@teamlearners/clawops 0.5.5 → 0.6.0

package/dist/agent/index.js

@@ -1,7 +1,8 @@
-import { DEFAULT_BASE_URL, AgentError, AgentConnectionError } from '../chunk-7S2OEBS6.js';
+import { DEFAULT_BASE_URL, AgentError, AgentConnectionError, VERSION } from '../chunk-2B74QZZJ.js';
 import pino from 'pino';
 import * as fs from 'fs';
 import * as path from 'path';
+import os from 'os';
 
 // src/agent/audio.ts
 var _BIAS = 132;
@@ -972,6 +973,36 @@ var AudioRecorder = class {
     }
   }
 };
+var MAX_ERRORS = 20;
+var MAX_ERROR_MESSAGE_LENGTH = 200;
+function getSdkInfo() {
+  return {
+    name: "clawops-node",
+    version: VERSION,
+    runtime: `node/${process.versions.node}`,
+    os: `${process.platform}/${os.arch()}`
+  };
+}
+function createCallMetrics() {
+  return {
+    firstResponseMs: null,
+    turnCount: 0,
+    toolCallCount: 0,
+    toolErrorCount: 0,
+    bargeInCount: 0,
+    endReason: null,
+    errors: []
+  };
+}
+function addMetricError(metrics, err) {
+  metrics.toolErrorCount++;
+  if (metrics.errors.length < MAX_ERRORS) {
+    metrics.errors.push({
+      type: err.name || "Error",
+      message: (err.message || "").slice(0, MAX_ERROR_MESSAGE_LENGTH)
+    });
+  }
+}
 
