@teamlearners/clawops 0.4.0 → 0.5.3

package/dist/agent/index.js

@@ -555,6 +555,20 @@ function buildMediaResponse(audioBase64) {
     }
   });
 }
+var VALID_DTMF_DIGITS = new Set("0123456789*#");
+function parseDtmfEvent(data) {
+  const dtmf = data["dtmf"];
+  return {
+    digit: dtmf["digit"] ?? "",
+    track: dtmf["track"] ?? ""
+  };
+}
+function buildDtmfMessage(digit) {
+  if (!VALID_DTMF_DIGITS.has(digit)) {
+    throw new Error(`\uC720\uD6A8\uD558\uC9C0 \uC54A\uC740 DTMF digit: ${digit}`);
+  }
+  return JSON.stringify({ event: "dtmf", dtmf: { digit } });
+}
 var MediaWebSocket = class {
   _ws = null;
   _audioQueue = [];
@@ -563,6 +577,8 @@ var MediaWebSocket = class {
   _onAudio = null;
   _onStart = null;
   _onClose = null;
+  _onDtmf = null;
+  _markWaiters = /* @__PURE__ */ new Map();
   /** Set the handler for inbound audio data. */
   onAudio(handler) {
     this._onAudio = handler;
@@ -575,6 +591,20 @@ var MediaWebSocket = class {
   onClose(handler) {
     this._onClose = handler;
   }
+  /** Set the handler for inbound DTMF events. */
+  onDtmf(handler) {
+    this._onDtmf = handler;
+  }
+  /** Send a single DTMF digit to the platform. */
+  sendDtmf(digit) {
+    if (this._ws && this._ws.readyState === 1) {
+      this._ws.send(buildDtmfMessage(digit));
+    }
+  }
+  /** Whether the WebSocket is connected. */
+  get isConnected() {
+    return this._ws !== null && this._ws.readyState === 1 && !this._closed;
+  }
   /** Connect to a media WebSocket URL with Bearer authentication. */
   async connect(url, apiKey) {
     const { WebSocket } = await import('ws');
@@ -634,6 +664,34 @@ var MediaWebSocket = class {
       );
     }
   }
+  /** Wait for all queued audio to be sent. */
+  flush() {
+    if (this._audioQueue.length === 0 || this._closed) return Promise.resolve();
+    return new Promise((resolve) => {
+      const check = () => {
+        if (this._audioQueue.length === 0 || this._closed) {
+          resolve();
+        } else {
+          setTimeout(check, 5);
+        }
+      };
+      setTimeout(check, 5);
+    });
+  }
+  /** Wait for a named mark to be echoed back by the server. */
+  waitForMark(name, timeoutMs = 5e3) {
+    if (this._closed) return Promise.resolve();
+    return new Promise((resolve) => {
+      const timer = setTimeout(() => {
+        this._markWaiters.delete(name);
+        resolve();
+      }, timeoutMs);
+      this._markWaiters.set(name, () => {
+        clearTimeout(timer);
+        resolve();
+      });
+    });
+  }
   /** Close the media WebSocket. */
   close() {
     this._closed = true;
@@ -659,6 +717,24 @@ var MediaWebSocket = class {
         }
         break;
       }
+      case "dtmf": {
+        const dtmfEvt = parseDtmfEvent(msg);
+        if (this._onDtmf) {
+          this._onDtmf(dtmfEvt.digit);
+        }
+        break;
+      }
+      case "mark": {
+        const markName = msg["mark"]?.["name"];
+        if (markName) {
+          const resolve = this._markWaiters.get(markName);
+          if (resolve) {
+            this._markWaiters.delete(markName);
+            resolve();
+          }
+        }
+        break;
+      }
       case "stop": {
         this.close();
         break;
