@mastra/voice-openai-realtime 0.11.3-alpha.0 → 0.11.3-alpha.2

package/CHANGELOG.md

@@ -1,5 +1,23 @@
 # @mastra/voice-openai-realtime
 
+## 0.11.3-alpha.2
+
+### Patch Changes
+
+- [#7394](https://github.com/mastra-ai/mastra/pull/7394) [`f0dfcac`](https://github.com/mastra-ai/mastra/commit/f0dfcac4458bdf789b975e2d63e984f5d1e7c4d3) Thanks [@NikAiyer](https://github.com/NikAiyer)! - updated core peerdep
+
+- Updated dependencies [[`7149d8d`](https://github.com/mastra-ai/mastra/commit/7149d8d4bdc1edf0008e0ca9b7925eb0b8b60dbe)]:
+  - @mastra/core@0.15.3-alpha.7
+
+## 0.11.3-alpha.1
+
+### Patch Changes
+
+- [#7343](https://github.com/mastra-ai/mastra/pull/7343) [`de3cbc6`](https://github.com/mastra-ai/mastra/commit/de3cbc61079211431bd30487982ea3653517278e) Thanks [@LekoArts](https://github.com/LekoArts)! - Update the `package.json` file to include additional fields like `repository`, `homepage` or `files`.
+
+- Updated dependencies [[`85ef90b`](https://github.com/mastra-ai/mastra/commit/85ef90bb2cd4ae4df855c7ac175f7d392c55c1bf), [`de3cbc6`](https://github.com/mastra-ai/mastra/commit/de3cbc61079211431bd30487982ea3653517278e)]:
+  - @mastra/core@0.15.3-alpha.5
+
 ## 0.11.3-alpha.0
 
 ### Patch Changes

package/package.json

@@ -1,10 +1,14 @@
 {
   "name": "@mastra/voice-openai-realtime",
-  "version": "0.11.3-alpha.0",
+  "version": "0.11.3-alpha.2",
   "description": "Mastra OpenAI Realtime API integration",
   "type": "module",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",
+  "files": [
+    "dist",
+    "CHANGELOG.md"
+  ],
   "exports": {
     ".": {
       "import": {
@@ -33,14 +37,23 @@
     "typescript": "^5.8.3",
     "vitest": "^3.2.4",
     "zod": "^3.25.76",
-    "@internal/types-builder": "0.0.9",
-    "@mastra/core": "0.15.3-alpha.4",
-    "@internal/lint": "0.0.34"
+    "@internal/lint": "0.0.34",
+    "@mastra/core": "0.15.3-alpha.7",
+    "@internal/types-builder": "0.0.9"
   },
   "peerDependencies": {
-    "@mastra/core": ">=0.15.2-0 <0.16.0-0",
+    "@mastra/core": ">=0.15.3-0 <0.16.0-0",
     "zod": "^3.25.0 || ^4.0.0"
   },
+  "homepage": "https://mastra.ai",
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/mastra-ai/mastra.git",
+    "directory": "voice/openai-realtime-api"
+  },
+  "bugs": {
+    "url": "https://github.com/mastra-ai/mastra/issues"
+  },
   "scripts": {
     "build": "tsup --silent --config tsup.config.ts",
     "build:watch": "tsup --watch --silent --config tsup.config.ts",

package/.turbo/turbo-build.log

@@ -1,4 +0,0 @@
-
-> @mastra/voice-openai-realtime@0.11.3-alpha.0 build /home/runner/work/mastra/mastra/voice/openai-realtime-api
-> tsup --silent --config tsup.config.ts
-

package/eslint.config.js

@@ -1,6 +0,0 @@
-import { createConfig } from '@internal/lint/eslint';
-
-const config = await createConfig();
-
-/** @type {import("eslint").Linter.Config[]} */
-export default [...config];