 // src/agent/session.ts
 var CallSession = class {
@@ -997,6 +1028,8 @@ var CallSession = class {
   _handlers = /* @__PURE__ */ new Map();
   _endedPromise;
   _resolveEnded;
+  _metrics = createCallMetrics();
+  _firstResponseSent = false;
   constructor(options) {
     this.callId = options.callId;
     this.fromNumber = options.fromNumber;
@@ -1019,6 +1052,30 @@ var CallSession = class {
   get duration() {
     return (Date.now() - this.startTime.getTime()) / 1e3;
   }
+  get metrics() {
+    return this._metrics;
+  }
+  recordFirstResponse() {
+    if (!this._firstResponseSent) {
+      this._firstResponseSent = true;
+      this._metrics.firstResponseMs = Date.now() - this.startTime.getTime();
+    }
+  }
+  recordTurn() {
+    this._metrics.turnCount++;
+  }
+  recordToolCall() {
+    this._metrics.toolCallCount++;
+  }
+  recordToolError(err) {
+    addMetricError(this._metrics, err);
+  }
+  recordBargeIn() {
+    this._metrics.bargeInCount++;
+  }
+  recordEndReason(reason) {
+    this._metrics.endReason = reason;
+  }
   /** Bind transport functions (called internally by the agent). */
   _bindTransport(send, clear, hangup, sendDtmf, isConnected) {
     this._sendAudioFn = send;
@@ -1459,6 +1516,7 @@ 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.");
     }
@@ -1482,6 +1540,10 @@ var ClawOpsAgent = class {
     try {
       await this._controlWs.connect();
       await this._controlWs.waitConnected();
+      try {
+        this._controlWs.send({ event: "agent.hello", sdk: getSdkInfo() });
+      } catch {
+      }
     } catch (err) {
       throw new AgentConnectionError(
         `Failed to connect to ClawOps: ${err instanceof Error ? err.message : String(err)}`
@@ -1695,9 +1757,11 @@ var ClawOpsAgent = class {
         session._bindTransport(
           (audio) => {
             mediaWs.sendAudio(audio.toString("base64"));
+            session.recordFirstResponse();
           },
           () => {
             mediaWs.sendClear();
+            session.recordBargeIn();
           },
           async () => {
             await mediaWs.flush();
@@ -1748,10 +1812,22 @@ var ClawOpsAgent = class {
           await mediaWs.connect(mediaWsUrl, this._apiKey);
           this._log.info("Media stream started: %s", session.callId);
           await sessionHandler.start(session, sessionTools);
+          const telemetry = sessionHandler.getTelemetry?.() ?? null;
+          if (telemetry) {
+            telemetry.toolCount = sessionTools?.size ?? 0;
+            telemetry.mcpServerCount = this._mcpServers?.length ?? 0;
+            telemetry.builtinTools = this._builtinTools ? [...this._builtinTools].map((t) => t.toString()) : [];
+            telemetry.recordingEnabled = this._recording;
+            try {
+              this._controlWs.send({ event: "call.telemetry", callId: session.callId, telemetry });
+            } catch {
+            }
+          }
           await session.wait();
           await sessionHandler.stop();
         } catch (err) {
           this._log.error({ err }, "Call session error: %s", session.callId);
+          session.recordEndReason("error");
         } finally {
           if (mcpClients.length > 0) {
             sessionTools.clearMcpTools();
@@ -1763,6 +1839,13 @@ var ClawOpsAgent = class {
           if (recorder) {
             recorder.stop();
           }
+          if (!session.metrics.endReason) {
+            session.recordEndReason(session.status === "ended" ? "user_hangup" : "agent_hangup");
+          }
+          try {
+            this._controlWs?.send({ event: "call.metrics", callId: session.callId, metrics: session.metrics });
+          } catch {
+          }
           session._emit("call_end");
           session._markEnded();
           this._activeSessions.delete(session.callId);
@@ -1773,20 +1856,13 @@ 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: {
@@ -1799,8 +1875,7 @@ 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: {
@@ -1808,492 +1883,419 @@ 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;
+  }
+  getTelemetry() {
+    const llm = this._llm;
+    const stt = this._stt;
+    const tts = this._tts;
+    return {
+      sessionType: "pipeline",
+      llm: llm.provider && llm.model ? { provider: llm.provider, model: llm.model } : null,
+      stt: stt.provider && stt.model ? { provider: stt.provider, model: stt.model } : null,
+      tts: tts.provider && tts.model ? { provider: tts.provider, model: tts.model } : null,
+      voice: tts.voiceId ?? null,
+      language: this._language,
+      greetingEnabled: this._greeting,
+      recordingEnabled: !!this._recorder,
+      toolCount: this._tools?.size ?? 0,
+      mcpServerCount: 0,
+      builtinTools: []
+    };
+  }
+  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")
-      });
+    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._closed = true;
-    if (this._ws) {
-      this._ws.close();
-      this._ws = null;
-    }
+    this._running = false;
+    this._log.info("PipelineSession stopped");
+    this._audioBuffer = [];
   }
-  _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 _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);
+      }
+    }
   }
-  _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);
-        }
-        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": {
-        if (this._call) {
-          this._call._emit("transcript", "assistant", msg["transcript"] ?? "");
-        }
-        break;
-      }
-      case "response.created": {
-        this._responseInProgress = true;
-        break;
-      }
-      case "response.done": {
-        this._responseInProgress = false;
-        if (this._onResponseDone) {
-          const cb = this._onResponseDone;
-          this._onResponseDone = null;
-          cb();
-        }
-        break;
-      }
-      case "error": {
-        this._log.error({ apiError: msg["error"] }, "OpenAI error");
-        break;
+  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));
       }
     }
   }
-  _handleAudioDelta(msg) {
-    if (this._responseStartTs === null) {
-      this._responseStartTs = Date.now();
-      this._sentAudioChunks = 0;
-    }
-    if (msg["item_id"]) {
-      this._lastAssistantItem = msg["item_id"];
-    }
-    const ulaw = Buffer.from(msg["delta"], "base64");
-    if (this._recorder) {
-      this._recorder.writeOutbound(ulawToPcm16(ulaw));
-    }
-    const combined = Buffer.concat([this._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++;
-    }
-    this._audioRemainder = combined.subarray(fullEnd);
+  async _generateGreeting() {
+    await new Promise((resolve) => setTimeout(resolve, 500));
+    await this._respond();
   }
-  _handleTruncation() {
-    if (!this._lastAssistantItem || this._responseStartTs === null) {
-      return;
-    }
-    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
+  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 (this._call) {
-      this._call.clearAudio();
+    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 () => ""
+      });
     }
-    this._lastAssistantItem = null;
-    this._responseStartTs = null;
-    this._sentAudioChunks = 0;
-    this._audioRemainder = Buffer.alloc(0);
+    return base;
   }
-  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();
+  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);
       }
-      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
-          });
-        } catch (err) {
-          result2 = `Error: ${err}`;
+    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;
         }
-        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) return;
