@mastra/voice-google-gemini-live 0.0.0-issue-7087-20250910004053 → 0.0.0-jail-fs-20260105160110

package/dist/index.js

@@ -562,7 +562,7 @@ var AudioStreamManager = class {
    */
   log(message, ...args) {
     if (this.debug) {
-      console.log(`[AudioStreamManager] ${message}`, ...args);
+      console.info(`[AudioStreamManager] ${message}`, ...args);
     }
   }
   /**
@@ -841,7 +841,7 @@ var ConnectionManager = class {
    */
   log(message, ...args) {
     if (this.debug) {
-      console.log(`[ConnectionManager] ${message}`, ...args);
+      console.info(`[ConnectionManager] ${message}`, ...args);
     }
   }
 };
@@ -1115,7 +1115,7 @@ var AuthManager = class {
    */
   log(message, ...args) {
     if (this.config.debug) {
-      console.log(`[AuthManager] ${message}`, ...args);
+      console.info(`[AuthManager] ${message}`, ...args);
     }
   }
 };
@@ -1231,7 +1231,7 @@ var EventManager = class {
    */
   log(message, ...args) {
     if (this.debug) {
-      console.log(`[EventManager] ${message}`, ...args);
+      console.info(`[EventManager] ${message}`, ...args);
     }
   }
 };
@@ -1260,7 +1260,7 @@ var GeminiLiveVoice = class _GeminiLiveVoice extends MastraVoice {
   sessionDurationTimeout;
   // Tool integration properties
   tools;
-  runtimeContext;
+  requestContext;
   // Store the configuration options
   options;
   /**
@@ -1497,69 +1497,68 @@ var GeminiLiveVoice = class _GeminiLiveVoice extends MastraVoice {
   /**
    * Establish connection to the Gemini Live API
    */
-  async connect({ runtimeContext } = {}) {
-    return this.traced(async () => {
-      if (this.state === "connected") {
-        this.log("Already connected to Gemini Live API");
-        return;
-      }
-      this.runtimeContext = runtimeContext;
-      this.emit("session", { state: "connecting" });
-      try {
-        let wsUrl;
-        let headers = {};
-        if (this.options.vertexAI) {
-          wsUrl = `wss://${this.options.location}-aiplatform.googleapis.com/ws/google.cloud.aiplatform.v1beta1.PredictionService.ServerStreamingPredict`;
-          await this.authManager.initialize();
-          const accessToken = await this.authManager.getAccessToken();
-          headers = { headers: { Authorization: `Bearer ${accessToken}` } };
-          this.log("Using Vertex AI authentication with OAuth token");
-        } else {
-          wsUrl = `wss://generativelanguage.googleapis.com/ws/google.ai.generativelanguage.v1alpha.GenerativeService.BidiGenerateContent`;
-          headers = {
-            headers: {
-              "x-goog-api-key": this.options.apiKey || "",
-              "Content-Type": "application/json"
-            }
-          };
-          this.log("Using Live API authentication with API key");
-        }
-        this.log("Connecting to:", wsUrl);
-        this.ws = new WebSocket(wsUrl, void 0, headers);
-        this.connectionManager.setWebSocket(this.ws);
-        this.setupEventListeners();
-        await this.connectionManager.waitForOpen();
-        if (this.isResuming && this.sessionHandle) {
-          await this.sendSessionResumption();
-        } else {
-          this.sendInitialConfig();
-          this.sessionStartTime = Date.now();
-          this.sessionId = randomUUID();
-        }
-        await this.waitForSessionCreated();
-        this.state = "connected";
-        this.emit("session", {
-          state: "connected",
-          config: {
-            sessionId: this.sessionId,
-            isResuming: this.isResuming,
-            toolCount: Object.keys(this.tools || {}).length
+  async connect({ requestContext } = {}) {
+    if (this.state === "connected") {
+      this.log("Already connected to Gemini Live API");
+      return;
+    }
+    this.requestContext = requestContext;
+    this.emit("session", { state: "connecting" });
+    try {
+      let wsUrl;
+      let headers = {};
+      if (this.options.vertexAI) {
+        const location = this.getVertexLocation();
+        wsUrl = `wss://${location}-aiplatform.googleapis.com/ws/google.cloud.aiplatform.v1beta1.LlmBidiService/BidiGenerateContent`;
+        await this.authManager.initialize();
+        const accessToken = await this.authManager.getAccessToken();
