@drawdream/livespeech 0.1.7 → 0.1.8

package/README.md

@@ -74,6 +74,7 @@ client.disconnect();
 connect() → startSession() → audioStart() → sendAudioChunk()* → audioEnd() → endSession()
                                     ↓
                           sendSystemMessage() (optional, during live session)
+                          sendToolResponse() (when toolCall received)
 ```
 
 | Step | Description |
@@ -83,6 +84,7 @@ connect() → startSession() → audioStart() → sendAudioChunk()* → audioEnd
 | `audioStart()` | Begin audio streaming |
 | `sendAudioChunk(data)` | Send PCM16 audio (call multiple times) |
 | `sendSystemMessage(msg)` | Inject context or trigger AI response (optional) |
+| `sendToolResponse(id, result)` | Send function result back to AI (after toolCall) |
 | `audioEnd()` | End streaming, triggers AI response |
 | `endSession()` | End conversation |
 | `disconnect()` | Close connection |
@@ -105,6 +107,7 @@ await client.startSession({
   pipelineMode: 'live',           // 'live' (default) or 'composed'
   aiSpeaksFirst: false,           // AI speaks first (live mode only)
   allowHarmCategory: false,       // Disable safety filtering (use with caution)
+  tools: [{ name: 'func', description: 'desc', parameters: {...} }],  // Function calling
 });
 ```
 
@@ -117,6 +120,7 @@ await client.startSession({
 | `pipelineMode` | `'live' \| 'composed'` | `'live'` | Audio processing mode |
 | `aiSpeaksFirst` | `boolean` | `false` | AI initiates conversation (live mode only) |
 | `allowHarmCategory` | `boolean` | `false` | Disable content safety filtering |
+| `tools` | `Tool[]` | `undefined` | Function definitions for AI to call |
 
 ### Pipeline Modes
 
@@ -152,6 +156,73 @@ await client.startSession({
 
 > ⚠️ **Warning**: Only use in controlled environments where content moderation is handled by other means.
 
+## Function Calling (Tool Use)
+
+Define functions that the AI can call during conversation. When the AI decides to call a function, you receive a `toolCall` event and must respond with `sendToolResponse()`.
+
+### Define Tools
+
+```typescript
+const tools = [
+  {
+    name: 'open_login',
+    description: 'Opens Google Login popup when user wants to sign in',
+    parameters: { type: 'OBJECT', properties: {}, required: [] }
+  },
+  {
+    name: 'get_price',
+    description: 'Gets product price by ID',
+    parameters: {
+      type: 'OBJECT',
+      properties: {
+        productId: { type: 'string', description: 'Product ID' }
+      },
+      required: ['productId']
+    }
+  }
+];
+
+await client.startSession({
+  prePrompt: 'You are a helpful assistant. Use tools when appropriate.',
+  tools,
+});
+```
+
+### Handle Tool Calls
+
+```typescript
+client.on('toolCall', (event) => {
+  console.log('AI wants to call:', event.name);
+  console.log('With arguments:', event.args);
+  
+  if (event.name === 'open_login') {
+    showLoginModal();
+    client.sendToolResponse(event.id, { success: true });
+  }
+  
+  if (event.name === 'get_price') {
+    const price = getProductPrice(event.args.productId);
+    client.sendToolResponse(event.id, { price, currency: 'USD' });
+  }
+});
+```
+
+### Tool Interface
+
+```typescript
+interface Tool {
+  name: string;                    // Function name
+  description: string;             // When AI should use this
+  parameters?: {
+    type: 'OBJECT';
+    properties: Record<string, unknown>;
+    required?: string[];
+  };
+}
+```
+
+> ⚠️ **Note**: Function calling only works with `pipelineMode: 'live'`
+
 ## System Messages
 
 During an active live session, you can inject text messages to the AI using `sendSystemMessage()`. This is useful for:
@@ -223,6 +294,7 @@ const client = new LiveSpeechClient({
 | `response` | AI's response text | `text`, `isFinal` |
 | `audio` | AI's audio output | `data`, `sampleRate` |
 | `turnComplete` | AI finished speaking | `timestamp` |
+| `toolCall` | AI wants to call a function | `id`, `name`, `args` |
 | `error` | Error occurred | `code`, `message` |
 
 ### Simple Handlers
@@ -249,6 +321,13 @@ client.setAudioHandler((data: Uint8Array) => {
 client.setErrorHandler((error) => {
   console.error(`Error [${error.code}]: ${error.message}`);
 });
+
+// Tool calls (function calling)
+client.on('toolCall', (event) => {
+  // Execute function and send result
+  const result = executeFunction(event.name, event.args);