@@ -847,6 +923,13 @@ var CallSession = class {
   _sendAudioFn = null;
   _clearAudioFn = null;
   _hangupFn = null;
+  /** @internal */
+  _sendDtmfFn = null;
+  /** @internal */
+  _isTransportConnected = null;
+  _dtmfCollectorActive = false;
+  _dtmfResolvers = [];
+  _dtmfBuffer = [];
   _handlers = /* @__PURE__ */ new Map();
   _endedPromise;
   _resolveEnded;
@@ -870,10 +953,12 @@ var CallSession = class {
     return (Date.now() - this.startTime.getTime()) / 1e3;
   }
   /** Bind transport functions (called internally by the agent). */
-  _bindTransport(send, clear, hangup) {
+  _bindTransport(send, clear, hangup, sendDtmf, isConnected) {
     this._sendAudioFn = send;
     this._clearAudioFn = clear;
     this._hangupFn = hangup;
+    if (sendDtmf) this._sendDtmfFn = sendDtmf;
+    if (isConnected) this._isTransportConnected = isConnected;
     this._status = "active";
   }
   /** Send PCM16 or ulaw audio to the caller. */
@@ -888,10 +973,76 @@ var CallSession = class {
       this._clearAudioFn();
     }
   }
-  /** Hang up the call. */
-  hangup() {
+  /** Hang up the call, waiting for pending audio to finish. */
+  async hangup() {
     if (this._hangupFn) {
-      this._hangupFn();
+      await this._hangupFn();
+    }
+  }
+  /** @internal Route a received DTMF digit to an active collector or buffer. */
+  _routeDtmf(digit) {
+    if (this._dtmfCollectorActive && this._dtmfResolvers.length > 0) {
+      const resolve = this._dtmfResolvers.shift();
+      resolve(digit);
+    } else {
+      this._dtmfBuffer.push(digit);
+    }
+  }
+  /** Collect DTMF digits from the caller. */
+  async collectDtmf(options) {
+    if (this._dtmfCollectorActive) {
+      throw new Error("\uC774\uBBF8 DTMF \uC218\uC9D1 \uC911\uC785\uB2C8\uB2E4");
+    }
+    const { maxDigits, finishOnKey = "#", timeout = 5 } = options;
+    this._dtmfCollectorActive = true;
+    const collected = [];
+    try {
+      while (collected.length < maxDigits) {
+        if (this._dtmfBuffer.length > 0) {
+          const digit2 = this._dtmfBuffer.shift();
+          if (digit2 === finishOnKey) break;
+          collected.push(digit2);
+          continue;
+        }
+        const digit = await Promise.race([
+          new Promise((resolve) => {
+            this._dtmfResolvers.push(resolve);
+          }),
+          new Promise((resolve) => {
+            setTimeout(() => resolve(null), timeout * 1e3);
+          })
+        ]);
+        if (digit === null) break;
+        if (digit === finishOnKey) break;
+        collected.push(digit);
+      }
+    } finally {
+      this._dtmfCollectorActive = false;
+      this._dtmfResolvers = [];
+      this._dtmfBuffer = [];
+    }
+    return collected.join("");
+  }
+  /** Send a sequence of DTMF digits. */
+  async sendDtmfSequence(digits) {
+    if (!this._sendDtmfFn) {
+      throw new Error("DTMF \uC804\uC1A1 \uD568\uC218\uAC00 \uBC14\uC778\uB529\uB418\uC9C0 \uC54A\uC558\uC2B5\uB2C8\uB2E4");
+    }
+    for (const ch of digits) {
+      if (this._isTransportConnected && !this._isTransportConnected()) {
+        throw new Error("DTMF \uC804\uC1A1 \uC911 \uC5F0\uACB0\uC774 \uB04A\uC5B4\uC84C\uC2B5\uB2C8\uB2E4");
+      }
+      if (ch === "w") {