+        headers = { headers: { Authorization: `Bearer ${accessToken}` } };
+        this.log("Using Vertex AI authentication with OAuth token");
+      } else {
+        wsUrl = `wss://generativelanguage.googleapis.com/ws/google.ai.generativelanguage.v1alpha.GenerativeService.BidiGenerateContent`;
+        headers = {
+          headers: {
+            "x-goog-api-key": this.options.apiKey || "",
+            "Content-Type": "application/json"
           }
-        });
-        this.log("Successfully connected to Gemini Live API", {
+        };
+        this.log("Using Live API authentication with API key");
+      }
+      this.log("Connecting to:", wsUrl);
+      this.ws = new WebSocket(wsUrl, void 0, headers);
+      this.connectionManager.setWebSocket(this.ws);
+      this.setupEventListeners();
+      await this.connectionManager.waitForOpen();
+      if (this.isResuming && this.sessionHandle) {
+        await this.sendSessionResumption();
+      } else {
+        this.sendInitialConfig();
+        this.sessionStartTime = Date.now();
+        this.sessionId = randomUUID();
+      }
+      await this.waitForSessionCreated();
+      this.state = "connected";
+      this.emit("session", {
+        state: "connected",
+        config: {
           sessionId: this.sessionId,
           isResuming: this.isResuming,
           toolCount: Object.keys(this.tools || {}).length
-        });
-        if (this.options.sessionConfig?.maxDuration) {
-          this.startSessionDurationMonitor();
         }
-      } catch (error) {
-        this.state = "disconnected";
-        this.log("Connection failed", error);
-        throw error;
+      });
+      this.log("Successfully connected to Gemini Live API", {
+        sessionId: this.sessionId,
+        isResuming: this.isResuming,
+        toolCount: Object.keys(this.tools || {}).length
+      });
+      if (this.options.sessionConfig?.maxDuration) {
+        this.startSessionDurationMonitor();
       }
-    }, "gemini-live.connect")();
+    } catch (error) {
+      this.state = "disconnected";
+      this.log("Connection failed", error);
+      throw error;
+    }
   }
   /**
    * Disconnect from the Gemini Live API
@@ -1597,172 +1596,164 @@ var GeminiLiveVoice = class _GeminiLiveVoice extends MastraVoice {
    * Send text to be converted to speech
    */
   async speak(input, options) {
-    return this.traced(async () => {
-      this.validateConnectionState();
-      if (typeof input !== "string") {
-        const chunks = [];
-        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 this.createAndEmitError("invalid_audio_format" /* INVALID_AUDIO_FORMAT */, "Input text is empty");
+    this.validateConnectionState();
+    if (typeof input !== "string") {
+      const chunks = [];
+      for await (const chunk of input) {
+        chunks.push(Buffer.isBuffer(chunk) ? chunk : Buffer.from(String(chunk)));
       }
-      this.addToContext("user", input);
-      const textMessage = {
-        client_content: {
-          turns: [
-            {
-              role: "user",
-              parts: [
-                {
-                  text: input
-                }
-              ]
-            }
-          ],
-          turnComplete: true
-        }
-      };
-      if (options && (options.speaker || options.languageCode || options.responseModalities)) {
-        const updateMessage = {
-          type: "session.update",
-          session: {
-            generation_config: {
-              ...options.responseModalities ? { response_modalities: options.responseModalities } : {},
-              speech_config: {
-                ...options.languageCode ? { language_code: options.languageCode } : {},
-                ...options.speaker ? { voice_config: { prebuilt_voice_config: { voice_name: options.speaker } } } : {}
+      input = Buffer.concat(chunks).toString("utf-8");
+    }
+    if (input.trim().length === 0) {
+      throw this.createAndEmitError("invalid_audio_format" /* INVALID_AUDIO_FORMAT */, "Input text is empty");
+    }
+    this.addToContext("user", input);
+    const textMessage = {
+      client_content: {
+        turns: [
+          {
+            role: "user",
+            parts: [