+      this._callSession?.recordToolCall();
+      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;
         }
-        await this._waitForResponseDone();
-        this._send({
-          type: "conversation.item.create",
-          item: {
-            type: "function_call_output",
-            call_id: callId,
-            output: result2
-          }
-        });
-        this._send({ type: "response.create" });
       }
-      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)
+      if (followUpText.trim()) {
+        this._conversation.push({ role: "assistant", content: followUpText });
+        await this._synthesizeAndSend(followUpText);
       }
-    });
-    this._send({ type: "response.create" });
-  }
-  _waitForResponseDone() {
-    if (!this._responseInProgress) return Promise.resolve();
-    return new Promise((resolve) => {
-      this._onResponseDone = resolve;
-    });
-  }
-  _send(data) {
-    if (this._ws && this._ws.readyState === 1 && !this._closed) {
-      this._ws.send(JSON.stringify(data));
-    }
-  }
-};
-
-// 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."
+    } catch (err) {
+      this._log.error({ err }, "Tool call failed: %s", name);
+      if (err instanceof Error) {
+        this._callSession?.recordToolError(err);
       }
-    },
-    required: ["digits"]
-  }
-};
-function resolveRef(ref, defs) {
-  const parts = ref.replace(/^#\//, "").split("/");
-  let result = defs;
-  for (const part of parts) {
-    if (result && typeof result === "object" && !Array.isArray(result)) {
-      result = result[part];
-    } else {
-      return {};
     }
   }
-  return typeof result === "object" && result !== null && !Array.isArray(result) ? result : {};
-}
-function sanitizeSchemaForGemini(schema, defs, depth = 0) {
-  if (depth > 15) return { type: "object", properties: {} };
-  if (!schema || typeof schema !== "object") return { type: "object", properties: {} };
-  if (defs === void 0) {
-    defs = schema["$defs"] ?? schema["definitions"] ?? {};
-  }
-  if (typeof schema["$ref"] === "string") {
-    const resolved = resolveRef(schema["$ref"], { $defs: defs, definitions: defs });
-    if (resolved && Object.keys(resolved).length > 0) {
-      return sanitizeSchemaForGemini(resolved, defs, depth + 1);
-    }
-    return { type: "object", properties: {} };
-  }
-  for (const comboKey of ["oneOf", "anyOf", "allOf"]) {
-    const variants = schema[comboKey];
-    if (Array.isArray(variants) && variants.length > 0) {
-      for (const v of variants) {
-        if (v && typeof v === "object") {
-          const resolved = sanitizeSchemaForGemini(v, defs, depth + 1);
-          if (resolved["type"] === "object" && resolved["properties"]) {
-            return resolved;
+  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);
         }
       }
-      const first = variants[0];
-      if (first && typeof first === "object") {
-        return sanitizeSchemaForGemini(first, defs, depth + 1);
-      }
-    }
-  }
-  const result = {};
-  let schemaType = schema["type"];
-  if (Array.isArray(schemaType)) {
-    const nonNull = schemaType.filter((t) => t !== "null");
-    schemaType = nonNull[0] ?? "string";
-  }
-  if (schemaType) result["type"] = schemaType;
-  if (schema["description"]) result["description"] = schema["description"];
-  if (schema["enum"]) result["enum"] = schema["enum"];
-  if (schema["required"]) result["required"] = schema["required"];
-  if (schema["properties"] && typeof schema["properties"] === "object") {
-    const props = {};
-    for (const [key, val] of Object.entries(schema["properties"])) {
-      if (val && typeof val === "object") {
-        props[key] = sanitizeSchemaForGemini(val, defs, depth + 1);
-      }
+    } catch (err) {
+      this._log.error({ err }, "TTS error");
+    } finally {
+      this._speaking = false;
     }
-    result["properties"] = props;
-  }
-  if (schema["items"] && typeof schema["items"] === "object" && !Array.isArray(schema["items"])) {
-    result["items"] = sanitizeSchemaForGemini(
-      schema["items"],
-      defs,
-      depth + 1
-    );
   }
-  if (!result["type"] && result["properties"]) result["type"] = "object";
-  if (result["type"] === "object" && !result["properties"]) result["properties"] = {};
-  return result;
-}
-var GeminiRealtime = class {
+};
+
+// 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;
-  // eslint-disable-next-line @typescript-eslint/no-explicit-any
-  _session = null;
+  _log = NOOP_LOGGER;
+  _builtinTools = null;
+  setLogger(logger) {
+    this._log = logger;
+  }
+  setBuiltinTools(tools) {
+    this._builtinTools = tools;
+  }
+  getTelemetry() {
+    return {
+      sessionType: "openai_realtime",
+      llm: { provider: "openai", model: this._model },
+      stt: null,
+      tts: null,
+      voice: this._voice,
+      language: this._language,
+      greetingEnabled: this._greeting,
+      recordingEnabled: !!this._recorder,
+      toolCount: this._tools?.size ?? 0,
+      mcpServerCount: 0,
+      builtinTools: []
+    };
+  }
+  _ws = null;
   _call = null;
   _tools = null;
   _recorder = null;
   _closed = false;
-  _sentAudioChunks = 0;
-  _audioRemainder = Buffer.alloc(0);
-  _builtinTools = null;
-  _toolCallInProgress = false;
-  _log = NOOP_LOGGER;
+  // 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["GOOGLE_API_KEY"] ?? "";
+    this._apiKey = options.apiKey ?? process.env["OPENAI_API_KEY"] ?? "";
     this._systemPrompt = options.systemPrompt ?? "";
-    this._model = options.model ?? "gemini-2.5-flash-native-audio-preview-12-2025";
-    this._voice = options.voice ?? "Kore";
+    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. */
@@ -2304,340 +2306,409 @@ var GeminiRealtime = class {
   setRecorder(recorder) {
     this._recorder = recorder;
   }
-  setBuiltinTools(tools) {
-    this._builtinTools = tools;
-  }
-  setLogger(logger) {
-    this._log = logger;
-  }
   async start(callSession, tools) {
     this._call = callSession;
     if (tools) this._tools = tools;
     this._closed = false;
-    this._sentAudioChunks = 0;
-    this._audioRemainder = Buffer.alloc(0);
+    this._playback = null;
+    this._latestMediaTs = 0;
     if (!this._apiKey) {
-      throw new Error("Google API key is required. Set GOOGLE_API_KEY or pass apiKey option.");
-    }
-    const { GoogleGenAI } = await import('@google/genai/node');
-    const client = new GoogleGenAI({ apiKey: this._apiKey });
-    const config = {
-      responseModalities: ["AUDIO"],
-      speechConfig: {
-        voiceConfig: {
-          prebuiltVoiceConfig: {
-            voiceName: this._voice
-          }
-        }
-      },
-      inputAudioTranscription: {},
-      outputAudioTranscription: {}
-    };
-    if (this._systemPrompt) {
-      config["systemInstruction"] = this._systemPrompt;
-    }
-    const toolSchemas = this._buildToolSchemas();
-    if (toolSchemas.length > 0) {
-      config["tools"] = [{ functionDeclarations: toolSchemas }];
+      throw new Error("OpenAI API key is required. Set OPENAI_API_KEY or pass apiKey option.");
     }
-    this._session = await client.live.connect({
-      model: this._model,
-      config,
-      callbacks: {
-        onmessage: (msg) => this._handleMessage(msg),
-        onerror: (err) => {
-          this._log.error({ err }, "Gemini SDK error");
-        },
-        onclose: (ev) => {
-          this._log.info({ code: ev?.code ?? "unknown" }, "Gemini connection closed");
-          this._closed = true;
-        }
+    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"
       }
     });
-    if (this._greeting) {
-      this._session.sendClientContent({
-        turns: [
