@teamlearners/clawops 0.4.0 → 0.5.1

package/dist/agent/index.cjs

@@ -1815,8 +1815,8 @@ var GeminiRealtime = class {
   _voice;
   _language;
   _greeting;
-  _generationConfig;
-  _ws = null;
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
+  _session = null;
   _call = null;
   _tools = null;
   _recorder = null;
@@ -1830,7 +1830,6 @@ var GeminiRealtime = class {
     this._voice = options.voice ?? "Kore";
     this._language = options.language ?? "ko";
     this._greeting = options.greeting ?? true;
-    this._generationConfig = options.generationConfig;
   }
   /** Inject per-call ToolRegistry. */
   setToolRegistry(registry) {
@@ -1849,83 +1848,78 @@ var GeminiRealtime = class {
     if (!this._apiKey) {
       throw new Error("Google API key is required. Set GOOGLE_API_KEY or pass apiKey option.");
     }
-    const { WebSocket } = await import('ws');
-    const url = `wss://generativelanguage.googleapis.com/ws/google.ai.generativelanguage.v1beta.GenerativeService.BidiGenerateContent?key=${this._apiKey}`;
-    this._ws = new WebSocket(url);
-    return new Promise((resolve, reject) => {
-      const ws = this._ws;
-      ws.on("open", () => {
-        this._sendSetup();
-        this._waitSetupComplete().then(() => {
-          if (this._greeting) {
-            this._sendGreeting();
+    const { GoogleGenAI } = await import('@google/genai/node');
+    const client = new GoogleGenAI({ apiKey: this._apiKey });
+    const config = {
+      responseModalities: ["AUDIO"],
+      speechConfig: {
+        voiceConfig: {
+          prebuiltVoiceConfig: {
+            voiceName: this._voice
           }
-          this._receiveLoop();
-          resolve();
-        }).catch(reject);
-      });
-      ws.on("close", () => {
-        this._closed = true;
-      });
-      ws.on("error", (err) => {
-        if (!this._ws) {
-          reject(err);
         }
-        console.error("[GeminiRealtime] WebSocket error:", err.message);
-      });
+      },
+      inputAudioTranscription: {},
+      outputAudioTranscription: {}
+    };
+    if (this._systemPrompt) {
+      config["systemInstruction"] = this._systemPrompt;
+    }
+    const toolSchemas = this._buildToolSchemas();
+    if (toolSchemas.length > 0) {
+      config["tools"] = [{ functionDeclarations: toolSchemas }];
+    }
+    this._session = await client.live.connect({
+      model: this._model,
+      config,
+      callbacks: {
+        onmessage: (msg) => this._handleMessage(msg),
+        onerror: (err) => {
+          console.error("[GeminiRealtime] SDK error:", err);
+        },
+        onclose: () => {
+          this._closed = true;
+        }
+      }
     });
+    if (this._greeting) {
+      this._session.sendClientContent({
+        turns: [
+          {
+            role: "user",
+            parts: [{ text: "\uC778\uC0AC\uD574 \uC8FC\uC138\uC694." }]
+          }
+        ],
+        turnComplete: true
+      });
+    }
   }
   feedAudio(audio) {
-    if (this._ws && this._ws.readyState === 1 && !this._closed) {
+    if (this._session && !this._closed) {
       const pcm8k = ulawToPcm16(audio);
+      if (this._recorder) {
+        this._recorder.writeInbound(pcm8k);
+      }
       const pcm16k = resamplePcm16(pcm8k, 8e3, 16e3);
-      this._ws.send(
-        JSON.stringify({
-          realtimeInput: {
-            mediaChunks: [
-              {
-                mimeType: "audio/pcm;rate=16000",
-                data: pcm16k.toString("base64")
-              }
-            ]
-          }
-        })
-      );
+      this._session.sendRealtimeInput({
+        audio: {
+          data: Buffer.from(pcm16k).toString("base64"),
+          mimeType: "audio/pcm;rate=16000"
+        }
+      });
     }
   }
   async stop() {
     this._closed = true;
-    if (this._ws) {
-      this._ws.close();
-      this._ws = null;
-    }
-  }
-  _sendSetup() {
-    if (!this._ws || this._ws.readyState !== 1) return;
-    const setupConfig = {
-      model: `models/${this._model}`,
-      generationConfig: {
-        responseModalities: ["AUDIO"],
-        speechConfig: {
-          voiceConfig: {
-            prebuiltVoiceConfig: {
-              voiceName: this._voice
-            }
-          }
-        },
-        ...this._generationConfig
-      },
-      realtimeInputConfig: {
-        automaticActivityDetection: {
-          disabled: false
-        }
+    if (this._session) {
+      try {
+        this._session.close();
+      } catch {
       }
-    };
-    if (this._systemPrompt) {
-      setupConfig["systemInstruction"] = {
-        parts: [{ text: this._systemPrompt }]
-      };
+      this._session = null;
     }
+  }
+  _buildToolSchemas() {
     const toolDefs = this._tools ? this._tools.toOpenAITools().map((t) => ({
       name: t.function.name,
       description: t.function.description,
@@ -1934,107 +1928,46 @@ var GeminiRealtime = class {
       )
     })) : [];
     toolDefs.push(HANG_UP_TOOL2);
