@teamlearners/clawops 0.5.4 → 0.5.8

package/dist/agent/index.js

@@ -335,6 +335,7 @@ function createPipelineLogger(parent) {
 // src/agent/control-ws.ts
 var INITIAL_RECONNECT_DELAY = 1e3;
 var MAX_RECONNECT_DELAY = 3e4;
+var PING_TIMEOUT = 6e4;
 function buildControlWsUrl(options) {
   const scheme = options.baseUrl.startsWith("https") ? "wss" : "ws";
   const host = options.baseUrl.replace(/^https?:\/\//, "").replace(/\/$/, "");
@@ -360,6 +361,7 @@ var ControlWebSocket = class {
   _connectedResolve = null;
   _connectedPromise;
   _log = NOOP_LOGGER;
+  _pingTimer = null;
   setLogger(logger) {
     this._log = logger;
   }
@@ -390,6 +392,7 @@ var ControlWebSocket = class {
   /** Close the WebSocket and stop reconnecting. */
   close() {
     this._closed = true;
+    this._clearPingTimer();
     if (this._ws) {
       this._ws.close();
       this._ws = null;
@@ -406,12 +409,16 @@ var ControlWebSocket = class {
     this._ws = ws;
     ws.on("open", () => {
       this._reconnectDelay = INITIAL_RECONNECT_DELAY;
+      this._resetPingTimer();
       if (this._connectedResolve) {
         this._connectedResolve();
         this._connectedResolve = null;
       }
       this._log.info("Control WS connected: %s", this._url);
     });
+    ws.on("ping", () => {
+      this._resetPingTimer();
+    });
     ws.on("message", (data) => {
       try {
         const msg = JSON.parse(data.toString());
@@ -421,6 +428,7 @@ var ControlWebSocket = class {
       }
     });
     ws.on("close", () => {
+      this._clearPingTimer();
       if (!this._closed) {
         this._scheduleReconnect();
       }
@@ -446,6 +454,21 @@ var ControlWebSocket = class {
       }
     }
   }
+  _resetPingTimer() {
+    this._clearPingTimer();
+    this._pingTimer = setTimeout(() => {
+      this._log.warn("Control WS ping timeout, closing connection");
+      if (this._ws) {
+        this._ws.terminate();
+      }
+    }, PING_TIMEOUT);
+  }
+  _clearPingTimer() {
+    if (this._pingTimer) {
+      clearTimeout(this._pingTimer);
+      this._pingTimer = null;
+    }
+  }
   _scheduleReconnect() {
     const delay = this._reconnectDelay;
     this._reconnectDelay = Math.min(this._reconnectDelay * 2, MAX_RECONNECT_DELAY);
@@ -1436,6 +1459,7 @@ var ClawOpsAgent = class {
   }
   /** Connect to the ClawOps platform and start listening for calls. */
   async connect() {
+    if (this._controlWs) return;
     if (!this._apiKey) {
       throw new AgentError("API key is required. Set CLAWOPS_API_KEY or pass apiKey option.");
     }
@@ -1750,20 +1774,13 @@ var ClawOpsAgent = class {
   }
 };
 