+              {
+                text: input
               }
+            ]
+          }
+        ],
+        turnComplete: true
+      }
+    };
+    if (options && (options.speaker || options.languageCode || options.responseModalities)) {
+      const updateMessage = {
+        type: "session.update",
+        session: {
+          generation_config: {
+            ...options.responseModalities ? { response_modalities: options.responseModalities } : {},
+            speech_config: {
+              ...options.languageCode ? { language_code: options.languageCode } : {},
+              ...options.speaker ? { voice_config: { prebuilt_voice_config: { voice_name: options.speaker } } } : {}
             }
           }
-        };
-        try {
-          this.sendEvent("session.update", updateMessage);
-          this.log("Applied per-turn runtime options", options);
-        } catch (error) {
-          this.log("Failed to apply per-turn runtime options", error);
         }
-      }
+      };
       try {
-        this.sendEvent("client_content", textMessage);
-        this.log("Text message sent", { text: input });
+        this.sendEvent("session.update", updateMessage);
+        this.log("Applied per-turn runtime options", options);
       } catch (error) {
-        this.log("Failed to send text message", error);
-        throw this.createAndEmitError("audio_processing_error" /* AUDIO_PROCESSING_ERROR */, "Failed to send text message", error);
+        this.log("Failed to apply per-turn runtime options", error);
       }
-    }, "gemini-live.speak")();
+    }
+    try {
+      this.sendEvent("client_content", textMessage);
+      this.log("Text message sent", { text: input });
+    } catch (error) {
+      this.log("Failed to send text message", error);
+      throw this.createAndEmitError("audio_processing_error" /* AUDIO_PROCESSING_ERROR */, "Failed to send text message", error);
+    }
   }
   /**
    * Send audio stream for processing
    */
   async send(audioData) {
-    return this.traced(async () => {
-      this.validateConnectionState();
-      if ("readable" in audioData && typeof audioData.on === "function") {
-        const stream = audioData;
-        stream.on("data", (chunk) => {
-          try {
-            const base64Audio = this.audioStreamManager.processAudioChunk(chunk);
-            const message = this.audioStreamManager.createAudioMessage(base64Audio, "realtime");
-            this.sendEvent("realtime_input", message);
-          } catch (error) {
-            this.log("Failed to process audio chunk", error);
-            this.createAndEmitError("audio_processing_error" /* AUDIO_PROCESSING_ERROR */, "Failed to process audio chunk", error);
-          }
-        });
-        stream.on("error", (error) => {
-          this.log("Audio stream error", error);
-          this.createAndEmitError("audio_stream_error" /* AUDIO_STREAM_ERROR */, "Audio stream error", error);
-        });
-        stream.on("end", () => {
-          this.log("Audio stream ended");
-        });
-      } else {
-        const validateAudio = this.audioStreamManager.validateAndConvertAudioInput(audioData);
-        const base64Audio = this.audioStreamManager.int16ArrayToBase64(validateAudio);
-        const message = this.audioStreamManager.createAudioMessage(base64Audio, "realtime");
-        this.sendEvent("realtime_input", message);
-      }
-    }, "gemini-live.send")();
+    this.validateConnectionState();
+    if ("readable" in audioData && typeof audioData.on === "function") {
+      const stream = audioData;
+      stream.on("data", (chunk) => {
+        try {
+          const base64Audio = this.audioStreamManager.processAudioChunk(chunk);
+          const message = this.audioStreamManager.createAudioMessage(base64Audio, "realtime");
+          this.sendEvent("realtime_input", message);
+        } catch (error) {
+          this.log("Failed to process audio chunk", error);
+          this.createAndEmitError("audio_processing_error" /* AUDIO_PROCESSING_ERROR */, "Failed to process audio chunk", error);
+        }
+      });
+      stream.on("error", (error) => {
+        this.log("Audio stream error", error);