-          {
-            role: "user",
-            parts: [{ text: "\uC778\uC0AC\uD574 \uC8FC\uC138\uC694." }]
-          }
-        ],
-        turnComplete: true
+    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();
       });
-    }
-  }
-  feedAudio(audio) {
-    if (this._session && !this._closed && !this._toolCallInProgress) {
-      const pcm8k = ulawToPcm16(audio);
-      if (this._recorder) {
-        this._recorder.writeInbound(pcm8k);
-      }
-      const pcm16k = resamplePcm16(pcm8k, 8e3, 16e3);
-      this._session.sendRealtimeInput({
-        audio: {
-          data: Buffer.from(pcm16k).toString("base64"),
-          mimeType: "audio/pcm;rate=16000"
+      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) {
-    if (this._session) {
-      this._session.sendClientContent({
-        turns: [{ role: "user", parts: [{ text: `[DTMF \uC785\uB825: ${digits}]` }] }],
-        turnComplete: true
+    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;
-    if (this._session) {
-      try {
-        this._session.close();
-      } catch {
-      }
-      this._session = null;
+    for (const [, controller] of this._pendingToolCalls) {
+      controller.abort();
+    }
+    this._pendingToolCalls.clear();
+    if (this._ws) {
+      this._ws.close();
+      this._ws = null;
     }
   }
-  _buildToolSchemas() {
-    const toolDefs = this._tools ? this._tools.toOpenAITools().map((t) => ({
-      name: t.function.name,
-      description: t.function.description,
-      parameters: sanitizeSchemaForGemini(
-        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);
-    return toolDefs;
+  _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) {
-    if (!this._call) return;
-    const serverContent = msg.serverContent;
-    if (serverContent) {
-      const modelTurn = serverContent.modelTurn;
-      if (modelTurn) {
-        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);
+    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;
       }
-      if (serverContent.turnComplete) {
-        this._log.debug("Turn complete");
-        this._flushAudioRemainder();
+      case "input_audio_buffer.speech_started": {
+        this._handleTruncation();
+        break;
       }
-      if (serverContent.interrupted) {
-        this._log.info("Barge-in detected");
+      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.clearAudio();
+          this._call._emit("transcript", "user", msg["transcript"] ?? "");
         }
-        this._sentAudioChunks = 0;
-        this._audioRemainder = Buffer.alloc(0);
+        break;
       }
-      const inputText = serverContent.inputTranscription?.text;
-      if (inputText && this._call) {
-        this._log.info("User: %s", inputText);
-        this._call._emit("transcript", "user", inputText);
+      case "response.audio_transcript.done": {
+        if (this._call) {
+          this._call._emit("transcript", "assistant", msg["transcript"] ?? "");
+        }
+        break;
       }
-      const outputText = serverContent.outputTranscription?.text;
-      if (outputText && this._call) {
-        this._log.info("Assistant: %s", outputText);
-        this._call._emit("transcript", "assistant", outputText);
+      case "response.created": {
+        this._responseInProgress = true;
+        break;
+      }
+      case "response.done": {
+        this._responseInProgress = false;
+        if (this._onResponseDone) {
+          const cb = this._onResponseDone;
+          this._onResponseDone = null;
+          cb();
+        }
+        break;
+      }
+      case "error": {
+        this._log.error({ apiError: msg["error"] }, "OpenAI error");
+        break;
       }
-    }
-    if (msg.toolCall) {
-      this._handleToolCall(msg.toolCall);
-    }
-    const toolCancellation = msg["toolCallCancellation"];
-    if (toolCancellation) {
-      this._log.info({ ids: toolCancellation.ids }, "Tool call cancelled");
     }
   }
-  _handleAudioData(b64Data) {
-    if (!this._call) return;
-    const pcm24k = Buffer.from(b64Data, "base64");
+  _handleAudioDelta(msg) {
+    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(resamplePcm16(pcm24k, 24e3, 8e3));
+      this._recorder.writeOutbound(ulawToPcm16(ulaw));
     }
-    const pcm8k = resamplePcm16(pcm24k, 24e3, 8e3);
-    const ulaw = pcm16ToUlaw(pcm8k);
-    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;
-    if (fullEnd > 0) {
-      this._call.sendAudio(combined.subarray(0, fullEnd));
-      this._sentAudioChunks += fullEnd / chunkSize;
+    for (let off = 0; off < fullEnd; off += chunkSize) {
+      if (this._call) {
+        this._call.sendAudio(combined.subarray(off, off + chunkSize));
+      }
+      pb.sentChunks++;
     }
-    this._audioRemainder = combined.subarray(fullEnd);
+    pb.audioRemainder = combined.subarray(fullEnd);
   }
-  _flushAudioRemainder() {
-    if (this._audioRemainder.length > 0 && this._call) {
-      const padded = Buffer.concat([
-        this._audioRemainder,
-        Buffer.alloc(160 - this._audioRemainder.length, 255)
-      ]);
-      this._call.sendAudio(padded);
-      this._sentAudioChunks++;
-      this._audioRemainder = Buffer.alloc(0);
+  _handleTruncation() {
+    for (const [, controller] of this._pendingToolCalls) {
+      controller.abort();
+    }
+    this._pendingToolCalls.clear();
+    if (this._call) {
+      this._call.clearAudio();
+    }
+    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(toolCall) {
-    const functionCalls = toolCall.functionCalls;
-    if (!functionCalls) return;
-    this._toolCallInProgress = true;
-    const responses = [];
-    for (const fc of functionCalls) {
-      const name = fc.name ?? "";
-      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();
+  async _handleToolCall(item) {
+    const funcName = item["name"];
+    const callId = item["call_id"];
+    this._log.info("Tool call: %s", funcName);
+    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
+            }
+          });
+          this._send({ type: "response.create" });
+          return;
         }
-        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}`;
+      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: JSON.stringify({ error: `Unknown tool: ${funcName}` })
           }
-          responses.push({
-            id: fcId,
-            name,
-            response: { result: result || "(\uD0C0\uC784\uC544\uC6C3 - \uC785\uB825 \uC5C6\uC74C)" }
-          });
+        });
+        this._send({ type: "response.create" });
+        return;
+      }
+      let result;
+      try {
+        const args = JSON.parse(item["arguments"] ?? "{}");
+        this._call?.recordToolCall();
+        result = await this._tools.call(funcName, args);
+      } catch (err) {
+        this._log.error({ err }, "Tool call failed: %s", funcName);
+        if (err instanceof Error) {
+          this._call?.recordToolError(err);
         }
-        continue;
+        result = `Error: ${err}`;
       }
-      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: ${err}`;
+      if (controller.signal.aborted) {
+        this._log.info("Tool call cancelled (user interrupted): %s", funcName);
+        return;
+      }
+      const resultStr = typeof result === "string" ? result : JSON.stringify(result);
+      await this._waitForResponseDone();
+      this._send({
+        type: "conversation.item.create",
+        item: {
+          type: "function_call_output",
+          call_id: callId,
+          output: resultStr
+        }
+      });
+      this._log.info("[ToolResult] %s call_id=%s len=%d", funcName, callId, resultStr.length);
+      this._send({ type: "response.create" });
