@mastra/voice-openai-realtime 0.0.5-alpha.0 → 0.1.0-alpha.1

package/dist/index.js

@@ -1,7 +1,8 @@
 import { MastraVoice } from '@mastra/core/voice';
-import { RealtimeClient } from 'openai-realtime-api';
-import { Readable } from 'stream';
+import { PassThrough, Readable } from 'stream';
 import { zodToJsonSchema } from 'zod-to-json-schema';
+import { WebSocket } from 'ws';
+import { EventEmitter } from 'events';
 
 // src/index.ts
 var transformTools = (tools) => {
@@ -27,6 +28,7 @@ var transformTools = (tools) => {
       continue;
     }
     const openaiTool = {
+      type: "function",
       name,
       description: tool.description || `Tool: ${name}`,
       parameters
@@ -61,22 +63,18 @@ var transformTools = (tools) => {
 var isReadableStream = (obj) => {
   return obj && obj instanceof Readable && typeof obj.read === "function" && typeof obj.pipe === "function" && obj.readable === true;
 };
-
-// src/index.ts
 var DEFAULT_VOICE = "alloy";
+var DEFAULT_URL = "wss://api.openai.com/v1/realtime";
 var DEFAULT_MODEL = "gpt-4o-mini-realtime-preview-2024-12-17";
-var DEFAULT_VAD_CONFIG = {
-  type: "server_vad",
-  threshold: 0.5,
-  prefix_padding_ms: 1e3,
-  silence_duration_ms: 1e3
-};
 var VOICES = ["alloy", "ash", "ballad", "coral", "echo", "sage", "shimmer", "verse"];
 var OpenAIRealtimeVoice = class extends MastraVoice {
-  client;
+  ws;
   state;
+  client;
   events;
+  instructions;
   tools;
+  debug;
   /**
    * Creates a new instance of OpenAIRealtimeVoice.
    *
@@ -85,13 +83,8 @@ var OpenAIRealtimeVoice = class extends MastraVoice {
    * @param options.chatModel.model - The model ID to use (defaults to GPT-4 Mini Realtime)
    * @param options.chatModel.apiKey - OpenAI API key. Falls back to process.env.OPENAI_API_KEY
    * @param options.chatModel.tools - Tools configuration for the model
-   * @param options.chatModel.options - Additional options for the realtime client
-   * @param options.chatModel.options.sessionConfig - Session configuration overrides
-   * @param options.chatModel.options.url - Custom WebSocket URL
-   * @param options.chatModel.options.dangerouslyAllowAPIKeyInBrowser - Whether to allow API key in browser
-   * @param options.chatModel.options.debug - Enable debug logging
-   * @param options.chatModel.options.tools - Additional tools configuration
    * @param options.speaker - Voice ID to use (defaults to 'alloy')
+   * @param options.debug - Enable debug mode
    *
    * @example
    * ```typescript
@@ -106,25 +99,26 @@ var OpenAIRealtimeVoice = class extends MastraVoice {
    */
   constructor({
     chatModel,
-    speaker
+    speaker,
+    debug = false
   } = {}) {
     super();
-    this.client = new RealtimeClient({
-      apiKey: chatModel?.apiKey || process.env.OPENAI_API_KEY,
-      model: chatModel?.model || DEFAULT_MODEL,
-      ...chatModel?.options,
-      sessionConfig: {
-        voice: speaker || DEFAULT_VOICE,
-        turn_detection: DEFAULT_VAD_CONFIG,
-        ...chatModel?.options?.sessionConfig
+    const url = `${chatModel?.url || DEFAULT_URL}?model=${chatModel?.model || DEFAULT_MODEL}`;
+    const apiKey = chatModel?.apiKey || process.env.OPENAI_API_KEY;
+    this.ws = new WebSocket(url, void 0, {
+      headers: {
+        Authorization: "Bearer " + apiKey,
+        "OpenAI-Beta": "realtime=v1"
       }
     });
+    this.client = new EventEmitter();
     this.state = "close";
     this.events = {};
+    this.tools = chatModel?.tools;
+    this.instructions = chatModel?.instructions;
+    this.speaker = speaker || DEFAULT_VOICE;
+    this.debug = debug;
     this.setupEventListeners();
-    if (chatModel?.tools) {
-      this.addTools(chatModel.tools);
-    }
   }
   /**
    * Returns a list of available voice speakers.