+        this.createAndEmitError("audio_stream_error" /* AUDIO_STREAM_ERROR */, "Audio stream error", error);
+      });
+      stream.on("end", () => {
+        this.log("Audio stream ended");
+      });
+    } else {
+      const validateAudio = this.audioStreamManager.validateAndConvertAudioInput(audioData);
+      const base64Audio = this.audioStreamManager.int16ArrayToBase64(validateAudio);
+      const message = this.audioStreamManager.createAudioMessage(base64Audio, "realtime");
+      this.sendEvent("realtime_input", message);
+    }
   }
   /**
    * Process speech from audio stream (traditional STT interface)
    */
   async listen(audioStream, _options) {
-    return this.traced(async () => {
-      this.validateConnectionState();
-      let transcriptionText = "";
-      const onWriting = (data) => {
-        if (data.role === "user") {
-          transcriptionText += data.text;
-          this.log("Received transcription text:", { text: data.text, total: transcriptionText });
-        }
-      };
-      const onError = (error) => {
-        throw new Error(`Transcription failed: ${error.message}`);
-      };
-      const onSession = (data) => {
-        if (data.state === "disconnected") {
-          throw new Error("Session disconnected during transcription");
-        }
-      };
-      this.on("writing", onWriting);
-      this.on("error", onError);
-      this.on("session", onSession);
-      try {
-        const result = await this.audioStreamManager.handleAudioTranscription(
-          audioStream,
-          (base64Audio) => {
-            return new Promise((resolve, reject) => {
-              try {
-                const message = this.audioStreamManager.createAudioMessage(base64Audio, "input");
-                const cleanup = () => {
-                  this.off("turnComplete", onTurnComplete);
-                  this.off("error", onErr);
-                };
-                const onTurnComplete = () => {
-                  cleanup();
-                  resolve(transcriptionText.trim());
-                };
-                const onErr = (e) => {
-                  cleanup();
-                  reject(new Error(e.message));
-                };
-                this.on("turnComplete", onTurnComplete);
-                this.on("error", onErr);
-                this.sendEvent("client_content", message);
-                this.log("Sent audio for transcription");
-              } catch (err) {
-                reject(err);
-              }
-            });
-          },
-          (error) => {
-            this.createAndEmitError("audio_processing_error" /* AUDIO_PROCESSING_ERROR */, "Audio transcription failed", error);
-          }
-        );
-        return result;
-      } finally {
-        this.off("writing", onWriting);
-        this.off("error", onError);
-        this.off("session", onSession);
+    this.validateConnectionState();
+    let transcriptionText = "";
+    const onWriting = (data) => {
+      if (data.role === "user") {
+        transcriptionText += data.text;
+        this.log("Received transcription text:", { text: data.text, total: transcriptionText });
       }
-    }, "gemini-live.listen")();
+    };
+    const onError = (error) => {
+      throw new Error(`Transcription failed: ${error.message}`);
+    };
+    const onSession = (data) => {
+      if (data.state === "disconnected") {
+        throw new Error("Session disconnected during transcription");
+      }
+    };
+    this.on("writing", onWriting);
+    this.on("error", onError);
+    this.on("session", onSession);
+    try {
+      const result = await this.audioStreamManager.handleAudioTranscription(
+        audioStream,
+        (base64Audio) => {
+          return new Promise((resolve, reject) => {
+            try {
+              const message = this.audioStreamManager.createAudioMessage(base64Audio, "input");
+              const cleanup = () => {
+                this.off("turnComplete", onTurnComplete);
+                this.off("error", onErr);
+              };
+              const onTurnComplete = () => {
+                cleanup();
+                resolve(transcriptionText.trim());
+              };
+              const onErr = (e) => {
+                cleanup();