-// src/agent/pipeline/openai-realtime.ts
-var OPENAI_REALTIME_URL = "wss://api.openai.com/v1/realtime?model=";
-var HANG_UP_TOOL = {
-  type: "function",
+// src/agent/pipeline/builtin-tool-schemas.ts
+var HANG_UP = {
   name: "hang_up",
   description: "End the phone call. Use when the conversation is finished or the caller says goodbye.",
-  parameters: {
-    type: "object",
-    properties: {},
-    required: []
-  }
+  parameters: { type: "object", properties: {} }
 };
-var COLLECT_DTMF_TOOL = {
-  type: "function",
+var COLLECT_DTMF = {
   name: "collect_dtmf",
   description: "\uC0AC\uC6A9\uC790\uB85C\uBD80\uD130 DTMF(\uC804\uD654 \uD0A4\uD328\uB4DC) \uC785\uB825\uC744 \uC218\uC9D1\uD569\uB2C8\uB2E4. \uBC18\uB4DC\uC2DC \uC0AC\uC6A9\uC790\uC5D0\uAC8C \uBB34\uC5C7\uC744 \uC785\uB825\uD574\uC57C \uD558\uB294\uC9C0 \uC548\uB0B4\uD55C \uD6C4 \uD638\uCD9C\uD558\uC138\uC694.",
   parameters: {
@@ -1776,8 +1793,7 @@ var COLLECT_DTMF_TOOL = {
     required: ["max_digits"]
   }
 };
-var SEND_DTMF_TOOL = {
-  type: "function",
+var SEND_DTMF = {
   name: "send_dtmf",
   description: "DTMF \uC2E0\uD638\uB97C \uC804\uC1A1\uD569\uB2C8\uB2E4. ARS \uBA54\uB274 \uD0D0\uC0C9\uC774\uB098 \uB0B4\uC120\uBC88\uD638 \uC785\uB825 \uC2DC \uC0AC\uC6A9\uD569\uB2C8\uB2E4.",
   parameters: {
@@ -1785,199 +1801,530 @@ var SEND_DTMF_TOOL = {
     properties: {
       digits: {
         type: "string",
-        description: "\uC804\uC1A1\uD560 \uBC88\uD638 (0-9, *, #). 'w'\uB294 500ms \uB300\uAE30, 'W'\uB294 1000ms \uB300\uAE30."
+        description: "\uC804\uC1A1\uD560 \uBC88\uD638 (0-9, *, #). 'w'\uB294 500ms \uB300\uAE30, 'W'\uB294 1000ms \uB300\uAE30. \uC608: '1', '1234#', '1w2'"
       }
     },
     required: ["digits"]
   }
 };
-var OpenAIRealtime = class {
-  _apiKey;
-  _systemPrompt;
-  _model;
-  _voice;
-  _language;
-  _eagerness;
-  _greeting;
-  _log = NOOP_LOGGER;
-  _builtinTools = null;
-  setLogger(logger) {
-    this._log = logger;
+var TOOL_MAP = /* @__PURE__ */ new Map([
+  ["hang_up" /* HANG_UP */, HANG_UP],
+  ["collect_dtmf" /* COLLECT_DTMF */, COLLECT_DTMF],
+  ["send_dtmf" /* SEND_DTMF */, SEND_DTMF]
+]);
+var BUILTIN_TOOL_NAMES = new Set(
+  Array.from(TOOL_MAP.values()).map((s) => s.name)
+);
+function toChatCompletions(schema) {
+  return {
+    type: "function",
+    function: {
+      name: schema.name,
+      description: schema.description,
+      parameters: schema.parameters
+    }
+  };
+}
+function toRealtime(schema) {
+  return {
+    type: "function",
+    name: schema.name,
+    description: schema.description,
+    parameters: schema.parameters
+  };
+}
+function toGemini(schema) {
+  return {
+    name: schema.name,
+    description: schema.description,
+    parameters: schema.parameters
+  };
+}
+var CONVERTERS = {
+  chat: toChatCompletions,
+  realtime: toRealtime,
+  gemini: toGemini
+};
+function getBuiltinToolSchemas(builtinTools, fmt) {
+  const converter = CONVERTERS[fmt];
+  const result = [];
+  for (const [toolEnum, schema] of TOOL_MAP) {
+    if (builtinTools === null || builtinTools.has(toolEnum)) {
+      result.push(converter(schema));
+    }
   }
-  setBuiltinTools(tools) {
-    this._builtinTools = tools;
+  return result;
+}
+function isBuiltinTool(name) {
+  return BUILTIN_TOOL_NAMES.has(name);
+}
+async function executeBuiltinTool(funcName, args, call) {
+  if (funcName === "hang_up") {
+    await call.hangup();
+    return "";
   }
-  _ws = null;
-  _call = null;
+  if (funcName === "collect_dtmf") {
+    try {
+      const result = await call.collectDtmf({
+        maxDigits: args["max_digits"] ?? 4,
+        finishOnKey: args["finish_on_key"] ?? "#",
+        timeout: args["timeout"] ?? 5
+      });
+      return result || "(\uD0C0\uC784\uC544\uC6C3 - \uC785\uB825 \uC5C6\uC74C)";
+    } catch (e) {
+      return `Error: ${e}`;
+    }
+  }
+  if (funcName === "send_dtmf") {
+    try {
+      await call.sendDtmfSequence(args["digits"] ?? "");
+      return "sent";
+    } catch (e) {
+      return `Error: ${e}`;
+    }
+  }
+  return null;
+}
+
+// src/agent/pipeline/pipeline-session.ts
+var PipelineSession = class {
+  _stt;
+  _llm;
+  _tts;
+  _systemPrompt;
+  _greeting;
+  _language;
+  _temperature;
+  _maxTokens;
+  _sampleRate;
+  _interruptOnSpeech;
+  _callSession = null;
   _tools = null;
   _recorder = null;
-  _closed = false;
-  // Truncation / barge-in tracking (matching Python SDK)
-  _lastAssistantItem = null;
-  _responseStartTs = null;
-  _sentAudioChunks = 0;
-  _audioRemainder = Buffer.alloc(0);
-  // Response state tracking — prevent sending response.create while one is active
-  _responseInProgress = false;
-  _onResponseDone = null;
-  constructor(options = {}) {
-    this._apiKey = options.apiKey ?? process.env["OPENAI_API_KEY"] ?? "";
-    this._systemPrompt = options.systemPrompt ?? "";
-    this._model = options.model ?? "gpt-realtime-1.5";
-    this._voice = options.voice ?? "marin";
-    this._language = options.language ?? "ko";
-    this._eagerness = options.eagerness ?? "high";
+  _conversation = [];
+  _audioBuffer = [];
+  _running = false;
+  _speaking = false;
+  _builtinTools = null;
+  _log = NOOP_LOGGER;
+  constructor(options) {
+    this._stt = options.stt;
+    this._llm = options.llm;
+    this._tts = options.tts;
+    this._systemPrompt = options.systemPrompt;
     this._greeting = options.greeting ?? true;
+    this._language = options.language ?? "ko";
+    this._temperature = options.temperature;
+    this._maxTokens = options.maxTokens;
+    this._sampleRate = options.sampleRate ?? 8e3;
+    this._interruptOnSpeech = options.interruptOnSpeech ?? true;
+    if (options.toolRegistry) this._tools = options.toolRegistry;
+    if (options.recorder) this._recorder = options.recorder;
   }
-  /** Inject per-call ToolRegistry. */
   setToolRegistry(registry) {
     this._tools = registry;
   }
-  /** Inject per-call AudioRecorder. */
   setRecorder(recorder) {
     this._recorder = recorder;
   }
-  async start(callSession, tools) {
-    this._call = callSession;
-    if (tools) this._tools = tools;
-    this._closed = false;
-    this._lastAssistantItem = null;
-    this._responseStartTs = null;
-    this._sentAudioChunks = 0;
-    this._audioRemainder = Buffer.alloc(0);
-    if (!this._apiKey) {
-      throw new Error("OpenAI API key is required. Set OPENAI_API_KEY or pass apiKey option.");
+  setBuiltinTools(tools) {
+    this._builtinTools = tools;
+  }
+  setLogger(logger) {
+    this._log = logger;
+    if ("setLogger" in this._stt && typeof this._stt.setLogger === "function") {
+      this._stt.setLogger(logger);
     }
-    const { WebSocket } = await import('ws');
-    const url = `${OPENAI_REALTIME_URL}${this._model}`;
-    this._ws = new WebSocket(url, {
-      headers: {
-        Authorization: `Bearer ${this._apiKey}`,
-        "OpenAI-Beta": "realtime=v1"
-      }
-    });
-    return new Promise((resolve, reject) => {
-      const ws = this._ws;
-      ws.on("open", () => {
-        this._sendSessionUpdate();
-        this._log.info("OpenAI Realtime connected");
-        if (this._greeting) {
-          this._send({ type: "response.create" });
-        }
-        resolve();
-      });
-      ws.on("message", (data) => {
-        try {
-          const msg = JSON.parse(data.toString());
-          this._handleMessage(msg);
-        } catch {
-        }
-      });
-      ws.on("close", () => {
-        this._closed = true;
+    if ("setLogger" in this._tts && typeof this._tts.setLogger === "function") {
+      this._tts.setLogger(logger);
+    }
+  }
+  async start(callSession, tools) {
+    this._callSession = callSession;