@@ -150,8 +144,8 @@ var OpenAIRealtimeVoice = class extends MastraVoice {
    * ```
    */
   close() {
-    if (!this.client) return;
-    this.client.disconnect();
+    if (!this.ws) return;
+    this.ws.close();
     this.state = "close";
   }
   /**
@@ -171,10 +165,10 @@ var OpenAIRealtimeVoice = class extends MastraVoice {
    * ```
    */
   addTools(tools) {
-    const transformedTools = transformTools(tools);
-    for (const tool of transformedTools) {
-      this.client.addTool(tool.openaiTool, tool.execute);
-    }
+    const openaiTools = transformTools(tools);
+    this.updateConfig({
+      tools: openaiTools.map((t) => t.openaiTool)
+    });
   }
   /**
    * Emits a speaking event using the configured voice model.
@@ -210,7 +204,7 @@ var OpenAIRealtimeVoice = class extends MastraVoice {
     if (input.trim().length === 0) {
       throw new Error("Input text is empty");
     }
-    this.client.realtime.send("response.create", {
+    this.sendEvent("response.create", {
       response: {
         instructions: `Repeat the following text: ${input}`,
         voice: options?.speaker ? options.speaker : void 0
@@ -236,7 +230,7 @@ var OpenAIRealtimeVoice = class extends MastraVoice {
    * ```
    */
   updateConfig(sessionConfig) {
-    this.client.updateSession(sessionConfig);
+    this.sendEvent("session.update", { session: sessionConfig });
   }
   /**
    * Processes audio input for speech recognition.
@@ -271,14 +265,14 @@ var OpenAIRealtimeVoice = class extends MastraVoice {
       const buffer = Buffer.concat(chunks);
       const int16Array = new Int16Array(buffer.buffer, buffer.byteOffset ?? 0, (buffer.byteLength ?? 0) / 2);
       const base64Audio = this.int16ArrayToBase64(int16Array);
-      this.client.realtime.send("conversation.item.create", {
+      this.sendEvent("conversation.item.create", {
         item: {
           type: "message",
           role: "user",
           content: [{ type: "input_audio", audio: base64Audio }]
         }
       });
-      this.client.realtime.send("response.create", {
+      this.sendEvent("response.create", {
         response: {
           modalities: ["text"],
           instructions: `ONLY repeat the input and DO NOT say anything else`
@@ -288,6 +282,16 @@ var OpenAIRealtimeVoice = class extends MastraVoice {
       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.
@@ -301,8 +305,17 @@ var OpenAIRealtimeVoice = class extends MastraVoice {
    * ```
    */
   async connect() {
-    await this.client.connect();
-    await this.client.waitForSessionCreated();
+    await this.waitForOpen();
+    await this.waitForSessionCreated();
+    const openaiTools = transformTools(this.tools);
+    this.updateConfig({
+      instructions: this.instructions,
+      tools: openaiTools.map((t) => t.openaiTool),
+      input_audio_transcription: {
+        model: "whisper-1"
+      },
+      voice: this.speaker
+    });
     this.state = "open";
   }
   /**
@@ -323,7 +336,7 @@ var OpenAIRealtimeVoice = class extends MastraVoice {
    * await voice.relay(micStream);
    * ```
    */
-  async send(audioData) {
+  async send(audioData, eventId) {
     if (!this.state || this.state !== "open") {
       console.warn("Cannot relay audio when not open. Call open() first.");
       return;
@@ -333,15 +346,14 @@ var OpenAIRealtimeVoice = class extends MastraVoice {
       stream.on("data", (chunk) => {
         try {
           const buffer = Buffer.isBuffer(chunk) ? chunk : Buffer.from(chunk);
-          const int16Array = new Int16Array(buffer.buffer, buffer.byteOffset, buffer.byteLength / 2);
-          this.client.appendInputAudio(int16Array);
+          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 {
-        this.client.appendInputAudio(audioData);
+        this.sendEvent("input_audio_buffer.append", { audio: audioData, event_id: eventId });
       } catch (err) {
         this.emit("error", err);
       }
@@ -368,7 +380,7 @@ var OpenAIRealtimeVoice = class extends MastraVoice {
    * });
    */
   async answer({ options }) {
-    this.client.realtime.send("response.create", { response: options ?? {} });
+    this.sendEvent("response.create", { response: options ?? {} });
   }
   /**
    * Registers an event listener for voice events.