+        await new Promise((r) => setTimeout(r, 500));
+      } else if (ch === "W") {
+        await new Promise((r) => setTimeout(r, 1e3));
+      } else if ("0123456789*#".includes(ch)) {
+        if (this._sendDtmfFn) {
+          await this._sendDtmfFn(ch);
+        }
+      } else {
+        throw new Error(`\uC720\uD6A8\uD558\uC9C0 \uC54A\uC740 DTMF \uBB38\uC790: ${ch}`);
+      }
     }
   }
   /** Register an event handler. */
@@ -933,6 +1084,33 @@ var CallSession = class {
   }
 };
 
+// src/agent/builtin-tool.ts
+var BuiltinTool = /* @__PURE__ */ ((BuiltinTool2) => {
+  BuiltinTool2["HANG_UP"] = "hang_up";
+  BuiltinTool2["COLLECT_DTMF"] = "collect_dtmf";
+  BuiltinTool2["SEND_DTMF"] = "send_dtmf";
+  BuiltinTool2["ALL"] = "all";
+  BuiltinTool2["NONE"] = "none";
+  return BuiltinTool2;
+})(BuiltinTool || {});
+var INDIVIDUAL_TOOLS = /* @__PURE__ */ new Set([
+  "hang_up" /* HANG_UP */,
+  "collect_dtmf" /* COLLECT_DTMF */,
+  "send_dtmf" /* SEND_DTMF */
+]);
+function resolveBuiltinTools(value) {
+  if (typeof value === "string") {
+    if (value === "all" /* ALL */) {
+      return new Set(INDIVIDUAL_TOOLS);
+    }
+    if (value === "none" /* NONE */) {
+      return /* @__PURE__ */ new Set();
+    }
+    return /* @__PURE__ */ new Set([value]);
+  }
+  return new Set(value.filter((t) => INDIVIDUAL_TOOLS.has(t)));
+}
+
 // src/agent/tool.ts
 function functionTool(fn) {
   return fn;
@@ -1141,6 +1319,12 @@ var ClawOpsAgent = class {
   _recording;
   _recordingPath;
   _activeSessions = /* @__PURE__ */ new Map();
+  _builtinTools;
+  _passiveDtmfDebounceMs;
+  _passiveDtmfBuffer = [];
+  _passiveDtmfTimer = null;
+  _passiveDtmfCallId = null;
+  _callSessions = /* @__PURE__ */ new Map();
   constructor(options) {
     this._apiKey = options.apiKey ?? process.env["CLAWOPS_API_KEY"] ?? "";
     this._accountId = options.accountId ?? process.env["CLAWOPS_ACCOUNT_ID"] ?? "";
@@ -1150,6 +1334,8 @@ var ClawOpsAgent = class {
     this._recording = options.recording ?? false;
     this._recordingPath = options.recordingPath ?? "./recordings";
     this._mcpServers = options.mcpServers ?? [];
+    this._builtinTools = resolveBuiltinTools(options.builtinTools ?? "all" /* ALL */);
+    this._passiveDtmfDebounceMs = options.passiveDtmfDebounceMs ?? 500;
     if (options.tracing) {
       setTracingConfig(options.tracing);
     }
@@ -1164,7 +1350,9 @@ var ClawOpsAgent = class {
   tool(nameOrTool, description, parameters, handler) {
     if (typeof nameOrTool === "string") {
       if (!description || !parameters || !handler) {
-        throw new AgentError("tool(name, description, parameters, handler) requires all arguments.");
+        throw new AgentError(
+          "tool(name, description, parameters, handler) requires all arguments."
+        );
       }
       this._tools.register({
         name: nameOrTool,
@@ -1200,7 +1388,9 @@ var ClawOpsAgent = class {
       throw new AgentError("API key is required. Set CLAWOPS_API_KEY or pass apiKey option.");
     }
     if (!this._accountId) {
-      throw new AgentError("Account ID is required. Set CLAWOPS_ACCOUNT_ID or pass accountId option.");
+      throw new AgentError(
+        "Account ID is required. Set CLAWOPS_ACCOUNT_ID or pass accountId option."