+  client.sendToolResponse(event.id, result);
+});
 ```
 
 ### Full Event API
@@ -283,6 +362,14 @@ client.on('turnComplete', () => {
 client.on('error', (event) => {
   console.error('Error:', event.code, event.message);
 });
+
+client.on('toolCall', (event) => {
+  // event.id: string - use with sendToolResponse
+  // event.name: string - function name
+  // event.args: object - function arguments
+  const result = handleToolCall(event.name, event.args);
+  client.sendToolResponse(event.id, result);
+});
 ```
 
 ## Audio Format
@@ -434,8 +521,10 @@ import type {
   ResponseEvent,
   AudioEvent,
   TurnCompleteEvent,
+  ToolCallEvent,
   ErrorEvent,
   ErrorCode,
+  Tool,
 } from '@drawdream/livespeech';
 ```
 

package/dist/index.d.mts

@@ -82,6 +82,58 @@ interface LiveSpeechConfig {
      */
     debug?: boolean;
 }
+/**
+ * Parameters schema for a function (for function calling)
+ */
+interface FunctionParameters {
+    /**
+     * Type of the parameters object
+     * @default "OBJECT"
+     */
+    type: 'OBJECT';
+    /**
+     * Properties of the parameters object
+     * Each key is a parameter name, and the value describes its type and description
+     * @example { "productId": { "type": "string", "description": "The product ID to look up" } }
+     */
+    properties: Record<string, unknown>;
+    /**
+     * List of required parameter names
+     */
+    required?: string[];
+}
+/**
+ * Tool definition for function calling
+ *
+ * Tools allow the AI to call functions defined by your application during conversation.
+ * When the AI decides to call a function, you'll receive a `toolCall` event.
+ *
+ * @example
+ * const tools: Tool[] = [{
+ *   name: 'open_google_login',
+ *   description: 'Opens the Google Login popup when user wants to sign in',
+ *   parameters: {
+ *     type: 'OBJECT',
+ *     properties: {},
+ *     required: []
+ *   }
+ * }];
+ */
+interface Tool {
+    /**
+     * Name of the function (must be unique within the session)
+     */
+    name: string;
+    /**
+     * Description of what the function does and when the AI should use it
+     */
+    description: string;
+    /**
+     * Parameters schema for the function
+     * If the function takes no parameters, use { type: 'OBJECT', properties: {}, required: [] }
+     */
+    parameters?: FunctionParameters;
+}
 /**
  * Session configuration options
  */
@@ -114,6 +166,23 @@ interface SessionConfig {
      * @default false
      */
     allowHarmCategory?: boolean;
+    /**
+     * Tools (functions) that the AI can call during conversation.
+     * When the AI calls a tool, you'll receive a `toolCall` event.
+     * You must respond with `sendToolResponse()` to continue the conversation.
+     *
+     * @example
+     * tools: [{
+     *   name: 'get_price',
+     *   description: 'Get the current price of a product',
+     *   parameters: {
+     *     type: 'OBJECT',
+     *     properties: { productId: { type: 'string', description: 'Product ID' } },
+     *     required: ['productId']
+     *   }
+     * }]
+     */
+    tools?: Tool[];
 }
 /**
  * Internal resolved configuration with defaults applied
@@ -132,7 +201,7 @@ interface ResolvedConfig {
 /**
  * Event types emitted by the LiveSpeech client
  */
-type LiveSpeechEventType = 'connected' | 'disconnected' | 'reconnecting' | 'sessionStarted' | 'sessionEnded' | 'ready' | 'userTranscript' | 'response' | 'audio' | 'turnComplete' | 'error';
+type LiveSpeechEventType = 'connected' | 'disconnected' | 'reconnecting' | 'sessionStarted' | 'sessionEnded' | 'ready' | 'userTranscript' | 'response' | 'audio' | 'turnComplete' | 'toolCall' | 'error';
 /**
  * Event payload for 'connected' event
  */
@@ -237,10 +306,40 @@ interface TurnCompleteEvent {
     type: 'turnComplete';
     timestamp: string;
 }
+/**
+ * Event payload for 'toolCall' event (function calling)
+ * Indicates the AI wants to call a function defined by your application.
+ * You must respond with `sendToolResponse()` to continue the conversation.
+ *
+ * @example
+ * client.on('toolCall', (event) => {
+ *   if (event.name === 'get_price') {
+ *     const price = await lookupPrice(event.args.productId);
+ *     client.sendToolResponse(event.id, { price });
+ *   }
+ * });
+ */
+interface ToolCallEvent {
+    type: 'toolCall';
+    /**
+     * Unique identifier for this tool call.
+     * Must be passed to `sendToolResponse()` to respond.