+    this._tools = tools ?? null;
+    this._running = true;
+    this._log.info("PipelineSession started");
+    this._conversation = [];
+    if (this._systemPrompt) {
+      this._conversation.push({
+        role: "system",
+        content: this._systemPrompt
       });
-      ws.on("error", (err) => {
-        if (!this._ws) {
-          reject(err);
-        }
-        this._log.error({ err }, "OpenAI Realtime WS error");
+    }
+    if (this._greeting) {
+      this._generateGreeting().catch((err) => {
+        this._log.error({ err }, "Greeting error");
       });
+    }
+    this._runSttLoop().catch((err) => {
+      this._log.error({ err }, "STT loop error");
     });
   }
-  async feedDtmf(digits) {
-    await this._waitForResponseDone();
-    this._send({
-      type: "conversation.item.create",
-      item: {
-        type: "message",
-        role: "user",
-        content: [{ type: "input_text", text: `[DTMF \uC785\uB825: ${digits}]` }]
-      }
-    });
-    this._send({ type: "response.create" });
-  }
   feedAudio(audio) {
-    if (this._ws && this._ws.readyState === 1 && !this._closed) {
-      this._send({
-        type: "input_audio_buffer.append",
-        audio: audio.toString("base64")
-      });
-    }
-  }
-  async stop() {
-    this._closed = true;
-    if (this._ws) {
-      this._ws.close();
-      this._ws = null;
+    if (this._running) {
+      this._audioBuffer.push(audio);
     }
   }
-  _sendSessionUpdate() {
-    if (!this._ws || this._ws.readyState !== 1) return;
-    const toolSchemas = this._tools ? this._tools.toOpenAITools().map((t) => ({ type: "function", ...t.function })) : [];
-    if (!this._builtinTools || this._builtinTools.has("hang_up" /* HANG_UP */))
-      toolSchemas.push(HANG_UP_TOOL);
-    if (!this._builtinTools || this._builtinTools.has("collect_dtmf" /* COLLECT_DTMF */))
-      toolSchemas.push(COLLECT_DTMF_TOOL);
-    if (!this._builtinTools || this._builtinTools.has("send_dtmf" /* SEND_DTMF */))
-      toolSchemas.push(SEND_DTMF_TOOL);
-    this._send({
-      type: "session.update",
-      session: {
-        modalities: ["text", "audio"],
-        voice: this._voice,
-        instructions: this._systemPrompt,
-        input_audio_format: "g711_ulaw",
-        output_audio_format: "g711_ulaw",
-        input_audio_transcription: {
-          model: "whisper-1",
-          language: this._language
-        },
-        input_audio_noise_reduction: { type: "far_field" },
-        turn_detection: {
-          type: "semantic_vad",
-          interrupt_response: true,
-          eagerness: this._eagerness
-        },
-        tools: toolSchemas
-      }
+  async feedDtmf(digits) {
+    this._conversation.push({
+      role: "user",
+      content: `[DTMF \uC785\uB825: ${digits}]`
     });
+    await this._respond();
   }
-  _handleMessage(msg) {
-    const type = msg["type"];
-    switch (type) {
-      case "response.audio.delta": {
-        this._handleAudioDelta(msg);
-        break;
-      }
-      case "response.audio.done": {
-        if (this._audioRemainder.length > 0) {
-          const padded = Buffer.concat([
-            this._audioRemainder,
-            Buffer.alloc(160 - this._audioRemainder.length, 255)
-          ]);
-          if (this._call) {
-            this._call.sendAudio(padded);
-          }
-          this._sentAudioChunks++;
-          this._audioRemainder = Buffer.alloc(0);
+  async stop() {
+    this._running = false;
+    this._log.info("PipelineSession stopped");
+    this._audioBuffer = [];
+  }
+  async _runSttLoop() {
+    const audioStream = this._createAudioStream();
+    for await (const event of this._stt.transcribe(audioStream, {
+      sampleRate: this._sampleRate
+    })) {
+      if (!this._running) break;
+      if (event.type === "interim" && this._speaking && this._interruptOnSpeech) {
+        this._speaking = false;
+        if (this._callSession) {
+          this._callSession.clearAudio();
         }
-        break;
-      }
-      case "input_audio_buffer.speech_started": {
-        this._handleTruncation();
-        break;
+        this._log.info('Barge-in: "%s"', event.transcript.substring(0, 30));
       }
-      case "response.output_item.done": {
-        const item = msg["item"];
-        if (item && item["type"] === "function_call") {
-          this._handleToolCall(item);
-        }
-        break;
+      if (event.type === "final" && event.transcript.trim()) {
+        this._log.info("STT: %s", event.transcript);
+        await this._handleUserSpeech(event.transcript);
       }
-      case "conversation.item.input_audio_transcription.completed": {
-        if (this._call) {
-          this._call._emit("transcript", "user", msg["transcript"] ?? "");
-        }
+    }
+  }
+  async *_createAudioStream() {
+    while (this._running) {
+      if (this._audioBuffer.length > 0) {
+        const ulaw = this._audioBuffer.shift();
+        const pcm8k = ulawToPcm16(ulaw);
+        const pcm16k = resamplePcm16(pcm8k, 8e3, 16e3);
+        yield pcm16k;
+      } else {
+        await new Promise((resolve) => setTimeout(resolve, 20));
+      }
+    }
+  }
+  async _generateGreeting() {
+    await new Promise((resolve) => setTimeout(resolve, 500));
+    await this._respond();
+  }
+  async _handleUserSpeech(transcript) {
+    this._conversation.push({ role: "user", content: transcript });
+    await this._respond();
+  }
+  _buildEffectiveTools() {
+    const builtinSchemas = getBuiltinToolSchemas(this._builtinTools, "chat");
+    const dtmfSchemas = builtinSchemas.filter((s) => {
+      const name = s["function"]?.["name"];
+      return name === "collect_dtmf" || name === "send_dtmf";
+    });
+    if (dtmfSchemas.length === 0) return this._tools ?? void 0;
+    const base = this._tools ? this._tools.fork() : new ToolRegistry();
+    for (const schema of dtmfSchemas) {
+      const fn = schema["function"];
+      const params = fn["parameters"];
+      base.register({
+        name: fn["name"],
+        description: fn["description"],
+        parameters: params["properties"] ?? {},
+        required: params["required"] ?? [],
+        handler: async () => ""
+      });
+    }
+    return base;
+  }
+  async _respond() {
+    let fullResponse = "";
+    const textChunks = [];
+    const effectiveTools = this._buildEffectiveTools();
+    const llmStream = this._llm.generate(this._conversation, {
+      tools: effectiveTools,
+      temperature: this._temperature,
+      maxTokens: this._maxTokens
+    });
+    for await (const chunk of llmStream) {
+      if (!this._running) break;
+      if (chunk.type === "text" && chunk.text) {
+        textChunks.push(chunk.text);
+        fullResponse += chunk.text;
+      } else if (chunk.type === "tool_call" && chunk.toolCall) {
+        await this._handleToolCall(chunk);
+      }
+    }
+    if (fullResponse.trim()) {
+      this._log.info("Assistant: %s", fullResponse.substring(0, 100));
+      this._conversation.push({ role: "assistant", content: fullResponse });
+      await this._synthesizeAndSend(fullResponse);
+    }
+  }
+  async _handleToolCall(chunk) {
+    if (!chunk.toolCall) return;
+    const { id, name, arguments: argsStr } = chunk.toolCall;
+    try {
+      const args = JSON.parse(argsStr);
+      if (BUILTIN_TOOL_NAMES.has(name) && this._callSession) {
+        const result2 = await executeBuiltinTool(name, args, this._callSession);
+        if (result2 !== null) {
+          if (name === "hang_up") return;
+          this._conversation.push({ role: "tool", content: result2, tool_call_id: id, name });
+          await this._respond();