package/src/index.test.ts

@@ -1,109 +0,0 @@
-import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
-import { OpenAIRealtimeVoice } from './index';
-
-// Mock RealtimeClient
-vi.mock('openai-realtime-api', () => {
-  return {
-    RealtimeClient: vi.fn().mockImplementation(() => ({
-      connect: vi.fn().mockResolvedValue(undefined),
-      disconnect: vi.fn(),
-      waitForSessionCreated: vi.fn().mockResolvedValue(undefined),
-      updateSession: vi.fn(),
-      appendInputAudio: vi.fn(),
-      on: vi.fn(),
-      emit: vi.fn(),
-    })),
-  };
-});
-
-vi.mock('ws', () => {
-  return {
-    WebSocket: vi.fn().mockImplementation(() => ({
-      send: vi.fn(),
-      close: vi.fn(),
-      on: vi.fn(),
-    })),
-  };
-});
-
-describe('OpenAIRealtimeVoice', () => {
-  let voice: OpenAIRealtimeVoice;
-
-  beforeEach(() => {
-    vi.clearAllMocks();
-    voice = new OpenAIRealtimeVoice({
-      apiKey: 'test-api-key',
-    });
-    voice.waitForOpen = () => Promise.resolve();
-    voice.waitForSessionCreated = () => Promise.resolve();
-  });
-
-  afterEach(() => {
-    voice?.disconnect();
-  });
-
-  describe('initialization', () => {
-    it('should initialize with default values', () => {
-      expect(voice).toBeInstanceOf(OpenAIRealtimeVoice);
-    });
-
-    it('should initialize with custom speaker', () => {
-      const customVoice = new OpenAIRealtimeVoice({
-        speaker: 'shimmer',
-      });
-      expect(customVoice).toBeInstanceOf(OpenAIRealtimeVoice);
-    });
-  });
-
-  describe('getSpeakers', () => {
-    it('should return array of available voices', async () => {
-      const speakers = await voice.getSpeakers();
-      expect(Array.isArray(speakers)).toBe(true);
-      expect(speakers.length).toBeGreaterThan(0);
-      expect(speakers[0]).toHaveProperty('voiceId');
-    });
-  });
-
-  describe('speak', () => {
-    it('should handle string input', async () => {
-      const testText = 'Hello, world!';
-      await voice.speak(testText);
-    });
-
-    it('should throw error on empty input', async () => {
-      await expect(voice.speak('')).rejects.toThrow('Input text is empty');
-    });
-  });
-
-  describe('send', () => {
-    it('should handle Int16Array input', async () => {
-      const testArray = new Int16Array([1, 2, 3]);
-
-      await voice.connect();
-      voice.send(testArray);
-    });
-  });
-
-  describe('event handling', () => {
-    it('should register and trigger event listeners', () => {
-      const mockCallback = vi.fn();
-      voice.on('speak', mockCallback);
-
-      // Simulate event emission
-      (voice as any).emit('speak', 'test');
-
-      expect(mockCallback).toHaveBeenCalledWith('test');
-    });
-
-    it('should remove event listeners', () => {
-      const mockCallback = vi.fn();
-      voice.on('speak', mockCallback);
-      voice.off('speak', mockCallback);
-
-      // Simulate event emission
-      (voice as any).emit('speak', 'test');
-
-      expect(mockCallback).not.toHaveBeenCalled();
-    });
-  });
-});