+     */
+    id: string;
+    /**
+     * Name of the function the AI wants to call.
+     */
+    name: string;
+    /**
+     * Arguments passed by the AI to the function.
+     */
+    args: Record<string, unknown>;
+    timestamp: string;
+}
 /**
  * Union type of all event payloads
  */
-type LiveSpeechEvent = ConnectedEvent | DisconnectedEvent | ReconnectingEvent | SessionStartedEvent | SessionEndedEvent | ReadyEvent | UserTranscriptEvent | ResponseEvent | AudioEvent | TurnCompleteEvent | ErrorEvent;
+type LiveSpeechEvent = ConnectedEvent | DisconnectedEvent | ReconnectingEvent | SessionStartedEvent | SessionEndedEvent | ReadyEvent | UserTranscriptEvent | ResponseEvent | AudioEvent | TurnCompleteEvent | ToolCallEvent | ErrorEvent;
 /**
  * Simplified event handlers for common use cases
  */
@@ -252,11 +351,11 @@ type ErrorHandler = (error: ErrorEvent) => void;
 /**
  * WebSocket message types sent from client to server
  */
-type ClientMessageType = 'startSession' | 'endSession' | 'audioStart' | 'audioChunk' | 'audioEnd' | 'systemMessage' | 'ping';
+type ClientMessageType = 'startSession' | 'endSession' | 'audioStart' | 'audioChunk' | 'audioEnd' | 'systemMessage' | 'toolResponse' | 'ping';
 /**
  * WebSocket message types received from server
  */
-type ServerMessageType = 'sessionStarted' | 'sessionEnded' | 'ready' | 'userTranscript' | 'response' | 'audio' | 'turnComplete' | 'error' | 'pong';
+type ServerMessageType = 'sessionStarted' | 'sessionEnded' | 'ready' | 'userTranscript' | 'response' | 'audio' | 'turnComplete' | 'toolCall' | 'error' | 'pong';
 /**
  * Base interface for client messages
  */
@@ -271,6 +370,9 @@ interface StartSessionMessage extends BaseClientMessage {
     prePrompt?: string;
     language?: string;
     pipelineMode?: 'live' | 'composed';
+    aiSpeaksFirst?: boolean;
+    allowHarmCategory?: boolean;
+    tools?: Tool[];
 }
 /**
  * End session message
@@ -319,10 +421,26 @@ interface SystemMessageMessage extends BaseClientMessage {
     action: 'systemMessage';
     payload: SystemMessagePayload;
 }
+/**
+ * Tool response payload
+ */
+interface ToolResponsePayload {
+    /** The tool call ID from the toolCall event */
+    id: string;
+    /** The result of the function execution */
+    response: unknown;
+}
+/**
+ * Tool response - send function execution result back to Gemini
+ */
+interface ToolResponseMessage extends BaseClientMessage {
+    action: 'toolResponse';
+    payload: ToolResponsePayload;
+}
 /**
  * Union type of all client messages
  */
-type ClientMessage = StartSessionMessage | EndSessionMessage | AudioStartMessage | AudioChunkMessage | AudioEndMessage | SystemMessageMessage | PingMessage;
+type ClientMessage = StartSessionMessage | EndSessionMessage | AudioStartMessage | AudioChunkMessage | AudioEndMessage | SystemMessageMessage | ToolResponseMessage | PingMessage;
 /**
  * Base interface for server messages
  */
@@ -396,10 +514,23 @@ interface ServerTurnCompleteMessage extends BaseServerMessage {
 interface ServerReadyMessage extends BaseServerMessage {
     type: 'ready';
 }
+/**
+ * Tool call message from server (function calling)
+ * Indicates the AI wants to call a function defined by the SDK user
+ */
+interface ServerToolCallMessage extends BaseServerMessage {
+    type: 'toolCall';
+    /** Unique identifier for this tool call */
+    id: string;
+    /** Name of the function to call */
+    name: string;
+    /** Arguments passed to the function */
+    args: Record<string, unknown>;
+}
 /**
  * Union type of all server messages
  */
-type ServerMessage = ServerSessionStartedMessage | ServerSessionEndedMessage | ServerReadyMessage | ServerUserTranscriptMessage | ServerResponseMessage | ServerAudioMessage | ServerTurnCompleteMessage | ServerErrorMessage | ServerPongMessage;
+type ServerMessage = ServerSessionStartedMessage | ServerSessionEndedMessage | ServerReadyMessage | ServerUserTranscriptMessage | ServerResponseMessage | ServerAudioMessage | ServerTurnCompleteMessage | ServerToolCallMessage | ServerErrorMessage | ServerPongMessage;
 
 /**
  * Connection state
@@ -420,6 +551,7 @@ type LiveSpeechEventMap = {
     response: ResponseEvent;
     audio: AudioEvent;
     turnComplete: TurnCompleteEvent;
+    toolCall: ToolCallEvent;
     error: ErrorEvent;
 };
 /**
@@ -515,6 +647,28 @@ declare class LiveSpeechClient {
         text: string;
         triggerResponse?: boolean;
     }): void;
+    /**
+     * Send a tool response back to the AI after executing a function
+     *
+     * When you receive a `toolCall` event, you must execute the function and send
+     * the result back using this method. The AI will use the result to continue
+     * the conversation.