+          return;
+        }
+      }
+      if (!this._tools) return;
+      const result = await this._tools.call(name, args);
+      this._conversation.push({
+        role: "assistant",
+        content: ""
+        // Tool call info stored in the message flow
+      });
+      this._conversation.push({
+        role: "tool",
+        content: typeof result === "string" ? result : JSON.stringify(result),
+        tool_call_id: id,
+        name
+      });
+      const effectiveTools = this._buildEffectiveTools();
+      let followUpText = "";
+      const followUpStream = this._llm.generate(this._conversation, {
+        tools: effectiveTools,
+        temperature: this._temperature,
+        maxTokens: this._maxTokens
+      });
+      for await (const followChunk of followUpStream) {
+        if (!this._running) break;
+        if (followChunk.type === "text" && followChunk.text) {
+          followUpText += followChunk.text;
+        }
+      }
+      if (followUpText.trim()) {
+        this._conversation.push({ role: "assistant", content: followUpText });
+        await this._synthesizeAndSend(followUpText);
+      }
+    } catch (err) {
+      this._log.error({ err }, "Tool call failed: %s", name);
+    }
+  }
+  async _synthesizeAndSend(text) {
+    if (!this._callSession || !this._running) return;
+    this._speaking = true;
+    try {
+      for await (const audioChunk of this._tts.synthesize(text, {
+        sampleRate: this._sampleRate
+      })) {
+        if (!this._running || !this._speaking) break;
+        if (this._recorder) {
+          const pcm8k2 = this._sampleRate !== 8e3 ? resamplePcm16(audioChunk, this._sampleRate, 8e3) : audioChunk;
+          this._recorder.writeOutbound(pcm8k2);
+        }
+        const pcm8k = resamplePcm16(audioChunk, this._sampleRate, 8e3);
+        const ulaw = pcm16ToUlaw(pcm8k);
+        for (let off = 0; off < ulaw.length; off += 160) {
+          let chunk = ulaw.subarray(off, off + 160);
+          if (chunk.length < 160) {
+            chunk = Buffer.concat([chunk, Buffer.alloc(160 - chunk.length, 255)]);
+          }
+          this._callSession.sendAudio(chunk);
+        }
+      }
+    } catch (err) {
+      this._log.error({ err }, "TTS error");
+    } finally {
+      this._speaking = false;
+    }
+  }
+};
+
+// src/agent/pipeline/realtime/openai-realtime.ts
+var OPENAI_REALTIME_URL = "wss://api.openai.com/v1/realtime?model=";
+var OpenAIRealtime = class {
+  _apiKey;
+  _systemPrompt;
+  _model;
+  _voice;
+  _language;
+  _turnDetection;
+  _greeting;
+  _log = NOOP_LOGGER;
+  _builtinTools = null;
+  setLogger(logger) {
+    this._log = logger;
+  }
+  setBuiltinTools(tools) {
+    this._builtinTools = tools;
+  }
+  _ws = null;
+  _call = null;
+  _tools = null;
+  _recorder = null;
+  _closed = false;
+  // PlaybackState — 현재 재생 중인 응답 상태
+  _playback = null;
+  _latestMediaTs = 0;
+  // Pending tool call tracking — 인터럽트 시 취소용
+  _pendingToolCalls = /* @__PURE__ */ new Map();
+  // Response state tracking — prevent sending response.create while one is active
+  _responseInProgress = false;
+  _onResponseDone = null;
+  constructor(options = {}) {
+    this._apiKey = options.apiKey ?? process.env["OPENAI_API_KEY"] ?? "";
+    this._systemPrompt = options.systemPrompt ?? "";
+    this._model = options.model ?? "gpt-realtime-1.5";
+    this._voice = options.voice ?? "marin";
+    this._language = options.language ?? "ko";
+    this._turnDetection = options.turnDetection !== void 0 ? options.turnDetection : { type: "semantic_vad", eagerness: "medium", interrupt_response: true };
+    this._greeting = options.greeting ?? true;
+  }
+  /** Inject per-call ToolRegistry. */
+  setToolRegistry(registry) {
+    this._tools = registry;
+  }
+  /** Inject per-call AudioRecorder. */
+  setRecorder(recorder) {
+    this._recorder = recorder;
+  }
+  async start(callSession, tools) {
+    this._call = callSession;
+    if (tools) this._tools = tools;
+    this._closed = false;
+    this._playback = null;
+    this._latestMediaTs = 0;
+    if (!this._apiKey) {
+      throw new Error("OpenAI API key is required. Set OPENAI_API_KEY or pass apiKey option.");
+    }
+    const { WebSocket } = await import('ws');
+    const url = `${OPENAI_REALTIME_URL}${this._model}`;
+    this._ws = new WebSocket(url, {
+      headers: {
+        Authorization: `Bearer ${this._apiKey}`,
+        "OpenAI-Beta": "realtime=v1"
+      }
+    });
+    return new Promise((resolve, reject) => {
+      const ws = this._ws;
+      ws.on("open", () => {
+        this._sendSessionUpdate();
+        this._log.info("OpenAI Realtime connected");
+        if (this._greeting) {
+          this._send({ type: "response.create" });
+        }
+        resolve();
+      });
+      ws.on("message", (data) => {
+        try {
+          const msg = JSON.parse(data.toString());
+          this._handleMessage(msg);
+        } catch {
+        }
+      });
+      ws.on("close", () => {
+        this._closed = true;
+      });
+      ws.on("error", (err) => {
+        if (!this._ws) {
+          reject(err);
+        }
+        this._log.error({ err }, "OpenAI Realtime WS error");
+      });
+    });
+  }
+  async feedDtmf(digits) {
+    await this._waitForResponseDone();
+    this._send({
+      type: "conversation.item.create",
+      item: {
+        type: "message",
+        role: "user",
+        content: [{ type: "input_text", text: `[DTMF \uC785\uB825: ${digits}]` }]
+      }
+    });
+    this._send({ type: "response.create" });
+  }
+  feedAudio(audio) {
+    this._latestMediaTs = Date.now();
+    if (this._ws && this._ws.readyState === 1 && !this._closed) {
+      this._send({
+        type: "input_audio_buffer.append",
+        audio: audio.toString("base64")
+      });
+    }
+  }
+  async stop() {
+    this._closed = true;
+    for (const [, controller] of this._pendingToolCalls) {
+      controller.abort();
+    }
+    this._pendingToolCalls.clear();
+    if (this._ws) {
+      this._ws.close();
+      this._ws = null;
+    }
+  }
+  _sendSessionUpdate() {
+    if (!this._ws || this._ws.readyState !== 1) return;
+    const toolSchemas = this._tools ? this._tools.toOpenAITools().map((t) => ({ type: "function", ...t.function })) : [];
+    toolSchemas.push(...getBuiltinToolSchemas(this._builtinTools, "realtime"));
+    this._send({
+      type: "session.update",
+      session: {
+        modalities: ["text", "audio"],
+        voice: this._voice,
+        instructions: this._systemPrompt,
+        input_audio_format: "g711_ulaw",
+        output_audio_format: "g711_ulaw",
+        input_audio_transcription: {
+          model: "whisper-1",
+          language: this._language
+        },
+        input_audio_noise_reduction: { type: "far_field" },
+        turn_detection: this._turnDetection,
+        tools: toolSchemas
+      }
+    });
+  }
+  _handleMessage(msg) {
+    const type = msg["type"];
+    switch (type) {
+      case "response.audio.delta": {
+        this._handleAudioDelta(msg);
+        break;
+      }
+      case "response.audio.done": {
+        if (this._playback) {
+          this._playback.generating = false;
+          if (this._playback.audioRemainder.length > 0) {
+            const padded = Buffer.concat([
+              this._playback.audioRemainder,
+              Buffer.alloc(160 - this._playback.audioRemainder.length, 255)
+            ]);
+            if (this._call) {
+              this._call.sendAudio(padded);
+            }
+            this._playback.sentChunks++;
+            this._playback.audioRemainder = Buffer.alloc(0);
+          }
+        }