package/src/index.ts

@@ -1,716 +0,0 @@
-import { EventEmitter } from 'events';
-import { PassThrough } from 'stream';
-import type { ToolsInput } from '@mastra/core/agent';
-import type { RuntimeContext } from '@mastra/core/runtime-context';
-import { MastraVoice } from '@mastra/core/voice';
-import type { Realtime, RealtimeServerEvents } from 'openai-realtime-api';
-import { WebSocket } from 'ws';
-import { isReadableStream, transformTools } from './utils';
-
-/**
- * Event callback function type
- */
-type EventCallback = (...args: any[]) => void;
-
-type StreamWithId = PassThrough & { id: string };
-
-/**
- * Map of event types to their callback arrays
- */
-type EventMap = {
-  transcribing: [{ text: string }];
-  writing: [{ text: string }];
-  speaking: [{ audio: string }];
-  speaker: [StreamWithId];
-  error: [Error];
-} & {
-  [key: string]: EventCallback[];
-};
-
-/** Default voice for text-to-speech. 'alloy' provides a neutral, balanced voice suitable for most use cases */
-const DEFAULT_VOICE: Realtime.Voice = 'alloy';
-
-const DEFAULT_TRANSCRIBER: Realtime.AudioTranscriptionModel = 'whisper-1';
-
-const DEFAULT_URL = 'wss://api.openai.com/v1/realtime';
-
-/**
- * Default model for real-time voice interactions.
- * This model is optimized for low-latency responses while maintaining high quality output.
- */
-const DEFAULT_MODEL = 'gpt-4o-mini-realtime-preview-2024-12-17';
-
-// /**
-//  * Default Voice Activity Detection (VAD) configuration.
-//  * These settings control how the system detects speech segments.
-//  *
-//  * @property {string} type - Uses server-side VAD for better accuracy
-//  * @property {number} threshold - Speech detection sensitivity (0.5 = balanced)
-//  * @property {number} prefix_padding_ms - Includes 1 second of audio before speech
-//  * @property {number} silence_duration_ms - Waits 1 second of silence before ending turn
-//  */
-// const DEFAULT_VAD_CONFIG = {
-//   type: 'server_vad',
-//   threshold: 0.5,
-//   prefix_padding_ms: 1000,
-//   silence_duration_ms: 1000,
-// } as Realtime.TurnDetection;
-
-type TTools = ToolsInput;
-
-/**
- * Available voice options for text-to-speech.
- * Each voice has unique characteristics suitable for different use cases:
- * - alloy: Neutral and balanced
- * - echo: Warm and natural
- * - shimmer: Clear and expressive
- * - And more...
- */
-const VOICES = ['alloy', 'ash', 'ballad', 'coral', 'echo', 'sage', 'shimmer', 'verse'];
-
-type RealtimeClientServerEventMap = {
-  [K in RealtimeServerEvents.EventType]: [RealtimeServerEvents.EventMap[K]];
-} & {
-  ['conversation.item.input_audio_transcription.delta']: [{ delta: string; response_id: string }];
-  ['conversation.item.input_audio_transcription.done']: [{ response_id: string }];
-};
-
-/**
- * OpenAIRealtimeVoice provides real-time voice interaction capabilities using OpenAI's
- * WebSocket-based API. It supports:
- * - Real-time text-to-speech
- * - Speech-to-text (transcription)
- * - Voice activity detection
- * - Multiple voice options
- * - Event-based audio streaming
- *
- * The class manages WebSocket connections, audio streaming, and event handling
- * for seamless voice interactions.
- *
- * @extends MastraVoice
- *
- * @example
- * ```typescript
- * const voice = new OpenAIRealtimeVoice({
- *   apiKey: process.env.OPENAI_API_KEY,
- *   model: 'gpt-4o-mini-realtime'
- * });
- *
- * await voice.open();
- * voice.on('speaking', (audioData) => {
- *   // Handle audio data
- * });
- *
- * await voice.speak('Hello, how can I help you today?');
- * ```
- */
-export class OpenAIRealtimeVoice extends MastraVoice {
-  private ws?: WebSocket;
-  private state: 'close' | 'open';
-  private client: EventEmitter<RealtimeClientServerEventMap>;
-  private events: EventMap;
-  private instructions?: string;
-  private tools?: TTools;
-  private debug: boolean;
-  private queue: unknown[] = [];
-  private transcriber: Realtime.AudioTranscriptionModel;
-  private runtimeContext?: RuntimeContext;
-  /**
-   * Creates a new instance of OpenAIRealtimeVoice.
-   *
-   * @param options - Configuration options for the voice instance
-   * @param options.url - The base URL for the OpenAI Realtime API
-   * @param options.model - The model ID to use (defaults to GPT-4 Mini Realtime)
-   * @param options.apiKey - OpenAI API key. Falls back to process.env.OPENAI_API_KEY
-   * @param options.speaker - Voice ID to use (defaults to 'alloy')
-   * @param options.debug - Enable debug mode
-   *
-   * @example
-   * ```typescript
-   * const voice = new OpenAIRealtimeVoice({
-   *   apiKey: 'your-api-key',
-   *   model: 'gpt-4o-mini-realtime',
-   *   speaker: 'alloy'
-   * });
-   * ```
-   */
-  constructor(
-    private options: {
-      model?: string;
-      url?: string;
-      apiKey?: string;
-      speaker?: Realtime.Voice;
-      transcriber?: Realtime.AudioTranscriptionModel;
-      debug?: boolean;
-    } = {},
-  ) {
-    super();
-
-    this.client = new EventEmitter();
-    this.state = 'close';
-    this.events = {} as EventMap;
-    this.speaker = options.speaker || DEFAULT_VOICE;
-    this.transcriber = options.transcriber || DEFAULT_TRANSCRIBER;
-    this.debug = options.debug || false;
-  }
-
-  /**
-   * Returns a list of available voice speakers.
-   *
-   * @returns Promise resolving to an array of voice objects, each containing at least a voiceId
-   *
-   * @example
-   * ```typescript
-   * const speakers = await voice.getSpeakers();
-   * // speakers = [{ voiceId: 'alloy' }, { voiceId: 'echo' }, ...]
-   * ```
-   */
-  getSpeakers(): Promise<Array<{ voiceId: string; [key: string]: any }>> {