+     *
+     * @param id - The tool call ID from the `toolCall` event
+     * @param response - The result of the function execution (will be JSON serialized)
+     *
+     * @example
+     * client.on('toolCall', async (event) => {
+     *   if (event.name === 'get_price') {
+     *     const price = await lookupPrice(event.args.productId);
+     *     client.sendToolResponse(event.id, { price, currency: 'USD' });
+     *   } else if (event.name === 'open_login') {
+     *     showLoginModal();
+     *     client.sendToolResponse(event.id, { success: true });
+     *   }
+     * });
+     */
+    sendToolResponse(id: string, response?: unknown): void;
     /**
      * Add event listener
      */
@@ -636,4 +790,4 @@ declare class AudioEncoder {
     wrapWav(data: Uint8Array): Uint8Array;
 }
 
-export { AudioEncoder, type AudioEncoderOptions, type AudioEvent, type AudioHandler, type ClientMessage, type ClientMessageType, type ConnectedEvent, type ConnectionState, type DisconnectReason, type DisconnectedEvent, type ErrorCode, type ErrorEvent, type ErrorHandler, LiveSpeechClient, type LiveSpeechConfig, type LiveSpeechEvent, type LiveSpeechEventMap, type LiveSpeechEventType, type PipelineMode, type ReadyEvent, type ReconnectingEvent, Region, Region as RegionType, type ResolvedConfig, type ResponseEvent, type ResponseHandler, type ServerMessage, type ServerMessageType, type SessionConfig, type SessionEndedEvent, type SessionStartedEvent, type TurnCompleteEvent, type UserTranscriptEvent, type UserTranscriptHandler, createWavHeader, decodeBase64ToAudio, encodeAudioToBase64, extractPcmFromWav, float32ToInt16, getEndpointForRegion, int16ToFloat32, int16ToUint8, isValidRegion, uint8ToInt16, wrapPcmInWav };
+export { AudioEncoder, type AudioEncoderOptions, type AudioEvent, type AudioHandler, type ClientMessage, type ClientMessageType, type ConnectedEvent, type ConnectionState, type DisconnectReason, type DisconnectedEvent, type ErrorCode, type ErrorEvent, type ErrorHandler, type FunctionParameters, LiveSpeechClient, type LiveSpeechConfig, type LiveSpeechEvent, type LiveSpeechEventMap, type LiveSpeechEventType, type PipelineMode, type ReadyEvent, type ReconnectingEvent, Region, Region as RegionType, type ResolvedConfig, type ResponseEvent, type ResponseHandler, type ServerMessage, type ServerMessageType, type SessionConfig, type SessionEndedEvent, type SessionStartedEvent, type Tool, type ToolCallEvent, type TurnCompleteEvent, type UserTranscriptEvent, type UserTranscriptHandler, createWavHeader, decodeBase64ToAudio, encodeAudioToBase64, extractPcmFromWav, float32ToInt16, getEndpointForRegion, int16ToFloat32, int16ToUint8, isValidRegion, uint8ToInt16, wrapPcmInWav };

package/dist/index.d.ts

@@ -82,6 +82,58 @@ interface LiveSpeechConfig {
      */
     debug?: boolean;
 }
+/**
+ * Parameters schema for a function (for function calling)
+ */
+interface FunctionParameters {
+    /**
+     * Type of the parameters object
+     * @default "OBJECT"
+     */
+    type: 'OBJECT';
+    /**
+     * Properties of the parameters object
+     * Each key is a parameter name, and the value describes its type and description
+     * @example { "productId": { "type": "string", "description": "The product ID to look up" } }
+     */
+    properties: Record<string, unknown>;
+    /**
+     * List of required parameter names
+     */
+    required?: string[];
+}
+/**
+ * Tool definition for function calling
+ *
+ * Tools allow the AI to call functions defined by your application during conversation.
+ * When the AI decides to call a function, you'll receive a `toolCall` event.