+        break;
+      }
+      case "input_audio_buffer.speech_started": {
+        this._handleTruncation();
+        break;
+      }
+      case "response.output_item.done": {
+        const item = msg["item"];
+        if (item && item["type"] === "function_call") {
+          this._handleToolCall(item);
+        }
+        break;
+      }
+      case "conversation.item.input_audio_transcription.completed": {
+        if (this._call) {
+          this._call._emit("transcript", "user", msg["transcript"] ?? "");
+        }
         break;
       }
       case "response.audio_transcript.done": {
@@ -2006,128 +2353,121 @@ var OpenAIRealtime = class {
     }
   }
   _handleAudioDelta(msg) {
-    if (this._responseStartTs === null) {
-      this._responseStartTs = Date.now();
-      this._sentAudioChunks = 0;
-    }
-    if (msg["item_id"]) {
-      this._lastAssistantItem = msg["item_id"];
+    if (this._playback === null) {
+      this._playback = {
+        itemId: msg["item_id"] || "",
+        startTs: this._latestMediaTs || Date.now(),
+        sentChunks: 0,
+        generating: true,
+        audioRemainder: Buffer.alloc(0)
+      };
+    } else if (msg["item_id"]) {
+      this._playback.itemId = msg["item_id"];
     }
+    const pb = this._playback;
     const ulaw = Buffer.from(msg["delta"], "base64");
     if (this._recorder) {
       this._recorder.writeOutbound(ulawToPcm16(ulaw));
     }
-    const combined = Buffer.concat([this._audioRemainder, ulaw]);
+    const combined = Buffer.concat([pb.audioRemainder, ulaw]);
     const chunkSize = 160;
     const fullEnd = Math.floor(combined.length / chunkSize) * chunkSize;
     for (let off = 0; off < fullEnd; off += chunkSize) {
       if (this._call) {
         this._call.sendAudio(combined.subarray(off, off + chunkSize));
       }
-      this._sentAudioChunks++;
+      pb.sentChunks++;
     }
-    this._audioRemainder = combined.subarray(fullEnd);
+    pb.audioRemainder = combined.subarray(fullEnd);
   }
   _handleTruncation() {
-    if (!this._lastAssistantItem || this._responseStartTs === null) {
-      return;
+    for (const [, controller] of this._pendingToolCalls) {
+      controller.abort();
     }
-    const audioEndMs = Math.max(0, this._sentAudioChunks * 20);
-    this._send({
-      type: "conversation.item.truncate",
-      item_id: this._lastAssistantItem,
-      content_index: 0,
-      audio_end_ms: audioEndMs
-    });
+    this._pendingToolCalls.clear();
     if (this._call) {
       this._call.clearAudio();
     }
-    this._lastAssistantItem = null;
-    this._responseStartTs = null;
-    this._sentAudioChunks = 0;
-    this._audioRemainder = Buffer.alloc(0);
+    const pb = this._playback;
+    if (pb === null) {
+      return;
+    }
+    const playedMs = Math.max(0, (this._latestMediaTs || Date.now()) - pb.startTs);
+    this._log.info(
+      "[Interrupt] item=%s played=%dms total=%dms",
+      pb.itemId,
+      playedMs,
+      pb.sentChunks * 20
+    );
+    this._playback = null;
   }
   async _handleToolCall(item) {
     const funcName = item["name"];
     const callId = item["call_id"];
     this._log.info("Tool call: %s", funcName);
-    if (funcName === "hang_up") {
-      if (this._call) {
-        await this._call.hangup();
-      }
-      return;
-    }
-    if (funcName === "collect_dtmf") {
-      if (this._call) {
-        let result2;
-        try {
-          const args = JSON.parse(item["arguments"] ?? "{}");
-          result2 = await this._call.collectDtmf({
-            maxDigits: args["max_digits"] ?? 4,
-            finishOnKey: args["finish_on_key"] ?? "#",
-            timeout: args["timeout"] ?? 5
+    const controller = new AbortController();
+    this._pendingToolCalls.set(callId, controller);
+    try {
+      if (BUILTIN_TOOL_NAMES.has(funcName) && this._call) {
+        const args = JSON.parse(item["arguments"] ?? "{}");
+        const result2 = await executeBuiltinTool(funcName, args, this._call);
+        if (result2 !== null) {
+          if (funcName === "hang_up") return;
+          if (controller.signal.aborted) return;
+          await this._waitForResponseDone();
+          this._send({
+            type: "conversation.item.create",
+            item: {
+              type: "function_call_output",
+              call_id: callId,
+              output: result2
+            }
           });
-        } catch (err) {
-          result2 = `Error: ${err}`;
+          this._send({ type: "response.create" });
+          return;
         }
-        await this._waitForResponseDone();
-        this._send({
-          type: "conversation.item.create",
-          item: {
-            type: "function_call_output",
-            call_id: callId,
-            output: result2 || "(\uD0C0\uC784\uC544\uC6C3 - \uC785\uB825 \uC5C6\uC74C)"
-          }
-        });
-        this._send({ type: "response.create" });
       }
-      return;
-    }
-    if (funcName === "send_dtmf") {
-      if (this._call) {
-        let result2;
-        try {
-          const args = JSON.parse(item["arguments"] ?? "{}");
-          await this._call.sendDtmfSequence(args["digits"] ?? "");
-          result2 = "sent";
-        } catch (err) {
-          result2 = `Error: ${err}`;
-        }
+      if (!this._tools || !this._tools.has(funcName)) {
+        this._log.error("Unknown tool: %s", funcName);
         await this._waitForResponseDone();
         this._send({
           type: "conversation.item.create",
           item: {
             type: "function_call_output",
             call_id: callId,
-            output: result2
+            output: JSON.stringify({ error: `Unknown tool: ${funcName}` })
           }
         });
         this._send({ type: "response.create" });
+        return;
       }
-      return;
-    }
-    if (!this._tools || !this._tools.has(funcName)) {
-      this._log.error("Unknown tool: %s", funcName);
-      return;
-    }
-    let result;
-    try {
-      const args = JSON.parse(item["arguments"] ?? "{}");
-      result = await this._tools.call(funcName, args);
-    } catch (err) {
-      this._log.error({ err }, "Tool call failed: %s", funcName);
-      result = `Error: ${err}`;
-    }
-    await this._waitForResponseDone();
-    this._send({
-      type: "conversation.item.create",
-      item: {
-        type: "function_call_output",
-        call_id: callId,
-        output: typeof result === "string" ? result : JSON.stringify(result)
+      let result;
+      try {
+        const args = JSON.parse(item["arguments"] ?? "{}");
+        result = await this._tools.call(funcName, args);
+      } catch (err) {
+        this._log.error({ err }, "Tool call failed: %s", funcName);
+        result = `Error: ${err}`;
       }
-    });
-    this._send({ type: "response.create" });
+      if (controller.signal.aborted) {
+        this._log.info("Tool call cancelled (user interrupted): %s", funcName);
+        return;
+      }
+      const resultStr = typeof result === "string" ? result : JSON.stringify(result);
+      await this._waitForResponseDone();
+      this._send({
+        type: "conversation.item.create",
+        item: {
+          type: "function_call_output",
+          call_id: callId,
+          output: resultStr
+        }
+      });
+      this._log.info("[ToolResult] %s call_id=%s len=%d", funcName, callId, resultStr.length);
+      this._send({ type: "response.create" });
+    } finally {
+      this._pendingToolCalls.delete(callId);
+    }
   }
   _waitForResponseDone() {
     if (!this._responseInProgress) return Promise.resolve();
@@ -2142,39 +2482,7 @@ var OpenAIRealtime = class {
   }
 };
 