-    return Promise.resolve(VOICES.map(v => ({ voiceId: v })));
-  }
-
-  /**
-   * Disconnects from the OpenAI realtime session and cleans up resources.
-   * Should be called when you're done with the voice instance.
-   *
-   * @example
-   * ```typescript
-   * voice.close(); // Disconnects and cleans up
-   * ```
-   */
-  close() {
-    if (!this.ws) return;
-    this.ws.close();
-    this.state = 'close';
-  }
-
-  /**
-   * Equips the voice instance with a set of instructions.
-   * Instructions allow the model to perform additional actions during conversations.
-   *
-   * @param instructions - Optional instructions to addInstructions
-   * @returns Transformed instructions ready for use with the model
-   *
-   * @example
-   * ```typescript
-   * voice.addInstructions('You are a helpful assistant.');
-   * ```
-   */
-  addInstructions(instructions?: string) {
-    this.instructions = instructions;
-  }
-
-  /**
-   * Equips the voice instance with a set of tools.
-   * Tools allow the model to perform additional actions during conversations.
-   *
-   * @param tools - Optional tools configuration to addTools
-   * @returns Transformed tools configuration ready for use with the model
-   *
-   * @example
-   * ```typescript
-   * const tools = {
-   *   search: async (query: string) => { ... },
-   *   calculate: (expression: string) => { ... }
-   * };
-   * voice.addTools(tools);
-   * ```
-   */
-  addTools(tools?: TTools) {
-    this.tools = tools || {};
-  }
-
-  /**
-   * Emits a speaking event using the configured voice model.
-   * Can accept either a string or a readable stream as input.
-   *
-   * @param input - The text to convert to speech, or a readable stream containing the text
-   * @param options - Optional configuration for this specific speech request
-   * @param options.speaker - Override the voice to use for this specific request
-   *
-   * @throws {Error} If the input text is empty
-   *
-   * @example
-   * ```typescript
-   * // Simple text to speech
-   * await voice.speak('Hello world');
-   *
-   * // With custom voice
-   * await voice.speak('Hello world', { speaker: 'echo' });
-   *
-   * // Using a stream
-   * const stream = fs.createReadStream('text.txt');
-   * await voice.speak(stream);
-   * ```
-   */
-  async speak(input: string | NodeJS.ReadableStream, options?: { speaker?: Realtime.Voice }): Promise<void> {
-    if (typeof input !== 'string') {
-      const chunks: Buffer[] = [];
-      for await (const chunk of input) {
-        chunks.push(Buffer.isBuffer(chunk) ? chunk : Buffer.from(String(chunk)));
-      }
-      input = Buffer.concat(chunks).toString('utf-8');
-    }
-
-    if (input.trim().length === 0) {
-      throw new Error('Input text is empty');
-    }
-
-    this.sendEvent('response.create', {
-      response: {
-        instructions: `Repeat the following text: ${input}`,
-        voice: options?.speaker ? options.speaker : undefined,
-      },
-    });
-  }
-
-  /**
-   * Updates the session configuration for the voice instance.
-   * This can be used to modify voice settings, turn detection, and other parameters.
-   *
-   * @param sessionConfig - New session configuration to apply
-   *
-   * @example
-   * ```typescript
-   * voice.updateConfig({
-   *   voice: 'echo',
-   *   turn_detection: {
-   *     type: 'server_vad',
-   *     threshold: 0.5,
-   *     silence_duration_ms: 1000
-   *   }
-   * });
-   * ```
-   */
-  updateConfig(sessionConfig: unknown): void {
-    this.sendEvent('session.update', { session: sessionConfig });
-  }
-
-  /**
-   * Checks if listening capabilities are enabled.
-   *
-   * @returns {Promise<{ enabled: boolean }>}
-   */
-  async getListener() {
-    return { enabled: true };
-  }
-
-  /**
-   * Processes audio input for speech recognition.
-   * Takes a readable stream of audio data and emits a writing event.
-   * The output of the writing event is int16 audio data.
-   *
-   * @param audioData - Readable stream containing the audio data to process
-   * @param options - Optional configuration for audio processing
-   *
-   * @throws {Error} If the audio data format is not supported
-   *
-   * @example
-   * ```typescript
-   * // Process audio from a file
-   * const audioStream = fs.createReadStream('audio.raw');
-   * await voice.listen(audioStream);
-   *
-   * // Process audio with options
-   * await voice.listen(microphoneStream, {
-   *   format: 'int16',
-   *   sampleRate: 24000
-   * });
-   * ```
-   */
-  async listen(audioData: NodeJS.ReadableStream): Promise<void> {
-    if (isReadableStream(audioData)) {
-      const chunks: Buffer[] = [];
-      for await (const chunk of audioData) {
-        const buffer = Buffer.isBuffer(chunk) ? chunk : Buffer.from(chunk);
-        chunks.push(buffer);
-      }
-
-      const buffer = Buffer.concat(chunks);
-      const int16Array = new Int16Array(buffer.buffer, buffer.byteOffset ?? 0, (buffer.byteLength ?? 0) / 2);
-      const base64Audio = this.int16ArrayToBase64(int16Array);
-
-      this.sendEvent('conversation.item.create', {
-        item: {
-          type: 'message',
-          role: 'user',
-          content: [{ type: 'input_audio', audio: base64Audio }],
-        },
-      });
-
-      this.sendEvent('response.create', {
-        response: {
-          modalities: ['text'],
-          instructions: `ONLY repeat the input and DO NOT say anything else`,
-        },
-      });
-    } else {
-      this.emit('error', new Error('Unsupported audio data format'));
-    }
-  }
-
-  waitForOpen() {
-    return new Promise(resolve => {
-      this.ws?.on('open', resolve);