+ *
+ * @example
+ * const tools: Tool[] = [{
+ *   name: 'open_google_login',
+ *   description: 'Opens the Google Login popup when user wants to sign in',
+ *   parameters: {
+ *     type: 'OBJECT',
+ *     properties: {},
+ *     required: []
+ *   }
+ * }];
+ */
+interface Tool {
+    /**
+     * Name of the function (must be unique within the session)
+     */
+    name: string;
+    /**
+     * Description of what the function does and when the AI should use it
+     */
+    description: string;
+    /**
+     * Parameters schema for the function
+     * If the function takes no parameters, use { type: 'OBJECT', properties: {}, required: [] }
+     */
+    parameters?: FunctionParameters;
+}
 /**
  * Session configuration options
  */
@@ -114,6 +166,23 @@ interface SessionConfig {
      * @default false
      */
     allowHarmCategory?: boolean;
+    /**
+     * Tools (functions) that the AI can call during conversation.
+     * When the AI calls a tool, you'll receive a `toolCall` event.
+     * You must respond with `sendToolResponse()` to continue the conversation.
+     *
+     * @example
+     * tools: [{
+     *   name: 'get_price',
+     *   description: 'Get the current price of a product',
+     *   parameters: {
+     *     type: 'OBJECT',
+     *     properties: { productId: { type: 'string', description: 'Product ID' } },
+     *     required: ['productId']
+     *   }
+     * }]
+     */
+    tools?: Tool[];
 }
 /**
  * Internal resolved configuration with defaults applied
@@ -132,7 +201,7 @@ interface ResolvedConfig {
 /**
  * Event types emitted by the LiveSpeech client
  */
-type LiveSpeechEventType = 'connected' | 'disconnected' | 'reconnecting' | 'sessionStarted' | 'sessionEnded' | 'ready' | 'userTranscript' | 'response' | 'audio' | 'turnComplete' | 'error';
+type LiveSpeechEventType = 'connected' | 'disconnected' | 'reconnecting' | 'sessionStarted' | 'sessionEnded' | 'ready' | 'userTranscript' | 'response' | 'audio' | 'turnComplete' | 'toolCall' | 'error';
 /**
  * Event payload for 'connected' event
  */
@@ -237,10 +306,40 @@ interface TurnCompleteEvent {
     type: 'turnComplete';
     timestamp: string;
 }
+/**
+ * Event payload for 'toolCall' event (function calling)
+ * Indicates the AI wants to call a function defined by your application.
+ * You must respond with `sendToolResponse()` to continue the conversation.
+ *
+ * @example
+ * client.on('toolCall', (event) => {
+ *   if (event.name === 'get_price') {
+ *     const price = await lookupPrice(event.args.productId);
+ *     client.sendToolResponse(event.id, { price });
+ *   }
+ * });
+ */
+interface ToolCallEvent {
+    type: 'toolCall';
+    /**
+     * Unique identifier for this tool call.
+     * Must be passed to `sendToolResponse()` to respond.
+     */
+    id: string;
+    /**
+     * Name of the function the AI wants to call.
+     */
+    name: string;
+    /**
+     * Arguments passed by the AI to the function.
+     */
+    args: Record<string, unknown>;
+    timestamp: string;
+}
 /**
  * Union type of all event payloads
  */
-type LiveSpeechEvent = ConnectedEvent | DisconnectedEvent | ReconnectingEvent | SessionStartedEvent | SessionEndedEvent | ReadyEvent | UserTranscriptEvent | ResponseEvent | AudioEvent | TurnCompleteEvent | ErrorEvent;
+type LiveSpeechEvent = ConnectedEvent | DisconnectedEvent | ReconnectingEvent | SessionStartedEvent | SessionEndedEvent | ReadyEvent | UserTranscriptEvent | ResponseEvent | AudioEvent | TurnCompleteEvent | ToolCallEvent | ErrorEvent;
 /**
  * Simplified event handlers for common use cases
  */
@@ -252,11 +351,11 @@ type ErrorHandler = (error: ErrorEvent) => void;
 /**
  * WebSocket message types sent from client to server
  */