+                reject(new Error(e.message));
+              };
+              this.on("turnComplete", onTurnComplete);
+              this.on("error", onErr);
+              this.sendEvent("client_content", message);
+              this.log("Sent audio for transcription");
+            } catch (err) {
+              reject(err);
+            }
+          });
+        },
+        (error) => {
+          this.createAndEmitError("audio_processing_error" /* AUDIO_PROCESSING_ERROR */, "Audio transcription failed", error);
+        }
+      );
+      return result;
+    } finally {
+      this.off("writing", onWriting);
+      this.off("error", onError);
+      this.off("session", onSession);
+    }
   }
   /**
    * Get available speakers/voices
    */
   async getSpeakers() {
-    return this.traced(async () => {
-      return [
-        { voiceId: "Puck", description: "Conversational, friendly" },
-        { voiceId: "Charon", description: "Deep, authoritative" },
-        { voiceId: "Kore", description: "Neutral, professional" },
-        { voiceId: "Fenrir", description: "Warm, approachable" }
-      ];
-    }, "gemini-live.getSpeakers")();
+    return [
+      { voiceId: "Puck", description: "Conversational, friendly" },
+      { voiceId: "Charon", description: "Deep, authoritative" },
+      { voiceId: "Kore", description: "Neutral, professional" },
+      { voiceId: "Fenrir", description: "Warm, approachable" }
+    ];
   }
   /**
    * Resume a previous session using a session handle
@@ -2267,6 +2258,18 @@ var GeminiLiveVoice = class _GeminiLiveVoice extends MastraVoice {
             role: "assistant"
           });
         }
+        if (part.functionCall) {
+          this.log("Found function call in serverContent.modelTurn.parts", part.functionCall);
+          const toolCallData = {
+            toolCall: {
+              name: part.functionCall.name,
+              args: part.functionCall.args || {},
+              id: part.functionCall.id || randomUUID()
+            }
+          };
+          void this.handleToolCall(toolCallData);
+          continue;
+        }
         if (part.inlineData?.mimeType?.includes("audio") && typeof part.inlineData.data === "string") {
           try {
             const audioData = part.inlineData.data;
@@ -2341,9 +2344,24 @@ var GeminiLiveVoice = class _GeminiLiveVoice extends MastraVoice {
     if (!data.toolCall) {
       return;
     }
-    const toolName = data.toolCall.name || "";
-    const toolArgs = data.toolCall.args || {};
-    const toolId = data.toolCall.id || randomUUID();
+    let toolCalls = [];
+    if (data.toolCall.functionCalls && Array.isArray(data.toolCall.functionCalls)) {
+      toolCalls = data.toolCall.functionCalls;
+    } else if (data.toolCall.name) {
+      toolCalls = [{ name: data.toolCall.name, args: data.toolCall.args, id: data.toolCall.id }];
+    }
+    for (const toolCall of toolCalls) {
+      const toolName = toolCall.name || "";
+      const toolArgs = toolCall.args || {};
+      const toolId = toolCall.id || randomUUID();
+      await this.processSingleToolCall(toolName, toolArgs, toolId);
+    }
+  }
+  /**
+   * Process a single tool call
+   * @private
+   */
+  async processSingleToolCall(toolName, toolArgs, toolId) {
     this.log("Processing tool call", { toolName, toolArgs, toolId });
     this.emit("toolCall", {
       name: toolName,
@@ -2363,36 +2381,38 @@ var GeminiLiveVoice = class _GeminiLiveVoice extends MastraVoice {
       let result;
       if (tool.execute) {
         this.log("Executing tool", { toolName, toolArgs });
-        result = await tool.execute(
-          { context: toolArgs, runtimeContext: this.runtimeContext },
-          {
-            toolCallId: toolId,
-            messages: []
-          }
-        );
+        result = await tool.execute(toolArgs, { requestContext: this.requestContext });
         this.log("Tool executed successfully", { toolName, result });
       } else {
         this.log("Tool has no execute function", { toolName });
         result = { error: "Tool has no execute function" };
       }
       const toolResultMessage = {
-        tool_result: {
-          tool_call_id: toolId,
-          result
