Server Configuration

Complete reference for LLMRTCServer configuration options.

Core Options

Option	Type	Default	Description
providers	ConversationProviders	required	LLM, STT, TTS provider instances
port	number	8787	TCP port to listen on
host	string	'127.0.0.1'	Bind address
systemPrompt	string	'You are a helpful realtime voice assistant.'	Base system prompt
historyLimit	number	8	Messages retained in context
streamingTTS	boolean	true	Enable streaming TTS (requires FFmpeg)
heartbeatTimeout	number	45000	Milliseconds before disconnect on no heartbeat
cors	CorsOptions	undefined	CORS configuration

interface ConversationProviders {
  llm: LLMProvider;
  stt: STTProvider;
  tts: TTSProvider;
  vision?: VisionProvider;  // Optional for image analysis
}

---

Playbook Mode Options

When using playbooks for multi-stage voice agents with tools:

Option	Type	Default	Description
playbook	Playbook	undefined	Playbook definition with stages/transitions
toolRegistry	ToolRegistry	undefined	Registry with tools (required with playbook)
playbookOptions	object	see below	Playbook orchestrator options

Playbook Orchestrator Options (Server Level)

At the server level, only the following options are supported:

interface ServerPlaybookOptions {
  /** Maximum tool calls per turn in Phase 1 */
  maxToolCallsPerTurn?: number;  // Default: 10

  /** Timeout for Phase 1 tool loop (ms) */
  phase1TimeoutMs?: number;  // Default: 60000 (1 minute)

  /** Enable debug logging */
  debug?: boolean;  // Default: false
}

Core SDK Options

Additional options like historyLimit and llmRetries are available when using PlaybookOrchestrator directly via the core SDK. See Core SDK Playbooks for details.

---

ICE Server Configuration

WebRTC requires ICE servers for NAT traversal. See Networking & TURN for details.

Option	Type	Description
iceServers	RTCIceServer[]	Custom STUN/TURN servers
metered	object	Metered TURN API configuration

// Option 1: Metered TURN (recommended)
const server = new LLMRTCServer({
  providers: { llm, stt, tts },
  metered: {
    appName: 'your-app',    // From Metered dashboard
    apiKey: 'your-key',     // API key
    region: 'us_east'       // Optional region preference
  }
});

// Option 2: Custom ICE servers
const server = new LLMRTCServer({
  providers: { llm, stt, tts },
  iceServers: [
    { urls: 'stun:stun.example.com:3478' },
    {
      urls: 'turn:turn.example.com:3478',
      username: 'user',
      credential: 'pass'
    }
  ]
});

---

Observability Options

Option	Type	Description
hooks	ServerHooks & OrchestratorHooks	Event callbacks for logging/monitoring
metrics	MetricsAdapter	Metrics adapter (Prometheus, DataDog, etc.)

import { createTimingHooks, ConsoleMetrics } from '@llmrtc/llmrtc-core';

const server = new LLMRTCServer({
  providers: { llm, stt, tts },
  hooks: createTimingHooks(),
  metrics: new ConsoleMetrics()  // For debugging
});

See Hooks & Metrics for complete hook reference.

---

Streaming TTS Options

Option	Type	Description
streamingTTS	boolean	Enable low-latency streaming (default: true)
sentenceChunker	(text: string) => string[]	Custom sentence boundary detection

Custom Sentence Chunker:

For non-Latin languages or custom punctuation:

const server = new LLMRTCServer({
  providers: { llm, stt, tts },
  streamingTTS: true,
  sentenceChunker: (text) => {
    // Custom logic for CJK languages
    return text.split(/[。！？]/g).filter(Boolean);
  }
});

---

Complete Example

import {
  LLMRTCServer,
  OpenAILLMProvider,
  OpenAIWhisperProvider,
  ElevenLabsTTSProvider,
  ToolRegistry,
  createTimingHooks
} from '@llmrtc/llmrtc-backend';

const server = new LLMRTCServer({
  // Required
  providers: {
    llm: new OpenAILLMProvider({ apiKey: process.env.OPENAI_API_KEY! }),
    stt: new OpenAIWhisperProvider({ apiKey: process.env.OPENAI_API_KEY! }),
    tts: new ElevenLabsTTSProvider({ apiKey: process.env.ELEVENLABS_API_KEY! })
  },

  // Server
  port: 8787,
  host: '0.0.0.0',

  // Conversation
  systemPrompt: 'You are a helpful voice assistant.',
  historyLimit: 8,
  streamingTTS: true,

  // Playbook mode (optional)
  playbook: myPlaybook,
  toolRegistry: myTools,
  playbookOptions: {
    maxToolCallsPerTurn: 10,
    phase1TimeoutMs: 60000,
    debug: false
  },

  // ICE servers
  metered: {
    appName: 'my-app',
    apiKey: process.env.METERED_API_KEY!
  },

  // Observability
  hooks: createTimingHooks(),

  // Heartbeat
  heartbeatTimeout: 45000
});

await server.start();

---

Related

• Environment Variables - CLI/env-based configuration
• Networking & TURN - ICE server setup
• Hooks & Metrics - Observability