-type ClientMessageType = 'startSession' | 'endSession' | 'audioStart' | 'audioChunk' | 'audioEnd' | 'systemMessage' | 'ping';
+type ClientMessageType = 'startSession' | 'endSession' | 'audioStart' | 'audioChunk' | 'audioEnd' | 'systemMessage' | 'toolResponse' | 'ping';
 /**
  * WebSocket message types received from server
  */
-type ServerMessageType = 'sessionStarted' | 'sessionEnded' | 'ready' | 'userTranscript' | 'response' | 'audio' | 'turnComplete' | 'error' | 'pong';
+type ServerMessageType = 'sessionStarted' | 'sessionEnded' | 'ready' | 'userTranscript' | 'response' | 'audio' | 'turnComplete' | 'toolCall' | 'error' | 'pong';
 /**
  * Base interface for client messages
  */
@@ -271,6 +370,9 @@ interface StartSessionMessage extends BaseClientMessage {
     prePrompt?: string;
     language?: string;
     pipelineMode?: 'live' | 'composed';
+    aiSpeaksFirst?: boolean;
+    allowHarmCategory?: boolean;
+    tools?: Tool[];
 }
 /**
  * End session message
@@ -319,10 +421,26 @@ interface SystemMessageMessage extends BaseClientMessage {
     action: 'systemMessage';
     payload: SystemMessagePayload;
 }
+/**
+ * Tool response payload
+ */
+interface ToolResponsePayload {
+    /** The tool call ID from the toolCall event */
+    id: string;
+    /** The result of the function execution */
+    response: unknown;
+}
+/**
+ * Tool response - send function execution result back to Gemini
+ */
+interface ToolResponseMessage extends BaseClientMessage {
+    action: 'toolResponse';
+    payload: ToolResponsePayload;
+}
 /**
  * Union type of all client messages
  */
-type ClientMessage = StartSessionMessage | EndSessionMessage | AudioStartMessage | AudioChunkMessage | AudioEndMessage | SystemMessageMessage | PingMessage;
+type ClientMessage = StartSessionMessage | EndSessionMessage | AudioStartMessage | AudioChunkMessage | AudioEndMessage | SystemMessageMessage | ToolResponseMessage | PingMessage;
 /**
  * Base interface for server messages
  */
@@ -396,10 +514,23 @@ interface ServerTurnCompleteMessage extends BaseServerMessage {
 interface ServerReadyMessage extends BaseServerMessage {
     type: 'ready';
 }
+/**
+ * Tool call message from server (function calling)
+ * Indicates the AI wants to call a function defined by the SDK user
+ */
+interface ServerToolCallMessage extends BaseServerMessage {
+    type: 'toolCall';
+    /** Unique identifier for this tool call */
+    id: string;
+    /** Name of the function to call */
+    name: string;
+    /** Arguments passed to the function */
+    args: Record<string, unknown>;
+}
 /**
  * Union type of all server messages
  */
-type ServerMessage = ServerSessionStartedMessage | ServerSessionEndedMessage | ServerReadyMessage | ServerUserTranscriptMessage | ServerResponseMessage | ServerAudioMessage | ServerTurnCompleteMessage | ServerErrorMessage | ServerPongMessage;
+type ServerMessage = ServerSessionStartedMessage | ServerSessionEndedMessage | ServerReadyMessage | ServerUserTranscriptMessage | ServerResponseMessage | ServerAudioMessage | ServerTurnCompleteMessage | ServerToolCallMessage | ServerErrorMessage | ServerPongMessage;
 
 /**
  * Connection state
@@ -420,6 +551,7 @@ type LiveSpeechEventMap = {
     response: ResponseEvent;
     audio: AudioEvent;
     turnComplete: TurnCompleteEvent;
+    toolCall: ToolCallEvent;
     error: ErrorEvent;
 };
 /**
@@ -515,6 +647,28 @@ declare class LiveSpeechClient {
         text: string;
         triggerResponse?: boolean;
     }): void;
+    /**
+     * Send a tool response back to the AI after executing a function
+     *
+     * When you receive a `toolCall` event, you must execute the function and send
+     * the result back using this method. The AI will use the result to continue
+     * the conversation.