-// src/agent/pipeline/gemini-realtime.ts
-var HANG_UP_TOOL2 = {
-  name: "hang_up",
-  description: "End the phone call. Use when the conversation is finished or the caller says goodbye.",
-  parameters: { type: "object", properties: {} }
-};
-var COLLECT_DTMF_TOOL2 = {
-  name: "collect_dtmf",
-  description: "\uC0AC\uC6A9\uC790\uB85C\uBD80\uD130 DTMF(\uC804\uD654 \uD0A4\uD328\uB4DC) \uC785\uB825\uC744 \uC218\uC9D1\uD569\uB2C8\uB2E4. \uBC18\uB4DC\uC2DC \uC0AC\uC6A9\uC790\uC5D0\uAC8C \uBB34\uC5C7\uC744 \uC785\uB825\uD574\uC57C \uD558\uB294\uC9C0 \uC548\uB0B4\uD55C \uD6C4 \uD638\uCD9C\uD558\uC138\uC694.",
-  parameters: {
-    type: "object",
-    properties: {
-      max_digits: { type: "integer", description: "\uC218\uC9D1\uD560 \uCD5C\uB300 \uC790\uB9BF\uC218" },
-      finish_on_key: { type: "string", description: "\uC785\uB825 \uC885\uB8CC \uD0A4 (\uAE30\uBCF8: #)" },
-      timeout: { type: "integer", description: "\uC785\uB825 \uB300\uAE30 \uC2DC\uAC04(\uCD08, \uAE30\uBCF8: 5)" }
-    },
-    required: ["max_digits"]
-  }
-};
-var SEND_DTMF_TOOL2 = {
-  name: "send_dtmf",
-  description: "DTMF \uC2E0\uD638\uB97C \uC804\uC1A1\uD569\uB2C8\uB2E4. ARS \uBA54\uB274 \uD0D0\uC0C9\uC774\uB098 \uB0B4\uC120\uBC88\uD638 \uC785\uB825 \uC2DC \uC0AC\uC6A9\uD569\uB2C8\uB2E4.",
-  parameters: {
-    type: "object",
-    properties: {
-      digits: {
-        type: "string",
-        description: "\uC804\uC1A1\uD560 \uBC88\uD638 (0-9, *, #). 'w'\uB294 500ms \uB300\uAE30, 'W'\uB294 1000ms \uB300\uAE30."
-      }
-    },
-    required: ["digits"]
-  }
-};
+// src/agent/pipeline/realtime/gemini-realtime.ts
 function resolveRef(ref, defs) {
   const parts = ref.replace(/^#\//, "").split("/");
   let result = defs;
@@ -2384,9 +2692,7 @@ var GeminiRealtime = class {
         t.function.parameters ?? { type: "object", properties: {} }
       )
     })) : [];
-    if (!this._builtinTools || this._builtinTools.has("hang_up" /* HANG_UP */)) toolDefs.push(HANG_UP_TOOL2);
-    if (!this._builtinTools || this._builtinTools.has("collect_dtmf" /* COLLECT_DTMF */)) toolDefs.push(COLLECT_DTMF_TOOL2);
-    if (!this._builtinTools || this._builtinTools.has("send_dtmf" /* SEND_DTMF */)) toolDefs.push(SEND_DTMF_TOOL2);
+    toolDefs.push(...getBuiltinToolSchemas(this._builtinTools, "gemini"));
     return toolDefs;
   }
   _handleMessage(msg) {
@@ -2474,51 +2780,17 @@ var GeminiRealtime = class {
       const fcId = fc.id ?? "";
       const args = fc.args ?? {};
       this._log.info({ tool: name, args }, "Tool call: %s", name);
-      if (name === "hang_up") {
-        this._log.info("hang_up: ending call");
-        if (this._call) {
-          await this._call.hangup();
-        }
-        return;
-      }
-      if (name === "collect_dtmf") {
-        if (this._call) {
-          let result;
-          try {
-            this._log.info({ maxDigits: args["max_digits"] ?? 4, timeout: args["timeout"] ?? 5 }, "collect_dtmf: waiting for digits");
-            result = await this._call.collectDtmf({
-              maxDigits: args["max_digits"] ?? 4,
-              finishOnKey: args["finish_on_key"] ?? "#",
-              timeout: args["timeout"] ?? 5
-            });
-            this._log.info("DTMF collected: %s", result || "(empty)");
-          } catch (err) {
-            this._log.error({ err }, "collect_dtmf error");
-            result = `Error: ${err}`;
-          }
-          responses.push({
-            id: fcId,
-            name,
-            response: { result: result || "(\uD0C0\uC784\uC544\uC6C3 - \uC785\uB825 \uC5C6\uC74C)" }
-          });
-        }
-        continue;
-      }
-      if (name === "send_dtmf") {
-        if (this._call) {
-          let result;
-          try {
-            this._log.info('send_dtmf: digits="%s"', args["digits"] ?? "");
-            await this._call.sendDtmfSequence(args["digits"] ?? "");
-            result = "sent";
-            this._log.info("send_dtmf: sent");
-          } catch (err) {
-            this._log.error({ err }, "send_dtmf error");
-            result = `Error: ${err}`;
+      if (BUILTIN_TOOL_NAMES.has(name) && this._call) {
+        const result = await executeBuiltinTool(name, args, this._call);
+        if (result !== null) {
+          if (name === "hang_up") {
+            this._log.info("hang_up: ending call");
+            return;
           }
+          this._log.info("Builtin tool result: %s -> %s", name, result);
           responses.push({ id: fcId, name, response: { result } });
+          continue;
         }
-        continue;
       }
       if (!this._tools || !this._tools.has(name)) {
         this._log.error("Unknown tool: %s", name);
@@ -2553,306 +2825,7 @@ var GeminiRealtime = class {
   }
 };
 