@@ -437,29 +449,105 @@ var OpenAIRealtimeVoice = class extends MastraVoice {
     }
   }
   setupEventListeners() {
-    this.client.on("error", (error) => {
-      this.emit("error", error);
+    const speakerStreams = /* @__PURE__ */ new Map();
+    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("conversation.created", (conversation) => {
-      this.emit("openAIRealtime:conversation.created", conversation);
+    this.client.on("session.created", (ev) => {
+      this.emit("session.created", ev);
     });
-    this.client.on("conversation.interrupted", () => {
-      this.emit("openAIRealtime:conversation.interrupted");
+    this.client.on("session.updated", (ev) => {
+      this.emit("session.updated", ev);
     });
-    this.client.on("conversation.updated", ({ delta }) => {
-      if (delta?.audio) {
-        this.emit("speaking", { audio: delta.audio });
-      }
+    this.client.on("response.created", (ev) => {
+      this.emit("response.created", ev);
+      const speakerStream = new PassThrough();
+      speakerStream.id = ev.response.id;
+      speakerStreams.set(ev.response.id, speakerStream);
+      this.emit("speaker", speakerStream);
     });
-    this.client.on("conversation.item.appended", (item) => {
-      this.emit("openAIRealtime:conversation.item.appended", item);
+    this.client.on("conversation.item.input_audio_transcription.delta", (ev) => {
+      this.emit("transcribing", { text: ev.delta, response_id: ev.response_id, role: "user" });
     });
-    this.client.on("conversation.item.completed", ({ item, delta }) => {
-      if (item.formatted.transcript) {
-        this.emit("writing", { text: item.formatted.transcript, role: item.role });
-      }
-      this.emit("openAIRealtime:conversation.item.completed", { item, delta });
+    this.client.on("conversation.item.input_audio_transcription.done", (ev) => {
+      this.emit("transcribing", { 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 });
     });
+    this.client.on("response.audio_transcript.done", (ev) => {
+      this.emit("writing", { text: "\n", response_id: ev.response_id });
+    });
+    this.client.on("response.text.delta", (ev) => {
+      this.emit("writing", { text: ev.delta, response_id: ev.response_id });
+    });
+    this.client.on("response.text.done", (ev) => {
+      this.emit("writing", { text: "\n", response_id: ev.response_id });
+    });
+    this.client.on("response.done", (ev) => {
+      this.handleFunctionCalls(ev);
+      this.emit("response.done", ev);
+      speakerStreams.delete(ev.response.id);
+    });
+  }
+  async handleFunctionCalls(ev) {
+    for (const output of ev.response?.output ?? []) {
+      if (output.type === "function_call") {
+        await this.handleFunctionCall(output);
+      }
+    }
+  }
+  async handleFunctionCall(output) {
+    try {
+      const context = JSON.parse(output.arguments);
+      const tool = this.tools?.[output.name];
+      if (!tool) {
+        console.warn(`Tool "${output.name}" not found`);
+        return;
+      }
+      const result = await tool?.execute?.(
+        { context },
+        {
+          toolCallId: "unknown",
+          messages: []
+        }
+      );
+      this.sendEvent("conversation.item.create", {
+        item: {
+          type: "function_call_output",
+          call_id: output.call_id,
+          output: JSON.stringify(result)
+        }
+      });
+    } catch (e) {
+      const err = e;
+      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", {});
+    }
   }
   int16ArrayToBase64(int16Array) {
     const buffer = new ArrayBuffer(int16Array.length * 2);
@@ -474,6 +562,14 @@ var OpenAIRealtimeVoice = class extends MastraVoice {
     }
     return btoa(binary);
   }
+  sendEvent(type, data) {
+    this.ws.send(
+      JSON.stringify({
+        type,
+        ...data
+      })
+    );
+  }
 };
 
 export { OpenAIRealtimeVoice };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mastra/voice-openai-realtime",
-  "version": "0.0.5-alpha.0",
+  "version": "0.1.0-alpha.1",
   "description": "Mastra OpenAI Realtime API integration",
   "type": "module",
   "main": "dist/index.js",
@@ -20,16 +20,18 @@
   },
   "dependencies": {
     "openai-realtime-api": "^1.0.7",
+    "ws": "^8.18.1",
     "zod-to-json-schema": "^3.24.1",
-    "@mastra/core": "^0.6.5-alpha.0"
+    "@mastra/core": "^0.7.0-alpha.1"
   },
   "devDependencies": {
     "@microsoft/api-extractor": "^7.49.2",
     "@types/node": "^22.13.1",
-    "eslint": "^9.20.1",
+    "@types/ws": "^8.18.0",
+    "eslint": "^9.23.0",
     "tsup": "^8.3.6",
     "typescript": "^5.7.3",
-    "vitest": "^2.1.8",
+    "vitest": "^2.1.9",
     "@internal/lint": "0.0.1"
   },
   "scripts": {