-    });
-  }
-
-  waitForSessionCreated() {
-    return new Promise(resolve => {
-      this.client.on('session.created', resolve);
-    });
-  }
-
-  /**
-   * Establishes a connection to the OpenAI realtime service.
-   * Must be called before using speak, listen, or relay functions.
-   *
-   * @throws {Error} If connection fails or session creation times out
-   *
-   * @example
-   * ```typescript
-   * await voice.open();
-   * // Now ready for voice interactions
-   * ```
-   */
-  async connect({ runtimeContext }: { runtimeContext?: RuntimeContext } = {}) {
-    const url = `${this.options.url || DEFAULT_URL}?model=${this.options.model || DEFAULT_MODEL}`;
-    const apiKey = this.options.apiKey || process.env.OPENAI_API_KEY;
-    this.runtimeContext = runtimeContext;
-
-    this.ws = new WebSocket(url, undefined, {
-      headers: {
-        Authorization: 'Bearer ' + apiKey,
-        'OpenAI-Beta': 'realtime=v1',
-      },
-    });
-
-    this.setupEventListeners();
-    await Promise.all([this.waitForOpen(), this.waitForSessionCreated()]);
-
-    const openaiTools = transformTools(this.tools);
-    this.updateConfig({
-      instructions: this.instructions,
-      tools: openaiTools.map(t => t.openaiTool),
-      input_audio_transcription: {
-        model: this.transcriber,
-      },
-      voice: this.speaker,
-    });
-    this.state = 'open';
-  }
-
-  disconnect() {
-    this.state = 'close';
-    this.ws?.close();
-  }
-
-  /**
-   * Streams audio data in real-time to the OpenAI service.
-   * Useful for continuous audio streaming scenarios like live microphone input.
-   * Must be in 'open' state before calling this method.
-   *
-   * @param audioData - Readable stream of audio data to relay
-   * @throws {Error} If audio format is not supported
-   *
-   * @example
-   * ```typescript
-   * // First connect
-   * await voice.open();
-   *
-   * // Then relay audio
-   * const micStream = getMicrophoneStream();
-   * await voice.relay(micStream);
-   * ```
-   */
-  async send(audioData: NodeJS.ReadableStream | Int16Array, eventId?: string): Promise<void> {
-    if (!this.state || this.state !== 'open') {
-      console.warn('Cannot relay audio when not open. Call open() first.');
-      return;
-    }
-
-    if (isReadableStream(audioData)) {
-      const stream = audioData as NodeJS.ReadableStream;
-      stream.on('data', chunk => {
-        try {
-          const buffer = Buffer.isBuffer(chunk) ? chunk : Buffer.from(chunk);
-          this.sendEvent('input_audio_buffer.append', { audio: buffer.toString('base64'), event_id: eventId });
-        } catch (err) {
-          this.emit('error', err);
-        }
-      });
-    } else if (audioData instanceof Int16Array) {
-      try {
-        const base64Audio = this.int16ArrayToBase64(audioData);
-        this.sendEvent('input_audio_buffer.append', { audio: base64Audio, event_id: eventId });
-      } catch (err) {
-        this.emit('error', err);
-      }
-    } else {
-      this.emit('error', new Error('Unsupported audio data format'));
-    }
-  }
-
-  /**
-   * Sends a response to the OpenAI Realtime API.
-   *
-   * Trigger a response to the real-time session.
-   *
-   * @param {Object} params - The parameters object
-   * @param {Realtime.ResponseConfig} params.options - Configuration options for the response
-   * @returns {Promise<void>} A promise that resolves when the response has been sent
-   *
-   * @example
-   * // Send a simple text response
-   * await realtimeVoice.answer({
-   *   options: {
-   *     content: "Hello, how can I help you today?",
-   *     voice: "alloy"
-   *   }
-   * });
-   */
-  async answer({ options }: { options?: Realtime.ResponseConfig }) {
-    this.sendEvent('response.create', { response: options ?? {} });
-  }
-
-  /**
-   * Registers an event listener for voice events.
-   * Available events: 'speaking', 'writing, 'error'
-   * Can listen to OpenAI Realtime events by prefixing with 'openAIRealtime:'
-   * Such as 'openAIRealtime:conversation.item.completed', 'openAIRealtime:conversation.updated', etc.
-   *
-   * @param event - Name of the event to listen for
-   * @param callback - Function to call when the event occurs
-   *
-   * @example
-   * ```typescript
-   * // Listen for speech events
-   * voice.on('speaking', (audioData: Int16Array) => {
-   *   // Handle audio data
-   * });
-   *
-   * // Handle errors
-   * voice.on('error', (error: Error) => {
-   *   console.error('Voice error:', error);
-   * });
-   * ```
-   */
-  on(event: string, callback: EventCallback): void {
-    if (!this.events[event]) {
-      this.events[event] = [];
-    }
-    this.events[event].push(callback);
-  }
-
-  /**
-   * Removes a previously registered event listener.
-   *
-   * @param event - Name of the event to stop listening to
-   * @param callback - The specific callback function to remove
-   *
-   * @example
-   * ```typescript
-   * // Create event handler
-   * const handleSpeech = (audioData: Int16Array) => {
-   *   // Handle audio data
-   * };
-   *
-   * // Add listener
-   * voice.on('speaking', handleSpeech);
-   *
-   * // Later, remove the listener
-   * voice.off('speaking', handleSpeech);
-   * ```
-   */
-  off(event: string, callback: EventCallback): void {
-    if (!this.events[event]) return;
-
-    const index = this.events[event].indexOf(callback);
-    if (index !== -1) {
-      this.events[event].splice(index, 1);
-    }
-  }
-
-  /**
-   * Emit an event with arguments
-   * @param event Event name
-   * @param args Arguments to pass to the callbacks
-   */
