docs
gemini-genai-rsL0 — Wire Protocolgemini_genai_rs
gemini-adk-rsL1 — Agent Runtimegemini_adk_rs
gemini-adk-fluent-rsL2 — Fluent DXgemini_adk_fluent_rs