+     *
+     * @param id - The tool call ID from the `toolCall` event
+     * @param response - The result of the function execution (will be JSON serialized)
+     *
+     * @example
+     * client.on('toolCall', async (event) => {
+     *   if (event.name === 'get_price') {
+     *     const price = await lookupPrice(event.args.productId);
+     *     client.sendToolResponse(event.id, { price, currency: 'USD' });
+     *   } else if (event.name === 'open_login') {
+     *     showLoginModal();
+     *     client.sendToolResponse(event.id, { success: true });
+     *   }
+     * });
+     */
+    sendToolResponse(id: string, response?: unknown): void;
     /**
      * Add event listener
      */
@@ -636,4 +790,4 @@ declare class AudioEncoder {
     wrapWav(data: Uint8Array): Uint8Array;
 }
 
-export { AudioEncoder, type AudioEncoderOptions, type AudioEvent, type AudioHandler, type ClientMessage, type ClientMessageType, type ConnectedEvent, type ConnectionState, type DisconnectReason, type DisconnectedEvent, type ErrorCode, type ErrorEvent, type ErrorHandler, LiveSpeechClient, type LiveSpeechConfig, type LiveSpeechEvent, type LiveSpeechEventMap, type LiveSpeechEventType, type PipelineMode, type ReadyEvent, type ReconnectingEvent, Region, Region as RegionType, type ResolvedConfig, type ResponseEvent, type ResponseHandler, type ServerMessage, type ServerMessageType, type SessionConfig, type SessionEndedEvent, type SessionStartedEvent, type TurnCompleteEvent, type UserTranscriptEvent, type UserTranscriptHandler, createWavHeader, decodeBase64ToAudio, encodeAudioToBase64, extractPcmFromWav, float32ToInt16, getEndpointForRegion, int16ToFloat32, int16ToUint8, isValidRegion, uint8ToInt16, wrapPcmInWav };
+export { AudioEncoder, type AudioEncoderOptions, type AudioEvent, type AudioHandler, type ClientMessage, type ClientMessageType, type ConnectedEvent, type ConnectionState, type DisconnectReason, type DisconnectedEvent, type ErrorCode, type ErrorEvent, type ErrorHandler, type FunctionParameters, LiveSpeechClient, type LiveSpeechConfig, type LiveSpeechEvent, type LiveSpeechEventMap, type LiveSpeechEventType, type PipelineMode, type ReadyEvent, type ReconnectingEvent, Region, Region as RegionType, type ResolvedConfig, type ResponseEvent, type ResponseHandler, type ServerMessage, type ServerMessageType, type SessionConfig, type SessionEndedEvent, type SessionStartedEvent, type Tool, type ToolCallEvent, type TurnCompleteEvent, type UserTranscriptEvent, type UserTranscriptHandler, createWavHeader, decodeBase64ToAudio, encodeAudioToBase64, extractPcmFromWav, float32ToInt16, getEndpointForRegion, int16ToFloat32, int16ToUint8, isValidRegion, uint8ToInt16, wrapPcmInWav };

package/dist/index.js

@@ -45,7 +45,7 @@ var Region = {
   US_WEST_2: "us-west-2"
 };
 var REGION_ENDPOINTS = {
-  "ap-northeast-2": "wss://talk.drawdream.co.kr",
+  "ap-northeast-2": "ws://localhost:9090",
   "us-west-2": "wss://talk.drawdream.ca"
   // Coming soon
 };
@@ -728,21 +728,15 @@ var LiveSpeechClient = class {
       };
       this.on("sessionStarted", onSessionStarted);
       this.on("error", onError);
-      const startMessage = {
-        action: "startSession"
-      };
-      if (config?.prePrompt) {
-        startMessage.prePrompt = config.prePrompt;
-      }
-      if (config?.language) {
-        startMessage.language = config.language;
-      }
-      startMessage.pipelineMode = config?.pipelineMode ?? "live";
-      if (config?.aiSpeaksFirst) {
-        startMessage.aiSpeaksFirst = config.aiSpeaksFirst;
-      }
-      startMessage.allowHarmCategory = config?.allowHarmCategory ?? false;
-      this.connection.send(startMessage);
+      this.connection.send({
+        action: "startSession",
+        ...config?.prePrompt && { prePrompt: config.prePrompt },
+        ...config?.language && { language: config.language },
+        pipelineMode: config?.pipelineMode ?? "live",
+        ...config?.aiSpeaksFirst && { aiSpeaksFirst: config.aiSpeaksFirst },
+        allowHarmCategory: config?.allowHarmCategory ?? false,
+        ...config?.tools && config.tools.length > 0 && { tools: config.tools }
+      });
     });
   }
   /**
@@ -849,6 +843,40 @@ var LiveSpeechClient = class {
       payload
     });
   }
+  /**
+   * Send a tool response back to the AI after executing a function
+   * 
+   * When you receive a `toolCall` event, you must execute the function and send
+   * the result back using this method. The AI will use the result to continue
+   * the conversation.