-    setupConfig["tools"] = [{ functionDeclarations: toolDefs }];
-    this._ws.send(JSON.stringify({ setup: setupConfig }));
-  }
-  _waitSetupComplete() {
-    return new Promise((resolve, reject) => {
-      if (!this._ws) {
-        reject(new Error("WebSocket not connected"));
-        return;
-      }
-      const onMessage = (data) => {
-        try {
-          const msg = JSON.parse(data.toString());
-          if ("setupComplete" in msg) {
-            this._ws?.removeListener("message", onMessage);
-            resolve();
-          }
-        } catch {
-        }
-      };
-      this._ws.on("message", onMessage);
-    });
-  }
-  _sendGreeting() {
-    if (!this._ws || this._ws.readyState !== 1) return;
-    this._ws.send(
-      JSON.stringify({
-        clientContent: {
-          turns: [
-            {
-              role: "user",
-              parts: [{ text: "\uC778\uC0AC\uD574 \uC8FC\uC138\uC694." }]
-            }
-          ],
-          turnComplete: true
-        }
-      })
-    );
-  }
-  _receiveLoop() {
-    if (!this._ws) return;
-    this._ws.on("message", (data) => {
-      try {
-        const msg = JSON.parse(data.toString());
-        this._handleMessage(msg);
-      } catch {
-      }
-    });
+    return toolDefs;
   }
   _handleMessage(msg) {
     if (!this._call) return;
-    const serverContent = msg["serverContent"];
+    const serverContent = msg.serverContent;
     if (serverContent) {
-      const modelTurn = serverContent["modelTurn"];
+      const modelTurn = serverContent.modelTurn;
       if (modelTurn) {
-        const parts = modelTurn["parts"];
-        if (parts) {
-          for (const part of parts) {
-            const inlineData = part["inlineData"];
-            if (inlineData && inlineData["data"]) {
-              const mimeType = inlineData["mimeType"] ?? "";
-              if (mimeType.includes("audio")) {
-                this._handleAudioData(inlineData["data"]);
-              }
-            }
-            const text = part["text"];
-            if (text && this._call) {
-              this._call._emit("transcript", "assistant", text);
+        for (const part of modelTurn.parts ?? []) {
+          const inlineData = part.inlineData;
+          if (inlineData?.data) {
+            const mimeType = inlineData.mimeType ?? "";
+            if (mimeType.includes("audio")) {
+              this._handleAudioData(inlineData.data);
             }
           }
         }
       }
-      if (serverContent["turnComplete"]) {
+      if (serverContent.turnComplete) {
         this._flushAudioRemainder();
       }
-      if (serverContent["interrupted"]) {
+      if (serverContent.interrupted) {
         if (this._call) {
           this._call.clearAudio();
         }
         this._sentAudioChunks = 0;
         this._audioRemainder = Buffer.alloc(0);
       }
-    }
-    const inputTranscription = msg["inputTranscription"];
-    if (inputTranscription) {
-      const text = inputTranscription["text"];
-      if (text && this._call) {
-        this._call._emit("transcript", "user", text);
+      const inputText = serverContent.inputTranscription?.text;
+      if (inputText && this._call) {
+        this._call._emit("transcript", "user", inputText);
       }
-    }
-    const outputTranscription = msg["outputTranscription"];
-    if (outputTranscription) {
-      const text = outputTranscription["text"];
-      if (text && this._call) {
-        this._call._emit("transcript", "assistant", text);
+      const outputText = serverContent.outputTranscription?.text;
+      if (outputText && this._call) {
+        this._call._emit("transcript", "assistant", outputText);
       }
     }
-    const toolCall = msg["toolCall"];
-    if (toolCall) {
-      this._handleToolCall(toolCall);
+    if (msg.toolCall) {
+      this._handleToolCall(msg.toolCall);
     }
-    if (msg["toolCallCancellation"]) ;
   }
   _handleAudioData(b64Data) {
     if (!this._call) return;
@@ -2065,13 +1998,13 @@ var GeminiRealtime = class {
     }
   }
   async _handleToolCall(toolCall) {
-    const functionCalls = toolCall["functionCalls"];
+    const functionCalls = toolCall.functionCalls;
     if (!functionCalls) return;
     const responses = [];
     for (const fc of functionCalls) {
-      const name = fc["name"];
-      const fcId = fc["id"] ?? "";
-      const args = fc["args"] ?? {};
+      const name = fc.name ?? "";
+      const fcId = fc.id ?? "";
+      const args = fc.args ?? {};
       if (name === "hang_up") {
         if (this._call) {
           this._call.hangup();
@@ -2099,14 +2032,10 @@ var GeminiRealtime = class {
         });
       }
     }
-    if (this._ws && this._ws.readyState === 1) {
-      this._ws.send(
-        JSON.stringify({
-          toolResponse: {
-            functionResponses: responses
-          }
-        })
-      );
+    if (this._session) {
+      this._session.sendToolResponse({
+        functionResponses: responses
+      });
     }
   }
 };