+    } finally {
+      this._pendingToolCalls.delete(callId);
+    }
+  }
+  _waitForResponseDone() {
+    if (!this._responseInProgress) return Promise.resolve();
+    return new Promise((resolve) => {
+      this._onResponseDone = resolve;
+    });
+  }
+  _send(data) {
+    if (this._ws && this._ws.readyState === 1 && !this._closed) {
+      this._ws.send(JSON.stringify(data));
+    }
+  }
+};
+
+// src/agent/pipeline/realtime/gemini-realtime.ts
+function resolveRef(ref, defs) {
+  const parts = ref.replace(/^#\//, "").split("/");
+  let result = defs;
+  for (const part of parts) {
+    if (result && typeof result === "object" && !Array.isArray(result)) {
+      result = result[part];
+    } else {
+      return {};
+    }
+  }
+  return typeof result === "object" && result !== null && !Array.isArray(result) ? result : {};
+}
+function sanitizeSchemaForGemini(schema, defs, depth = 0) {
+  if (depth > 15) return { type: "object", properties: {} };
+  if (!schema || typeof schema !== "object") return { type: "object", properties: {} };
+  if (defs === void 0) {
+    defs = schema["$defs"] ?? schema["definitions"] ?? {};
+  }
+  if (typeof schema["$ref"] === "string") {
+    const resolved = resolveRef(schema["$ref"], { $defs: defs, definitions: defs });
+    if (resolved && Object.keys(resolved).length > 0) {
+      return sanitizeSchemaForGemini(resolved, defs, depth + 1);
+    }
+    return { type: "object", properties: {} };
+  }
+  for (const comboKey of ["oneOf", "anyOf", "allOf"]) {
+    const variants = schema[comboKey];
+    if (Array.isArray(variants) && variants.length > 0) {
+      for (const v of variants) {
+        if (v && typeof v === "object") {
+          const resolved = sanitizeSchemaForGemini(v, defs, depth + 1);
+          if (resolved["type"] === "object" && resolved["properties"]) {
+            return resolved;
           }
-          responses.push({ id: fcId, name, response: { result } });
         }
-        continue;
-      }
-      if (!this._tools || !this._tools.has(name)) {
-        this._log.error("Unknown tool: %s", name);
-        responses.push({ id: fcId, name, response: { error: `Unknown tool: ${name}` } });
-        continue;
       }
-      try {
-        const result = await this._tools.call(name, args);
-        const resultStr = typeof result === "string" ? result : JSON.stringify(result);
-        this._log.info("Tool result: %s -> %s", name, resultStr.substring(0, 200));
-        responses.push({
-          id: fcId,
-          name,
-          response: { result: resultStr }
-        });
-      } catch (err) {
-        this._log.error({ err }, "Tool call failed: %s", name);
-        responses.push({
-          id: fcId,
-          name,
-          response: { error: String(err) }
-        });
+      const first = variants[0];
+      if (first && typeof first === "object") {
+        return sanitizeSchemaForGemini(first, defs, depth + 1);
       }
     }
-    if (responses.length > 0 && this._session) {
-      this._log.debug("Sending %d tool response(s)", responses.length);
-      this._session.sendToolResponse({
-        functionResponses: responses
-      });
-    }
-    this._toolCallInProgress = false;
   }
-};
-
-// 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"]
-    }
+  const result = {};
+  let schemaType = schema["type"];
+  if (Array.isArray(schemaType)) {
+    const nonNull = schemaType.filter((t) => t !== "null");
+    schemaType = nonNull[0] ?? "string";
   }
-};
-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"]
+  if (schemaType) result["type"] = schemaType;
+  if (schema["description"]) result["description"] = schema["description"];
+  if (schema["enum"]) result["enum"] = schema["enum"];
+  if (schema["required"]) result["required"] = schema["required"];
+  if (schema["properties"] && typeof schema["properties"] === "object") {
+    const props = {};
+    for (const [key, val] of Object.entries(schema["properties"])) {
+      if (val && typeof val === "object") {
+        props[key] = sanitizeSchemaForGemini(val, defs, depth + 1);
+      }
     }
+    result["properties"] = props;
   }
-};
-var PipelineSession = class {
-  _stt;
-  _llm;
-  _tts;
+  if (schema["items"] && typeof schema["items"] === "object" && !Array.isArray(schema["items"])) {
+    result["items"] = sanitizeSchemaForGemini(
+      schema["items"],
+      defs,
+      depth + 1
+    );
+  }
+  if (!result["type"] && result["properties"]) result["type"] = "object";
+  if (result["type"] === "object" && !result["properties"]) result["properties"] = {};
+  return result;
+}
+var GeminiRealtime = class {
+  _apiKey;
   _systemPrompt;
-  _greeting;
+  _model;
+  _voice;
   _language;
-  _temperature;
-  _maxTokens;
-  _sampleRate;
-  _interruptOnSpeech;
-  _callSession = null;
+  _greeting;
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
+  _session = null;
+  _call = null;
   _tools = null;
   _recorder = null;
-  _conversation = [];
-  _audioBuffer = [];
-  _running = false;
-  _speaking = false;
+  _closed = false;
+  _sentAudioChunks = 0;
+  _audioRemainder = Buffer.alloc(0);
   _builtinTools = null;
+  _toolCallInProgress = false;
   _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;
+  constructor(options = {}) {
+    this._apiKey = options.apiKey ?? process.env["GOOGLE_API_KEY"] ?? "";
+    this._systemPrompt = options.systemPrompt ?? "";
+    this._model = options.model ?? "gemini-2.5-flash-native-audio-preview-12-2025";
+    this._voice = options.voice ?? "Kore";
     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;
+    this._greeting = options.greeting ?? true;
   }
+  /** Inject per-call ToolRegistry. */
   setToolRegistry(registry) {
     this._tools = registry;
   }
+  /** Inject per-call AudioRecorder. */
   setRecorder(recorder) {
     this._recorder = recorder;
   }
@@ -2646,239 +2717,266 @@ var PipelineSession = class {
   }
   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);
-    }
+  }
+  getTelemetry() {
+    return {
+      sessionType: "gemini_realtime",
+      llm: { provider: "gemini", model: this._model },
+      stt: null,
+      tts: null,
+      voice: this._voice,
+      language: this._language,
+      greetingEnabled: this._greeting,
+      recordingEnabled: !!this._recorder,
+      toolCount: this._tools?.size ?? 0,
+      mcpServerCount: 0,
+      builtinTools: []
+    };
   }
   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
-      });
+    this._call = callSession;
+    if (tools) this._tools = tools;
+    this._closed = false;
+    this._sentAudioChunks = 0;
+    this._audioRemainder = Buffer.alloc(0);
+    if (!this._apiKey) {
+      throw new Error("Google API key is required. Set GOOGLE_API_KEY or pass apiKey option.");
     }
-    if (this._greeting) {
-      this._generateGreeting().catch((err) => {
-        this._log.error({ err }, "Greeting error");
-      });
+    const { GoogleGenAI } = await import('@google/genai/node');
+    const client = new GoogleGenAI({ apiKey: this._apiKey });
+    const config = {
+      responseModalities: ["AUDIO"],
+      speechConfig: {
+        voiceConfig: {
+          prebuiltVoiceConfig: {
+            voiceName: this._voice
+          }
+        }
+      },
+      inputAudioTranscription: {},
+      outputAudioTranscription: {}
+    };
+    if (this._systemPrompt) {
+      config["systemInstruction"] = this._systemPrompt;
     }
-    this._runSttLoop().catch((err) => {
-      this._log.error({ err }, "STT loop error");
-    });
-  }
-  feedAudio(audio) {
-    if (this._running) {
-      this._audioBuffer.push(audio);
+    const toolSchemas = this._buildToolSchemas();