-// src/agent/pipeline/pipeline-session.ts
-var COLLECT_DTMF_TOOL3 = {
-  function: {
-    description: "\uC0AC\uC6A9\uC790\uB85C\uBD80\uD130 DTMF(\uC804\uD654 \uD0A4\uD328\uB4DC) \uC785\uB825\uC744 \uC218\uC9D1\uD569\uB2C8\uB2E4.",
-    parameters: {
-      properties: {
-        max_digits: { type: "integer", description: "\uC218\uC9D1\uD560 \uCD5C\uB300 \uC790\uB9BF\uC218" },
-        finish_on_key: { type: "string", description: "\uC785\uB825 \uC885\uB8CC \uD0A4 (\uAE30\uBCF8: #)" },
-        timeout: { type: "integer", description: "\uC785\uB825 \uB300\uAE30 \uC2DC\uAC04(\uCD08, \uAE30\uBCF8: 5)" }
-      },
-      required: ["max_digits"]
-    }
-  }
-};
-var SEND_DTMF_TOOL3 = {
-  function: {
-    description: "DTMF \uC2E0\uD638\uB97C \uC804\uC1A1\uD569\uB2C8\uB2E4. ARS \uBA54\uB274 \uD0D0\uC0C9\uC774\uB098 \uB0B4\uC120\uBC88\uD638 \uC785\uB825 \uC2DC \uC0AC\uC6A9\uD569\uB2C8\uB2E4.",
-    parameters: {
-      properties: {
-        digits: { type: "string", description: "\uC804\uC1A1\uD560 \uBC88\uD638 (0-9, *, #). 'w'\uB294 500ms \uB300\uAE30, 'W'\uB294 1000ms \uB300\uAE30." }
-      },
-      required: ["digits"]
-    }
-  }
-};
-var PipelineSession = class {
-  _stt;
-  _llm;
-  _tts;
-  _systemPrompt;
-  _greeting;
-  _language;
-  _temperature;
-  _maxTokens;
-  _sampleRate;
-  _interruptOnSpeech;
-  _callSession = null;
-  _tools = null;
-  _recorder = null;
-  _conversation = [];
-  _audioBuffer = [];
-  _running = false;
-  _speaking = false;
-  _builtinTools = null;
-  _log = NOOP_LOGGER;
-  constructor(options) {
-    this._stt = options.stt;
-    this._llm = options.llm;
-    this._tts = options.tts;
-    this._systemPrompt = options.systemPrompt;
-    this._greeting = options.greeting ?? true;
-    this._language = options.language ?? "ko";
-    this._temperature = options.temperature;
-    this._maxTokens = options.maxTokens;
-    this._sampleRate = options.sampleRate ?? 8e3;
-    this._interruptOnSpeech = options.interruptOnSpeech ?? true;
-    if (options.toolRegistry) this._tools = options.toolRegistry;
-    if (options.recorder) this._recorder = options.recorder;
-  }
-  setToolRegistry(registry) {
-    this._tools = registry;
-  }
-  setRecorder(recorder) {
-    this._recorder = recorder;
-  }
-  setBuiltinTools(tools) {
-    this._builtinTools = tools;
-  }
-  setLogger(logger) {
-    this._log = logger;
-    if ("setLogger" in this._stt && typeof this._stt.setLogger === "function") {
-      this._stt.setLogger(logger);
-    }
-    if ("setLogger" in this._tts && typeof this._tts.setLogger === "function") {
-      this._tts.setLogger(logger);
-    }
-  }
-  async start(callSession, tools) {
-    this._callSession = callSession;
-    this._tools = tools ?? null;
-    this._running = true;
-    this._log.info("PipelineSession started");
-    this._conversation = [];
-    if (this._systemPrompt) {
-      this._conversation.push({
-        role: "system",
-        content: this._systemPrompt
-      });
-    }
-    if (this._greeting) {
-      this._generateGreeting().catch((err) => {
-        this._log.error({ err }, "Greeting error");
-      });
-    }
-    this._runSttLoop().catch((err) => {
-      this._log.error({ err }, "STT loop error");
-    });
-  }
-  feedAudio(audio) {
-    if (this._running) {
-      this._audioBuffer.push(audio);
-    }
-  }
-  async feedDtmf(digits) {
-    this._conversation.push({
-      role: "user",
-      content: `[DTMF \uC785\uB825: ${digits}]`
-    });
-    await this._respond();
-  }
-  async stop() {
-    this._running = false;
-    this._log.info("PipelineSession stopped");
-    this._audioBuffer = [];
-  }
-  async _runSttLoop() {
-    const audioStream = this._createAudioStream();
-    for await (const event of this._stt.transcribe(audioStream, {
-      sampleRate: this._sampleRate
-    })) {
-      if (!this._running) break;
-      if (event.type === "interim" && this._speaking && this._interruptOnSpeech) {
-        this._speaking = false;
-        if (this._callSession) {
-          this._callSession.clearAudio();
-        }
-        this._log.info('Barge-in: "%s"', event.transcript.substring(0, 30));
-      }
-      if (event.type === "final" && event.transcript.trim()) {
-        this._log.info("STT: %s", event.transcript);
-        await this._handleUserSpeech(event.transcript);
-      }
-    }
-  }
-  async *_createAudioStream() {
-    while (this._running) {
-      if (this._audioBuffer.length > 0) {
-        const ulaw = this._audioBuffer.shift();
-        const pcm8k = ulawToPcm16(ulaw);
-        const pcm16k = resamplePcm16(pcm8k, 8e3, 16e3);
-        yield pcm16k;
-      } else {
-        await new Promise((resolve) => setTimeout(resolve, 20));
-      }
-    }
-  }
-  async _generateGreeting() {
-    await new Promise((resolve) => setTimeout(resolve, 500));
-    await this._respond();
-  }
-  async _handleUserSpeech(transcript) {
-    this._conversation.push({ role: "user", content: transcript });
-    await this._respond();
-  }
-  _buildEffectiveTools() {
-    const includeCollectDtmf = !this._builtinTools || this._builtinTools.has("collect_dtmf" /* COLLECT_DTMF */);
-    const includeSendDtmf = !this._builtinTools || this._builtinTools.has("send_dtmf" /* SEND_DTMF */);
-    if (!includeCollectDtmf && !includeSendDtmf) return this._tools ?? void 0;
-    const base = this._tools ? this._tools.fork() : new ToolRegistry();
-    if (includeCollectDtmf) {
-      base.register({
-        name: "collect_dtmf",
-        description: COLLECT_DTMF_TOOL3.function.description,
-        parameters: COLLECT_DTMF_TOOL3.function.parameters.properties,
-        required: COLLECT_DTMF_TOOL3.function.parameters.required,
-        handler: async () => ""
-      });
-    }
-    if (includeSendDtmf) {
-      base.register({
-        name: "send_dtmf",
-        description: SEND_DTMF_TOOL3.function.description,
-        parameters: SEND_DTMF_TOOL3.function.parameters.properties,
-        required: SEND_DTMF_TOOL3.function.parameters.required,
-        handler: async () => ""
-      });
-    }
-    return base;
-  }
-  async _respond() {
-    let fullResponse = "";
-    const textChunks = [];
-    const effectiveTools = this._buildEffectiveTools();
-    const llmStream = this._llm.generate(this._conversation, {
-      tools: effectiveTools,
-      temperature: this._temperature,
-      maxTokens: this._maxTokens
-    });
-    for await (const chunk of llmStream) {
-      if (!this._running) break;
-      if (chunk.type === "text" && chunk.text) {
-        textChunks.push(chunk.text);
-        fullResponse += chunk.text;
-      } else if (chunk.type === "tool_call" && chunk.toolCall) {
-        await this._handleToolCall(chunk);
-      }
-    }
-    if (fullResponse.trim()) {
-      this._log.info("Assistant: %s", fullResponse.substring(0, 100));
-      this._conversation.push({ role: "assistant", content: fullResponse });
-      await this._synthesizeAndSend(fullResponse);
-    }
-  }
-  async _handleToolCall(chunk) {
-    if (!chunk.toolCall) return;
-    const { id, name, arguments: argsStr } = chunk.toolCall;
-    try {
-      const args = JSON.parse(argsStr);
-      if (name === "collect_dtmf" && this._callSession) {
-        let result2;
-        try {
-          result2 = await this._callSession.collectDtmf({
-            maxDigits: args["max_digits"] ?? 4,
-            finishOnKey: args["finish_on_key"] ?? "#",
-            timeout: args["timeout"] ?? 5
-          });
-        } catch (err) {
-          result2 = `Error: ${err}`;
-        }
-        this._conversation.push({ role: "tool", content: result2 || "(\uD0C0\uC784\uC544\uC6C3 - \uC785\uB825 \uC5C6\uC74C)", tool_call_id: id, name });
-        await this._respond();
-        return;
-      }
-      if (name === "send_dtmf" && this._callSession) {
-        let result2;
-        try {
-          await this._callSession.sendDtmfSequence(args["digits"] ?? "");
-          result2 = "sent";
-        } catch (err) {
-          result2 = `Error: ${err}`;
-        }
-        this._conversation.push({ role: "tool", content: result2, tool_call_id: id, name });
-        await this._respond();
-        return;
-      }
-      if (!this._tools) return;
-      const result = await this._tools.call(name, args);
-      this._conversation.push({
-        role: "assistant",
-        content: ""
-        // Tool call info stored in the message flow
-      });
-      this._conversation.push({
-        role: "tool",
-        content: typeof result === "string" ? result : JSON.stringify(result),
-        tool_call_id: id,
-        name
-      });
-      const effectiveTools = this._buildEffectiveTools();
-      let followUpText = "";
-      const followUpStream = this._llm.generate(this._conversation, {
-        tools: effectiveTools,
-        temperature: this._temperature,
-        maxTokens: this._maxTokens
-      });
-      for await (const followChunk of followUpStream) {
-        if (!this._running) break;
-        if (followChunk.type === "text" && followChunk.text) {
-          followUpText += followChunk.text;
-        }
-      }
-      if (followUpText.trim()) {
-        this._conversation.push({ role: "assistant", content: followUpText });
-        await this._synthesizeAndSend(followUpText);
-      }
-    } catch (err) {
-      this._log.error({ err }, "Tool call failed: %s", name);
-    }
-  }
-  async _synthesizeAndSend(text) {
-    if (!this._callSession || !this._running) return;
-    this._speaking = true;
-    try {
-      for await (const audioChunk of this._tts.synthesize(text, {
-        sampleRate: this._sampleRate
-      })) {
-        if (!this._running || !this._speaking) break;
-        if (this._recorder) {
-          const pcm8k2 = this._sampleRate !== 8e3 ? resamplePcm16(audioChunk, this._sampleRate, 8e3) : audioChunk;
-          this._recorder.writeOutbound(pcm8k2);
-        }
-        const pcm8k = resamplePcm16(audioChunk, this._sampleRate, 8e3);
-        const ulaw = pcm16ToUlaw(pcm8k);
-        for (let off = 0; off < ulaw.length; off += 160) {
-          let chunk = ulaw.subarray(off, off + 160);
-          if (chunk.length < 160) {
-            chunk = Buffer.concat([chunk, Buffer.alloc(160 - chunk.length, 255)]);
-          }
-          this._callSession.sendAudio(chunk);
-        }
-      }
-    } catch (err) {
-      this._log.error({ err }, "TTS error");
-    } finally {
-      this._speaking = false;
-    }
-  }
-};
-
-// src/agent/pipeline/deepgram-stt.ts
+// src/agent/pipeline/stt/deepgram-stt.ts
 var DeepgramSTT = class {
   _options;
   _log = NOOP_LOGGER;
@@ -2973,7 +2946,7 @@ var DeepgramSTT = class {
   }
 };
 