+      );
     }
     this._controlWs = new ControlWebSocket({
       baseUrl: this._baseUrl,
@@ -1247,6 +1437,7 @@ var ClawOpsAgent = class {
       session._markEnded();
     }
     this._activeSessions.clear();
+    this._callSessions.clear();
     console.log("[ClawOpsAgent] Disconnected");
   }
   /**
@@ -1283,7 +1474,9 @@ var ClawOpsAgent = class {
       }
     }
     this._activeSessions.set(callSession.callId, callSession);
-    console.log(`[ClawOpsAgent] Outbound call initiated: ${this._fromNumber} -> ${to} (${callSession.callId})`);
+    console.log(
+      `[ClawOpsAgent] Outbound call initiated: ${this._fromNumber} -> ${to} (${callSession.callId})`
+    );
     return callSession;
   }
   _handleIncoming(event) {
@@ -1361,6 +1554,30 @@ var ClawOpsAgent = class {
       this._activeSessions.delete(callId);
     }
   }
+  _onDtmfEvent(callSession, digit) {
+    callSession._emit("dtmf", digit);
+    callSession._routeDtmf(digit);
+    if (callSession._dtmfCollectorActive) {
+      callSession.clearAudio();
+      return;
+    }
+    this._passiveDtmfBuffer.push(digit);
+    this._passiveDtmfCallId = callSession.callId;
+    if (this._passiveDtmfTimer) {
+      clearTimeout(this._passiveDtmfTimer);
+    }
+    this._passiveDtmfTimer = setTimeout(() => {
+      const digits = this._passiveDtmfBuffer.join("");
+      this._passiveDtmfBuffer = [];
+      const sessionHandler = this._passiveDtmfCallId ? this._callSessions.get(this._passiveDtmfCallId) : null;
+      this._passiveDtmfCallId = null;
+      if (digits && sessionHandler && sessionHandler.feedDtmf) {
+        sessionHandler.feedDtmf(digits).catch((err) => {
+          console.error("[ClawOpsAgent] feedDtmf error:", err);
+        });
+      }
+    }, this._passiveDtmfDebounceMs);
+  }
   async _startCallSession(session, mediaWsUrl) {
     await withSpan(
       "clawops.call_session",
@@ -1398,9 +1615,17 @@ var ClawOpsAgent = class {
           () => {
             mediaWs.sendClear();
           },
-          () => {
+          async () => {
+            await mediaWs.flush();
+            const markName = `hangup-${Date.now()}`;
+            mediaWs.sendMark(markName);
+            await mediaWs.waitForMark(markName, 5e3);
             mediaWs.close();
-          }
+          },
+          async (digit) => {
+            mediaWs.sendDtmf(digit);
+          },
+          () => mediaWs.isConnected
         );
         const sessionHandler = this._session;
         if ("setToolRegistry" in sessionHandler && typeof sessionHandler.setToolRegistry === "function") {
@@ -1409,6 +1634,10 @@ var ClawOpsAgent = class {
         if (recorder && "setRecorder" in sessionHandler && typeof sessionHandler.setRecorder === "function") {
           sessionHandler.setRecorder(recorder);
         }
+        if ("setBuiltinTools" in sessionHandler && typeof sessionHandler.setBuiltinTools === "function") {
+          sessionHandler.setBuiltinTools(this._builtinTools);
+        }
+        this._callSessions.set(session.callId, sessionHandler);
         mediaWs.onAudio((ulawAudio, _timestamp) => {
           if (sessionHandler) {
             sessionHandler.feedAudio(ulawAudio);
@@ -1417,6 +1646,9 @@ var ClawOpsAgent = class {
             recorder.writeInbound(ulawToPcm16(ulawAudio));
           }
         });
+        mediaWs.onDtmf((digit) => {
+          this._onDtmfEvent(session, digit);
+        });
         mediaWs.onClose(() => {
           if (recorder) {
             recorder.stop();
@@ -1445,6 +1677,7 @@ var ClawOpsAgent = class {
           session._emit("call_end");
           session._markEnded();
           this._activeSessions.delete(session.callId);
+          this._callSessions.delete(session.callId);
         }
       }
     );
@@ -1459,6 +1692,32 @@ var HANG_UP_TOOL = {
   description: "End the phone call. Use when the conversation is finished or the caller says goodbye.",
   parameters: { type: "object", properties: {}, required: [] }
 };
+var COLLECT_DTMF_TOOL = {
+  type: "function",
+  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_TOOL = {
+  type: "function",
+  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"]
+  }
+};
 var OpenAIRealtime = class {
   _apiKey;
   _systemPrompt;
@@ -1467,6 +1726,10 @@ var OpenAIRealtime = class {
   _language;
   _eagerness;
   _greeting;
+  _builtinTools = null;
+  setBuiltinTools(tools) {
+    this._builtinTools = tools;
+  }
   _ws = null;
   _call = null;
   _tools = null;