+    if (toolSchemas.length > 0) {
+      config["tools"] = [{ functionDeclarations: toolSchemas }];
     }
-  }
-  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._session = await client.live.connect({
+      model: this._model,
+      config,
+      callbacks: {
+        onmessage: (msg) => this._handleMessage(msg),
+        onerror: (err) => {
+          this._log.error({ err }, "Gemini SDK error");
+        },
+        onclose: (ev) => {
+          this._log.info({ code: ev?.code ?? "unknown" }, "Gemini connection closed");
+          this._closed = true;
         }
-        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);
       }
+    });
+    if (this._greeting) {
+      this._session.sendClientContent({
+        turns: [
+          {
+            role: "user",
+            parts: [{ text: "\uC778\uC0AC\uD574 \uC8FC\uC138\uC694." }]
+          }
+        ],
+        turnComplete: true
+      });
     }
   }
-  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));
+  feedAudio(audio) {
+    if (this._session && !this._closed && !this._toolCallInProgress) {
+      const pcm8k = ulawToPcm16(audio);
+      if (this._recorder) {
+        this._recorder.writeInbound(pcm8k);
       }
-    }
-  }
-  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 () => ""
+      const pcm16k = resamplePcm16(pcm8k, 8e3, 16e3);
+      this._session.sendRealtimeInput({
+        audio: {
+          data: Buffer.from(pcm16k).toString("base64"),
+          mimeType: "audio/pcm;rate=16000"
+        }
       });
     }
-    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 () => ""
+  }
+  async feedDtmf(digits) {
+    if (this._session) {
+      this._session.sendClientContent({
+        turns: [{ role: "user", parts: [{ text: `[DTMF \uC785\uB825: ${digits}]` }] }],
+        turnComplete: true
       });
     }
-    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);
+  async stop() {
+    this._closed = true;
+    if (this._session) {
+      try {
+        this._session.close();
+      } catch {
       }
-    }
-    if (fullResponse.trim()) {
-      this._log.info("Assistant: %s", fullResponse.substring(0, 100));
-      this._conversation.push({ role: "assistant", content: fullResponse });
-      await this._synthesizeAndSend(fullResponse);
+      this._session = null;
     }
   }
-  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}`;
+  _buildToolSchemas() {
+    const toolDefs = this._tools ? this._tools.toOpenAITools().map((t) => ({
+      name: t.function.name,
+      description: t.function.description,
+      parameters: sanitizeSchemaForGemini(
+        t.function.parameters ?? { type: "object", properties: {} }
+      )
+    })) : [];
+    toolDefs.push(...getBuiltinToolSchemas(this._builtinTools, "gemini"));
+    return toolDefs;
+  }
+  _handleMessage(msg) {
+    if (!this._call) return;
+    const serverContent = msg.serverContent;
+    if (serverContent) {
+      const modelTurn = serverContent.modelTurn;
+      if (modelTurn) {
+        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);
+            }
+          }
         }
-        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 (serverContent.turnComplete) {
+        this._log.debug("Turn complete");
+        this._flushAudioRemainder();
       }
-      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 (serverContent.interrupted) {
+        this._log.info("Barge-in detected");
+        if (this._call) {
+          this._call.clearAudio();
         }
+        this._sentAudioChunks = 0;
+        this._audioRemainder = Buffer.alloc(0);
       }
-      if (followUpText.trim()) {
-        this._conversation.push({ role: "assistant", content: followUpText });
-        await this._synthesizeAndSend(followUpText);
+      const inputText = serverContent.inputTranscription?.text;
+      if (inputText && this._call) {
+        this._log.info("User: %s", inputText);
+        this._call._emit("transcript", "user", inputText);
       }
-    } catch (err) {
-      this._log.error({ err }, "Tool call failed: %s", name);
+      const outputText = serverContent.outputTranscription?.text;
+      if (outputText && this._call) {
+        this._log.info("Assistant: %s", outputText);
+        this._call._emit("transcript", "assistant", outputText);
+      }
+    }
+    if (msg.toolCall) {
+      this._handleToolCall(msg.toolCall);
+    }
+    const toolCancellation = msg["toolCallCancellation"];
+    if (toolCancellation) {
+      this._log.info({ ids: toolCancellation.ids }, "Tool call cancelled");
     }
   }
-  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)]);
+  _handleAudioData(b64Data) {
+    if (!this._call) return;
+    const pcm24k = Buffer.from(b64Data, "base64");
+    if (this._recorder) {
+      this._recorder.writeOutbound(resamplePcm16(pcm24k, 24e3, 8e3));
+    }
+    const pcm8k = resamplePcm16(pcm24k, 24e3, 8e3);
+    const ulaw = pcm16ToUlaw(pcm8k);
+    const combined = Buffer.concat([this._audioRemainder, ulaw]);
+    const chunkSize = 160;
+    const fullEnd = Math.floor(combined.length / chunkSize) * chunkSize;
+    if (fullEnd > 0) {
+      this._call.sendAudio(combined.subarray(0, fullEnd));
+      this._sentAudioChunks += fullEnd / chunkSize;
+    }
+    this._audioRemainder = combined.subarray(fullEnd);
+  }
+  _flushAudioRemainder() {
+    if (this._audioRemainder.length > 0 && this._call) {
+      const padded = Buffer.concat([
+        this._audioRemainder,
+        Buffer.alloc(160 - this._audioRemainder.length, 255)
+      ]);
+      this._call.sendAudio(padded);
+      this._sentAudioChunks++;
+      this._audioRemainder = Buffer.alloc(0);
+    }
+  }
+  async _handleToolCall(toolCall) {
+    const functionCalls = toolCall.functionCalls;
+    if (!functionCalls) return;
+    this._toolCallInProgress = true;
+    const responses = [];
+    for (const fc of functionCalls) {
+      const name = fc.name ?? "";
+      const fcId = fc.id ?? "";
+      const args = fc.args ?? {};
+      this._log.info({ tool: name, args }, "Tool call: %s", name);
+      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._callSession.sendAudio(chunk);
+          this._log.info("Builtin tool result: %s -> %s", name, result);
+          responses.push({ id: fcId, name, response: { result } });
+          continue;
         }
       }
-    } catch (err) {
-      this._log.error({ err }, "TTS error");
-    } finally {
-      this._speaking = false;
+      if (!this._tools || !this._tools.has(name)) {
+        this._log.error("Unknown tool: %s", name);
+        responses.push({ id: fcId, name, response: { error: `Unknown tool: ${name}` } });
+        continue;
+      }
+      try {
+        this._call?.recordToolCall();
+        const result = await this._tools.call(name, args);
+        const resultStr = typeof result === "string" ? result : JSON.stringify(result);
+        this._log.info("Tool result: %s -> %s", name, resultStr.substring(0, 200));
+        responses.push({
+          id: fcId,
+          name,
+          response: { result: resultStr }
+        });
+      } catch (err) {
+        this._log.error({ err }, "Tool call failed: %s", name);
+        if (err instanceof Error) {
+          this._call?.recordToolError(err);
+        }
+        responses.push({
+          id: fcId,
+          name,
+          response: { error: String(err) }
+        });
+      }
+    }
+    if (responses.length > 0 && this._session) {
+      this._log.debug("Sending %d tool response(s)", responses.length);
+      this._session.sendToolResponse({
+        functionResponses: responses
+      });
     }
+    this._toolCallInProgress = false;
   }
 };
 