+        toolResponse: {
+          functionResponses: [
+            {
+              id: toolId,
+              response: result
+            }
+          ]
         }
       };
-      this.sendEvent("tool_result", toolResultMessage);
+      this.sendEvent("toolResponse", toolResultMessage);
       this.log("Tool result sent", { toolName, toolId, result });
     } catch (error) {
       const errorMessage = error instanceof Error ? error.message : "Unknown error";
       this.log("Tool execution failed", { toolName, error: errorMessage });
       const errorResultMessage = {
-        tool_result: {
-          tool_call_id: toolId,
-          result: { error: errorMessage }
+        toolResponse: {
+          functionResponses: [
+            {
+              id: toolId,
+              response: { error: errorMessage }
+            }
+          ]
         }
       };
-      this.sendEvent("tool_result", errorResultMessage);
+      this.sendEvent("toolResponse", errorResultMessage);
       this.createAndEmitError("tool_execution_error" /* TOOL_EXECUTION_ERROR */, `Tool execution failed: ${errorMessage}`, {
         toolName,
         toolArgs,
@@ -2452,6 +2472,31 @@ var GeminiLiveVoice = class _GeminiLiveVoice extends MastraVoice {
     }
     return "text";
   }
+  /**
+   * Resolve Vertex AI location with sensible default
+   * @private
+   */
+  getVertexLocation() {
+    return this.options.location?.trim() || "us-central1";
+  }
+  /**
+   * Resolve the correct model identifier for Gemini API or Vertex AI
+   * @private
+   */
+  resolveModelIdentifier() {
+    const model = this.options.model ?? DEFAULT_MODEL;
+    if (!this.options.vertexAI) {
+      return `models/${model}`;
+    }
+    if (!this.options.project) {
+      throw this.createAndEmitError(
+        "project_id_missing" /* PROJECT_ID_MISSING */,
+        "Google Cloud project ID is required when using Vertex AI."
+      );
+    }
+    const location = this.getVertexLocation();
+    return `projects/${this.options.project}/locations/${location}/publishers/google/models/${model}`;
+  }
   /**
    * Send initial configuration to Gemini Live API
    * @private
@@ -2462,7 +2507,7 @@ var GeminiLiveVoice = class _GeminiLiveVoice extends MastraVoice {
     }
     const setupMessage = {
       setup: {
-        model: `models/${this.options.model}`
+        model: this.resolveModelIdentifier()
       }
     };
     if (this.options.instructions) {
@@ -2611,6 +2656,8 @@ var GeminiLiveVoice = class _GeminiLiveVoice extends MastraVoice {
       message = data;
     } else if (type === "realtime_input" && data.realtime_input) {
       message = data;
+    } else if (type === "toolResponse" && data.toolResponse) {
+      message = data;
     } else if (type === "session.update" && data.session) {
       message = data;
     } else {
@@ -2636,14 +2683,14 @@ var GeminiLiveVoice = class _GeminiLiveVoice extends MastraVoice {
    *   inputSchema: z.object({
    *     location: z.string().describe("The city and state, e.g. San Francisco, CA"),
    *   }),
-   *   execute: async ({ context }) => {
+   *   execute: async (inputData) => {
    *     // Fetch weather data from an API
    *     const response = await fetch(
-   *       `https://api.weather.com?location=${encodeURIComponent(context.location)}`,
+   *       `https://api.weather.com?location=${encodeURIComponent(inputData.location)}`,
    *     );
    *     const data = await response.json();
    *     return {
-   *       message: `The current temperature in ${context.location} is ${data.temperature}°F with ${data.conditions}.`,
+   *       message: `The current temperature in ${inputData.location} is ${data.temperature}°F with ${data.conditions}.`,
    *     };
    *   },
    * });
@@ -2661,12 +2708,12 @@ var GeminiLiveVoice = class _GeminiLiveVoice extends MastraVoice {
    * Get the current tools configured for this voice instance
    * @returns Object containing the current tools
    */
-  getTools() {
+  listTools() {
     return this.tools;
   }
   log(message, ...args) {
     if (this.debug) {
-      console.log(`[GeminiLiveVoice] ${message}`, ...args);
+      console.info(`[GeminiLiveVoice] ${message}`, ...args);
     }
   }
   /**