-  private emit(event: string, ...args: any[]): void {
-    if (!this.events[event]) return;
-
-    for (const callback of this.events[event]) {
-      callback(...args);
-    }
-  }
-
-  private setupEventListeners(): void {
-    const speakerStreams = new Map<string, StreamWithId>();
-
-    if (!this.ws) {
-      throw new Error('WebSocket not initialized');
-    }
-
-    this.ws.on('message', message => {
-      const data = JSON.parse(message.toString());
-      this.client.emit(data.type, data);
-
-      if (this.debug) {
-        const { delta, ...fields } = data;
-        console.log(data.type, fields, delta?.length < 100 ? delta : '');
-      }
-    });
-
-    this.client.on('session.created', ev => {
-      this.emit('session.created', ev);
-
-      const queue = this.queue.splice(0, this.queue.length);
-      for (const ev of queue) {
-        this.ws?.send(JSON.stringify(ev));
-      }
-    });
-    this.client.on('session.updated', ev => {
-      this.emit('session.updated', ev);
-    });
-    this.client.on('response.created', ev => {
-      this.emit('response.created', ev);
-
-      const speakerStream = new PassThrough() as StreamWithId;
-
-      speakerStream.id = ev.response.id;
-
-      speakerStreams.set(ev.response.id, speakerStream);
-      this.emit('speaker', speakerStream);
-    });
-    this.client.on('conversation.item.input_audio_transcription.delta', ev => {
-      this.emit('writing', { text: ev.delta, response_id: ev.response_id, role: 'user' });
-    });
-    this.client.on('conversation.item.input_audio_transcription.done', ev => {
-      this.emit('writing', { text: '\n', response_id: ev.response_id, role: 'user' });
-    });
-    this.client.on('response.audio.delta', ev => {
-      const audio = Buffer.from(ev.delta, 'base64');
-      this.emit('speaking', { audio, response_id: ev.response_id });
-
-      const stream = speakerStreams.get(ev.response_id);
-      stream?.write(audio);
-    });
-    this.client.on('response.audio.done', ev => {
-      this.emit('speaking.done', { response_id: ev.response_id });
-
-      const stream = speakerStreams.get(ev.response_id);
-      stream?.end();
-    });
-    this.client.on('response.audio_transcript.delta', ev => {
-      this.emit('writing', { text: ev.delta, response_id: ev.response_id, role: 'assistant' });
-    });
-    this.client.on('response.audio_transcript.done', ev => {
-      this.emit('writing', { text: '\n', response_id: ev.response_id, role: 'assistant' });
-    });
-    this.client.on('response.text.delta', ev => {
-      this.emit('writing', { text: ev.delta, response_id: ev.response_id, role: 'assistant' });
-    });
-    this.client.on('response.text.done', ev => {
-      this.emit('writing', { text: '\n', response_id: ev.response_id, role: 'assistant' });
-    });
-    this.client.on('response.done', async ev => {
-      await this.handleFunctionCalls(ev);
-      this.emit('response.done', ev);
-      speakerStreams.delete(ev.response.id);
-    });
-    this.client.on('error', async ev => {
-      this.emit('error', ev);
-    });
-  }
-
-  private async handleFunctionCalls(ev: any) {
-    for (const output of ev.response?.output ?? []) {
-      if (output.type === 'function_call') {
-        await this.handleFunctionCall(output);
-      }
-    }
-  }
-
-  private async handleFunctionCall(output: any) {
-    try {
-      const context = JSON.parse(output.arguments);
-      const tool = this.tools?.[output.name];
-      if (!tool) {
-        console.warn(`Tool "${output.name}" not found`);
-        return;
-      }
-
-      if (tool?.execute) {
-        this.emit('tool-call-start', {
-          toolCallId: output.call_id,
-          toolName: output.name,
-          toolDescription: tool.description,
-          args: context,
-        });
-      }
-
-      const result = await tool?.execute?.(
-        { context, runtimeContext: this.runtimeContext },
-        {
-          toolCallId: output.call_id,
-          messages: [],
-        },
-      );
-
-      this.emit('tool-call-result', {
-        toolCallId: output.call_id,
-        toolName: output.name,
-        toolDescription: tool.description,
-        args: context,
-        result,
-      });
-
-      this.sendEvent('conversation.item.create', {
-        item: {
-          type: 'function_call_output',
-          call_id: output.call_id,
-          output: JSON.stringify(result),
-        },
-      });
-    } catch (e) {
-      const err = e as Error;
-      console.warn(`Error calling tool "${output.name}":`, err.message);
-      this.sendEvent('conversation.item.create', {
-        item: {
-          type: 'function_call_output',
-          call_id: output.call_id,
-          output: JSON.stringify({ error: err.message }),
-        },
-      });
-    } finally {
-      this.sendEvent('response.create', {});
-    }
-  }
-
-  private int16ArrayToBase64(int16Array: Int16Array): string {
-    const buffer = new ArrayBuffer(int16Array.length * 2);
-    const view = new DataView(buffer);
-    for (let i = 0; i < int16Array.length; i++) {
-      view.setInt16(i * 2, int16Array[i]!, true);
-    }
-    const uint8Array = new Uint8Array(buffer);
-    let binary = '';
-    for (let i = 0; i < uint8Array.length; i++) {
-      binary += String.fromCharCode(uint8Array[i]!);
-    }
-    return btoa(binary);
-  }
-
-  private sendEvent(type: string, data: any) {
-    if (!this.ws || this.ws.readyState !== this.ws.OPEN) {
-      this.queue.push({ type: type, ...data });
-    } else {
-      this.ws?.send(
-        JSON.stringify({
-          type: type,
-          ...data,
-        }),
-      );
-    }
-  }
-}