-// src/agent/pipeline/deepgram-stt.ts
+// src/agent/pipeline/stt/deepgram-stt.ts
 var DeepgramSTT = class {
   _options;
   _log = NOOP_LOGGER;
+  get provider() {
+    return "deepgram";
+  }
+  get model() {
+    return this._options.model ?? "nova-3";
+  }
   setLogger(logger) {
     this._log = logger;
   }
@@ -2996,10 +3094,19 @@ var DeepgramSTT = class {
   }
 };
 
-// src/agent/pipeline/elevenlabs-tts.ts
+// src/agent/pipeline/tts/elevenlabs-tts.ts
 var ElevenLabsTTS = class {
   _options;
   _log = NOOP_LOGGER;
+  get provider() {
+    return "elevenlabs";
+  }
+  get model() {
+    return this._options.model ?? "eleven_flash_v2_5";
+  }
+  get voiceId() {
+    return this._options.voiceId ?? "EXAVITQu4vr4xnSDxMaL";
+  }
   setLogger(logger) {
     this._log = logger;
   }
@@ -3149,9 +3256,15 @@ var ElevenLabsTTS = class {
   }
 };
 
-// src/agent/pipeline/openai-llm.ts
+// src/agent/pipeline/llm/openai-llm.ts
 var OpenAILLM = class {
   _options;
+  get provider() {
+    return "openai";
+  }
+  get model() {
+    return this._options.model;
+  }
   constructor(options = {}) {
     this._options = {
       model: "gpt-4o-mini",
@@ -3235,9 +3348,15 @@ var OpenAILLM = class {
   }
 };
 