+   * 
+   * @param id - The tool call ID from the `toolCall` event
+   * @param response - The result of the function execution (will be JSON serialized)
+   * 
+   * @example
+   * client.on('toolCall', async (event) => {
+   *   if (event.name === 'get_price') {
+   *     const price = await lookupPrice(event.args.productId);
+   *     client.sendToolResponse(event.id, { price, currency: 'USD' });
+   *   } else if (event.name === 'open_login') {
+   *     showLoginModal();
+   *     client.sendToolResponse(event.id, { success: true });
+   *   }
+   * });
+   */
+  sendToolResponse(id, response = {}) {
+    if (!this.isConnected) {
+      throw new Error("Not connected");
+    }
+    if (!this.isStreaming) {
+      throw new Error("No active Live session. Call audioStart() first.");
+    }
+    this.logger.info("Sending tool response:", id);
+    this.connection.send({
+      action: "toolResponse",
+      payload: { id, response }
+    });
+  }
   // ==================== Event System ====================
   /**
    * Add event listener
@@ -1014,6 +1042,18 @@ var LiveSpeechClient = class {
         this.emit("turnComplete", turnCompleteEvent);
         break;
       }
+      case "toolCall": {
+        const toolCallEvent = {
+          type: "toolCall",
+          id: message.id,
+          name: message.name,
+          args: message.args ?? {},
+          timestamp: message.timestamp
+        };
+        this.logger.info("Tool call received:", toolCallEvent.name);
+        this.emit("toolCall", toolCallEvent);
+        break;
+      }
       case "error":
         this.handleError(message.code, message.message);
         break;

package/dist/index.mjs

@@ -6,7 +6,7 @@ var Region = {
   US_WEST_2: "us-west-2"
 };
 var REGION_ENDPOINTS = {
-  "ap-northeast-2": "wss://talk.drawdream.co.kr",
+  "ap-northeast-2": "ws://localhost:9090",
   "us-west-2": "wss://talk.drawdream.ca"
   // Coming soon
 };
@@ -689,21 +689,15 @@ var LiveSpeechClient = class {
       };
       this.on("sessionStarted", onSessionStarted);
       this.on("error", onError);
-      const startMessage = {
-        action: "startSession"
-      };
-      if (config?.prePrompt) {
-        startMessage.prePrompt = config.prePrompt;
-      }
-      if (config?.language) {
-        startMessage.language = config.language;
-      }
-      startMessage.pipelineMode = config?.pipelineMode ?? "live";
-      if (config?.aiSpeaksFirst) {
-        startMessage.aiSpeaksFirst = config.aiSpeaksFirst;
-      }
-      startMessage.allowHarmCategory = config?.allowHarmCategory ?? false;
-      this.connection.send(startMessage);
+      this.connection.send({
+        action: "startSession",
+        ...config?.prePrompt && { prePrompt: config.prePrompt },
+        ...config?.language && { language: config.language },
+        pipelineMode: config?.pipelineMode ?? "live",
+        ...config?.aiSpeaksFirst && { aiSpeaksFirst: config.aiSpeaksFirst },
+        allowHarmCategory: config?.allowHarmCategory ?? false,
+        ...config?.tools && config.tools.length > 0 && { tools: config.tools }
+      });
     });
   }
   /**
@@ -810,6 +804,40 @@ var LiveSpeechClient = class {
       payload
     });
   }
+  /**
+   * Send a tool response back to the AI after executing a function
+   * 
+   * When you receive a `toolCall` event, you must execute the function and send
+   * the result back using this method. The AI will use the result to continue
+   * the conversation.
+   * 
+   * @param id - The tool call ID from the `toolCall` event
+   * @param response - The result of the function execution (will be JSON serialized)
+   * 
+   * @example
+   * client.on('toolCall', async (event) => {
+   *   if (event.name === 'get_price') {
+   *     const price = await lookupPrice(event.args.productId);
+   *     client.sendToolResponse(event.id, { price, currency: 'USD' });
+   *   } else if (event.name === 'open_login') {
+   *     showLoginModal();
+   *     client.sendToolResponse(event.id, { success: true });
+   *   }
+   * });
+   */
+  sendToolResponse(id, response = {}) {
+    if (!this.isConnected) {
+      throw new Error("Not connected");
+    }
+    if (!this.isStreaming) {
+      throw new Error("No active Live session. Call audioStart() first.");
+    }
+    this.logger.info("Sending tool response:", id);
+    this.connection.send({
+      action: "toolResponse",
+      payload: { id, response }
+    });
+  }
   // ==================== Event System ====================
   /**
    * Add event listener
@@ -975,6 +1003,18 @@ var LiveSpeechClient = class {
         this.emit("turnComplete", turnCompleteEvent);
         break;
       }
+      case "toolCall": {
+        const toolCallEvent = {
+          type: "toolCall",
+          id: message.id,
+          name: message.name,
+          args: message.args ?? {},
+          timestamp: message.timestamp
+        };
+        this.logger.info("Tool call received:", toolCallEvent.name);
+        this.emit("toolCall", toolCallEvent);
+        break;
+      }
       case "error":
         this.handleError(message.code, message.message);
         break;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@drawdream/livespeech",
-  "version": "0.1.7",
+  "version": "0.1.8",
   "description": "Real-time speech-to-speech AI conversation SDK",
   "main": "dist/index.js",
   "module": "dist/index.mjs",