package/src/utils.test.ts

@@ -1,119 +0,0 @@
-import { createTool } from '@mastra/core/tools';
-import { describe, it, expect } from 'vitest';
-import { z } from 'zod';
-import { transformTools } from './utils';
-
-// Vitest provides these globals automatically, but we can import them explicitly for clarity
-
-describe('transformTools', () => {
-  describe('Basic Tool Transformation', () => {
-    it('should transform a tool with Zod inputSchema to OpenAI format', () => {
-      // Create a test tool with Zod schema
-      const tool = createTool({
-        id: 'zodTool',
-        description: 'A tool with Zod schema',
-        inputSchema: z.object({
-          name: z.string(),
-          age: z.number().optional(),
-        }),
-        outputSchema: z.string(),
-        execute: async ({ context }) => {
-          return `Hello, ${context.name}`;
-        },
-      });
-
-      // Transform the tool
-      const transformedTools = transformTools({
-        zodTool: tool,
-      });
-
-      // Assert the transformation results
-      expect(transformedTools).toHaveLength(1);
-      const { openaiTool } = transformedTools[0];
-
-      expect(openaiTool).toMatchObject({
-        type: 'function',
-        name: 'zodTool',
-        description: 'A tool with Zod schema',
-        parameters: expect.objectContaining({
-          type: 'object',
-          properties: expect.objectContaining({
-            name: expect.objectContaining({ type: 'string' }),
-            age: expect.objectContaining({ type: 'number' }),
-          }),
-          required: ['name'],
-        }),
-      });
-    });
-
-    it('should transform a tool with JSON schema parameters to OpenAI format', () => {
-      // Create a test tool with direct JSON schema
-      const tool = {
-        id: 'jsonTool',
-        description: 'A tool with JSON schema',
-        parameters: {
-          type: 'object',
-          properties: {
-            query: { type: 'string' },
-            limit: { type: 'integer' },
-          },
-          required: ['query'],
-        },
-        execute: async (args: { query: string; limit?: number }) => {
-          return `Searched for: ${args.query}`;
-        },
-      };
-
-      // Transform the tool
-      const transformedTools = transformTools({
-        jsonTool: tool,
-      });
-
-      // Assert the transformation results
-      expect(transformedTools).toHaveLength(1);
-      const { openaiTool } = transformedTools[0];
-
-      expect(openaiTool).toMatchObject({
-        type: 'function',
-        name: 'jsonTool',
-        description: 'A tool with JSON schema',
-        parameters: expect.objectContaining({
-          type: 'object',
-          properties: expect.objectContaining({
-            query: expect.objectContaining({ type: 'string' }),
-            limit: expect.objectContaining({ type: 'integer' }),
-          }),
-          required: ['query'],
-        }),
-      });
-    });
-  });
-
-  describe('Tool Execution Tests', () => {
-    it('should create an adapter function for tool execution', async () => {
-      // Create a tool that expects context
-      const tool = createTool({
-        id: 'messageTool',
-        description: 'A tool that processes a message',
-        inputSchema: z.object({
-          message: z.string(),
-        }),
-        outputSchema: z.string(),
-        execute: async ({ context }) => {
-          return `Processed: ${context.message}`;
-        },
-      });
-
-      // Transform the tool
-      const transformedTools = transformTools({
-        messageTool: tool,
-      });
-
-      // Execute the transformed tool
-      const result = await transformedTools[0].execute({ message: 'Hello' });
-
-      // Verify the adapter correctly passes the context
-      expect(result).toBe('Processed: Hello');
-    });
-  });
-});