-// src/agent/pipeline/anthropic-llm.ts
+// src/agent/pipeline/llm/anthropic-llm.ts
 var AnthropicLLM = class {
   _options;
+  get provider() {
+    return "anthropic";
+  }
+  get model() {
+    return this._options.model;
+  }
   constructor(options = {}) {
     this._options = {
       model: "claude-sonnet-4-6",
@@ -3340,9 +3459,15 @@ var AnthropicLLM = class {
   }
 };
 
-// src/agent/pipeline/gemini-llm.ts
+// src/agent/pipeline/llm/gemini-llm.ts
 var GeminiLLM = class {
   _options;
+  get provider() {
+    return "gemini";
+  }
+  get model() {
+    return this._options.model;
+  }
   constructor(options = {}) {
     this._options = {
       model: "gemini-2.5-flash",
@@ -3424,9 +3549,15 @@ var GeminiLLM = class {
   }
 };
 
-// src/agent/pipeline/openai-compat-llm.ts
+// src/agent/pipeline/llm/openai-compat-llm.ts
 var OpenAICompatLLM = class {
   _options;
+  get provider() {
+    return "openai_compatible";
+  }
+  get model() {
+    return this._options.model;
+  }
   constructor(options) {
     this._options = options;
   }
@@ -3507,9 +3638,15 @@ var OpenAICompatLLM = class {
   }
 };
 
-// src/agent/pipeline/ollama-llm.ts
+// src/agent/pipeline/llm/ollama-llm.ts
 var OllamaLLM = class {
   _inner;
+  get provider() {
+    return "ollama";
+  }
+  get model() {
+    return this._inner.model;
+  }
   constructor(options = {}) {
     const baseUrl = options.baseUrl ?? process.env["OLLAMA_BASE_URL"] ?? "http://localhost:11434/v1";
     this._inner = new OpenAICompatLLM({
@@ -3525,9 +3662,15 @@ var OllamaLLM = class {
   }
 };
 
-// src/agent/pipeline/mistral-llm.ts
+// src/agent/pipeline/llm/mistral-llm.ts
 var MistralLLM = class {
   _inner;
+  get provider() {
+    return "mistral";
+  }
+  get model() {
+    return this._inner.model;
+  }
   constructor(options = {}) {
     this._inner = new OpenAICompatLLM({
       apiKey: options.apiKey ?? process.env["MISTRAL_API_KEY"],
@@ -3542,9 +3685,15 @@ var MistralLLM = class {
   }
 };
 
-// src/agent/pipeline/groq-llm.ts
+// src/agent/pipeline/llm/groq-llm.ts
 var GroqLLM = class {
   _inner;
+  get provider() {
+    return "groq";
+  }
+  get model() {
+    return this._inner.model;
+  }
   constructor(options = {}) {
     this._inner = new OpenAICompatLLM({
       apiKey: options.apiKey ?? process.env["GROQ_API_KEY"],
@@ -3559,9 +3708,15 @@ var GroqLLM = class {
   }
 };
 
-// src/agent/pipeline/perplexity-llm.ts
+// src/agent/pipeline/llm/perplexity-llm.ts
 var PerplexityLLM = class {
   _inner;
+  get provider() {
+    return "perplexity";
+  }
+  get model() {
+    return this._inner.model;
+  }
   constructor(options = {}) {
     this._inner = new OpenAICompatLLM({
       apiKey: options.apiKey ?? process.env["PERPLEXITY_API_KEY"],
@@ -3576,9 +3731,15 @@ var PerplexityLLM = class {
   }
 };
 
-// src/agent/pipeline/together-llm.ts
+// src/agent/pipeline/llm/together-llm.ts
 var TogetherLLM = class {
   _inner;
+  get provider() {
+    return "together";
+  }
+  get model() {
+    return this._inner.model;
+  }
   constructor(options = {}) {
     this._inner = new OpenAICompatLLM({
       apiKey: options.apiKey ?? process.env["TOGETHER_API_KEY"],
@@ -3593,9 +3754,15 @@ var TogetherLLM = class {
   }
 };
 
-// src/agent/pipeline/fireworks-llm.ts
+// src/agent/pipeline/llm/fireworks-llm.ts
 var FireworksLLM = class {
   _inner;
+  get provider() {
+    return "fireworks";
+  }
+  get model() {
+    return this._inner.model;
+  }
   constructor(options = {}) {
     this._inner = new OpenAICompatLLM({
       apiKey: options.apiKey ?? process.env["FIREWORKS_API_KEY"],
@@ -3610,9 +3777,15 @@ var FireworksLLM = class {
   }
 };
 
-// src/agent/pipeline/deepseek-llm.ts
+// src/agent/pipeline/llm/deepseek-llm.ts
 var DeepSeekLLM = class {
   _inner;
+  get provider() {
+    return "deepseek";
+  }
+  get model() {
+    return this._inner.model;
+  }
   constructor(options = {}) {
     this._inner = new OpenAICompatLLM({
       apiKey: options.apiKey ?? process.env["DEEPSEEK_API_KEY"],
@@ -3627,9 +3800,15 @@ var DeepSeekLLM = class {
   }
 };
 
-// src/agent/pipeline/xai-llm.ts
+// src/agent/pipeline/llm/xai-llm.ts
 var XaiLLM = class {
   _inner;
+  get provider() {
+    return "xai";
+  }
+  get model() {
+    return this._inner.model;
+  }
   constructor(options = {}) {
     this._inner = new OpenAICompatLLM({
       apiKey: options.apiKey ?? process.env["XAI_API_KEY"],
@@ -3663,6 +3842,6 @@ 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