-// src/agent/pipeline/elevenlabs-tts.ts
+// src/agent/pipeline/tts/elevenlabs-tts.ts
 var ElevenLabsTTS = class {
   _options;
   _log = NOOP_LOGGER;
@@ -3126,7 +3099,7 @@ var ElevenLabsTTS = class {
   }
 };
 
-// src/agent/pipeline/openai-llm.ts
+// src/agent/pipeline/llm/openai-llm.ts
 var OpenAILLM = class {
   _options;
   constructor(options = {}) {
@@ -3212,7 +3185,7 @@ var OpenAILLM = class {
   }
 };
 
-// src/agent/pipeline/anthropic-llm.ts
+// src/agent/pipeline/llm/anthropic-llm.ts
 var AnthropicLLM = class {
   _options;
   constructor(options = {}) {
@@ -3317,7 +3290,7 @@ var AnthropicLLM = class {
   }
 };
 
-// src/agent/pipeline/gemini-llm.ts
+// src/agent/pipeline/llm/gemini-llm.ts
 var GeminiLLM = class {
   _options;
   constructor(options = {}) {
@@ -3401,7 +3374,7 @@ var GeminiLLM = class {
   }
 };
 
-// src/agent/pipeline/openai-compat-llm.ts
+// src/agent/pipeline/llm/openai-compat-llm.ts
 var OpenAICompatLLM = class {
   _options;
   constructor(options) {
@@ -3484,7 +3457,7 @@ var OpenAICompatLLM = class {
   }
 };
 
-// src/agent/pipeline/ollama-llm.ts
+// src/agent/pipeline/llm/ollama-llm.ts
 var OllamaLLM = class {
   _inner;
   constructor(options = {}) {
@@ -3502,7 +3475,7 @@ var OllamaLLM = class {
   }
 };
 
-// src/agent/pipeline/mistral-llm.ts
+// src/agent/pipeline/llm/mistral-llm.ts
 var MistralLLM = class {
   _inner;
   constructor(options = {}) {
@@ -3519,7 +3492,7 @@ var MistralLLM = class {
   }
 };
 
-// src/agent/pipeline/groq-llm.ts
+// src/agent/pipeline/llm/groq-llm.ts
 var GroqLLM = class {
   _inner;
   constructor(options = {}) {
@@ -3536,7 +3509,7 @@ var GroqLLM = class {
   }
 };
 
-// src/agent/pipeline/perplexity-llm.ts
+// src/agent/pipeline/llm/perplexity-llm.ts
 var PerplexityLLM = class {
   _inner;
   constructor(options = {}) {
@@ -3553,7 +3526,7 @@ var PerplexityLLM = class {
   }
 };
 
-// src/agent/pipeline/together-llm.ts
+// src/agent/pipeline/llm/together-llm.ts
 var TogetherLLM = class {
   _inner;
   constructor(options = {}) {
@@ -3570,7 +3543,7 @@ var TogetherLLM = class {
   }
 };
 
-// src/agent/pipeline/fireworks-llm.ts
+// src/agent/pipeline/llm/fireworks-llm.ts
 var FireworksLLM = class {
   _inner;
   constructor(options = {}) {
@@ -3587,7 +3560,7 @@ var FireworksLLM = class {
   }
 };
 
-// src/agent/pipeline/deepseek-llm.ts
+// src/agent/pipeline/llm/deepseek-llm.ts
 var DeepSeekLLM = class {
   _inner;
   constructor(options = {}) {
@@ -3604,7 +3577,7 @@ var DeepSeekLLM = class {
   }
 };
 
-// src/agent/pipeline/xai-llm.ts
+// src/agent/pipeline/llm/xai-llm.ts
 var XaiLLM = class {
   _inner;
   constructor(options = {}) {
@@ -3640,6 +3613,6 @@ function mcpServerHTTP(options) {
   };
 }
 
-export { AnthropicLLM, AudioRecorder, BuiltinTool, CallSession, ClawOpsAgent, DECODE_TABLE, DeepSeekLLM, DeepgramSTT, ElevenLabsTTS, FireworksLLM, GeminiLLM, GeminiRealtime, GroqLLM, MCPClient, MistralLLM, OllamaLLM, OpenAICompatLLM, OpenAILLM, OpenAIRealtime, PerplexityLLM, PipelineSession, TogetherLLM, ToolRegistry, XaiLLM, createAgentLogger, createPipelineLogger, functionTool, getTracingConfig, mcpServerHTTP, mcpServerStdio, pcm16ToUlaw, resamplePcm16, resetTracingConfig, setTracingConfig, ulawToPcm16, zodToToolParams };
+export { AnthropicLLM, AudioRecorder, BUILTIN_TOOL_NAMES, BuiltinTool, CallSession, ClawOpsAgent, DECODE_TABLE, DeepSeekLLM, DeepgramSTT, ElevenLabsTTS, FireworksLLM, GeminiLLM, GeminiRealtime, GroqLLM, MCPClient, MistralLLM, OllamaLLM, OpenAICompatLLM, OpenAILLM, OpenAIRealtime, PerplexityLLM, PipelineSession, TogetherLLM, ToolRegistry, XaiLLM, createAgentLogger, createPipelineLogger, executeBuiltinTool, functionTool, getBuiltinToolSchemas, getTracingConfig, isBuiltinTool, mcpServerHTTP, mcpServerStdio, pcm16ToUlaw, resamplePcm16, resetTracingConfig, setTracingConfig, ulawToPcm16, zodToToolParams };
 //# sourceMappingURL=index.js.map
 //# sourceMappingURL=index.js.map