package/src/utils.ts

@@ -1,106 +0,0 @@
-import { Readable } from 'stream';
-import type { ToolsInput } from '@mastra/core/agent';
-import { zodToJsonSchema } from 'zod-to-json-schema';
-
-export type OpenAIExecuteFunction = (args: any) => Promise<any>;
-type ToolDefinition = {
-  type: 'function';
-  name: string;
-  description: string;
-  parameters: {
-    [key: string]: any;
-  };
-};
-
-type TTools = ToolsInput;
-export const transformTools = (tools?: TTools) => {
-  const openaiTools: { openaiTool: ToolDefinition; execute: OpenAIExecuteFunction }[] = [];
-  for (const [name, tool] of Object.entries(tools || {})) {
-    let parameters: { [key: string]: any };
-
-    if ('inputSchema' in tool && tool.inputSchema) {
-      if (isZodObject(tool.inputSchema)) {
-        parameters = zodToJsonSchema(tool.inputSchema);
-        delete parameters.$schema;
-      } else {
-        parameters = tool.inputSchema;
-      }
-    } else if ('parameters' in tool) {
-      if (isZodObject(tool.parameters)) {
-        parameters = zodToJsonSchema(tool.parameters);
-        delete parameters.$schema;
-      } else {
-        parameters = tool.parameters;
-      }
-    } else {
-      console.warn(`Tool ${name} has neither inputSchema nor parameters, skipping`);
-      continue;
-    }
-    const openaiTool: ToolDefinition = {
-      type: 'function',
-      name,
-      description: tool.description || `Tool: ${name}`,
-      parameters,
-    };
-
-    if (tool.execute) {
-      // Create an adapter function that works with both ToolAction and VercelTool execute functions
-      const executeAdapter = async (args: any) => {
-        try {
-          if (!tool.execute) {
-            throw new Error(`Tool ${name} has no execute function`);
-          }
-
-          // For ToolAction, the first argument is a context object with the args in a 'context' property
-          if ('inputSchema' in tool) {
-            return await tool.execute(
-              { context: args },
-              {
-                toolCallId: 'unknown',
-                messages: [],
-              },
-            );
-          }
-          // For VercelTool, pass args directly
-          else {
-            // Create a minimal ToolExecutionOptions object with required properties
-            const options = {
-              toolCallId: 'unknown',
-              messages: [],
-            };
-            return await tool.execute(args, options);
-          }
-        } catch (error) {
-          console.error(`Error executing tool ${name}:`, error);
-          throw error;
-        }
-      };
-      openaiTools.push({ openaiTool, execute: executeAdapter });
-    } else {
-      console.warn(`Tool ${name} has no execute function, skipping`);
-    }
-  }
-  return openaiTools;
-};
-
-export const isReadableStream = (obj: unknown) => {
-  return (
-    obj &&
-    obj instanceof Readable &&
-    typeof obj.read === 'function' &&
-    typeof obj.pipe === 'function' &&
-    obj.readable === true
-  );
-};
-
-function isZodObject(schema: unknown) {
-  return (
-    !!schema &&
-    typeof schema === 'object' &&
-    '_def' in schema &&
-    schema._def &&
-    typeof schema._def === 'object' &&
-    'typeName' in schema._def &&
-    schema._def.typeName === 'ZodObject'
-  );
-}

package/tsconfig.build.json

@@ -1,9 +0,0 @@
-{
-  "extends": ["./tsconfig.json", "../../tsconfig.build.json"],
-  "compilerOptions": {
-    "outDir": "./dist",
-    "rootDir": "./src"
-  },
-  "include": ["src/**/*"],
-  "exclude": ["node_modules", "**/*.test.ts", "src/**/*.mock.ts"]
-}

package/tsconfig.json

@@ -1,5 +0,0 @@
-{
-  "extends": "../../tsconfig.node.json",
-  "include": ["src/**/*", "tsup.config.ts"],
-  "exclude": ["node_modules", "**/*.test.ts"]
-}

package/tsup.config.ts

@@ -1,17 +0,0 @@
-import { generateTypes } from '@internal/types-builder';
-import { defineConfig } from 'tsup';
-
-export default defineConfig({
-  entry: ['src/index.ts'],
-  format: ['esm', 'cjs'],
-  clean: true,
-  dts: false,
-  splitting: true,
-  treeshake: {
-    preset: 'smallest',
-  },
-  sourcemap: true,
-  onSuccess: async () => {
-    await generateTypes(process.cwd());
-  },
-});

package/vitest.config.ts

@@ -1,8 +0,0 @@
-import { defineConfig } from 'vitest/config';
-
-export default defineConfig({
-  test: {
-    globals: true,
-    include: ['src/**/*.test.ts'],
-  },
-});