@@ -1477,6 +1740,9 @@ var OpenAIRealtime = class {
   _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 ?? "";
@@ -1540,6 +1806,18 @@ var OpenAIRealtime = class {
       });
     });
   }
+  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({
@@ -1558,7 +1836,9 @@ var OpenAIRealtime = class {
   _sendSessionUpdate() {
     if (!this._ws || this._ws.readyState !== 1) return;
     const toolSchemas = this._tools ? this._tools.toOpenAITools().map((t) => ({ type: "function", ...t.function })) : [];
-    toolSchemas.push(HANG_UP_TOOL);
+    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: {
@@ -1625,6 +1905,19 @@ var OpenAIRealtime = class {
         }
         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": {
         console.error("[OpenAIRealtime] API error:", msg["error"]);
         break;
@@ -1678,7 +1971,56 @@ var OpenAIRealtime = class {
     const callId = item["call_id"];
     if (funcName === "hang_up") {
       if (this._call) {
-        this._call.hangup();
+        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
+          });
+        } catch (err) {
+          result2 = `Error: ${err}`;
+        }
+        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}`;
+        }
+        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;
     }
@@ -1694,6 +2036,7 @@ var OpenAIRealtime = class {
       console.error(`[OpenAIRealtime] Tool call failed: ${funcName}:`, err);
       result = `Error: ${err}`;
     }
+    await this._waitForResponseDone();
     this._send({
       type: "conversation.item.create",
       item: {
@@ -1704,6 +2047,12 @@ var OpenAIRealtime = class {
     });
     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));
@@ -1717,6 +2066,33 @@ var HANG_UP_TOOL2 = {
   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"]
+  }
+};
 function resolveRef(ref, defs) {
   const parts = ref.replace(/^#\//, "").split("/");
   let result = defs;
@@ -1779,7 +2155,11 @@ function sanitizeSchemaForGemini(schema, defs, depth = 0) {
     result["properties"] = props;
   }
   if (schema["items"] && typeof schema["items"] === "object" && !Array.isArray(schema["items"])) {
-    result["items"] = sanitizeSchemaForGemini(schema["items"], defs, depth + 1);
+    result["items"] = sanitizeSchemaForGemini(
+      schema["items"],
+      defs,
+      depth + 1
+    );
   }
   if (!result["type"] && result["properties"]) result["type"] = "object";
   if (result["type"] === "object" && !result["properties"]) result["properties"] = {};
@@ -1792,14 +2172,16 @@ var GeminiRealtime = class {
   _voice;
   _language;
   _greeting;
-  _generationConfig;
-  _ws = null;
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
+  _session = null;
   _call = null;
   _tools = null;
   _recorder = null;
   _closed = false;
   _sentAudioChunks = 0;
   _audioRemainder = Buffer.alloc(0);
+  _builtinTools = null;
+  _toolCallInProgress = false;
   constructor(options = {}) {
     this._apiKey = options.apiKey ?? process.env["GOOGLE_API_KEY"] ?? "";
     this._systemPrompt = options.systemPrompt ?? "";
@@ -1807,7 +2189,6 @@ var GeminiRealtime = class {
     this._voice = options.voice ?? "Kore";
     this._language = options.language ?? "ko";
     this._greeting = options.greeting ?? true;
-    this._generationConfig = options.generationConfig;
   }
   /** Inject per-call ToolRegistry. */
   setToolRegistry(registry) {
@@ -1817,6 +2198,9 @@ var GeminiRealtime = class {
   setRecorder(recorder) {
     this._recorder = recorder;
   }
+  setBuiltinTools(tools) {
+    this._builtinTools = tools;
+  }
   async start(callSession, tools) {
     this._call = callSession;
     if (tools) this._tools = tools;
@@ -1826,83 +2210,89 @@ var GeminiRealtime = class {
     if (!this._apiKey) {
       throw new Error("Google API key is required. Set GOOGLE_API_KEY or pass apiKey option.");
     }
-    const { WebSocket } = await import('ws');
-    const url = `wss://generativelanguage.googleapis.com/ws/google.ai.generativelanguage.v1beta.GenerativeService.BidiGenerateContent?key=${this._apiKey}`;
-    this._ws = new WebSocket(url);
-    return new Promise((resolve, reject) => {
-      const ws = this._ws;
-      ws.on("open", () => {
-        this._sendSetup();
-        this._waitSetupComplete().then(() => {
-          if (this._greeting) {
-            this._sendGreeting();
+    const { GoogleGenAI } = await import('@google/genai/node');
+    const client = new GoogleGenAI({ apiKey: this._apiKey });
+    const config = {
+      responseModalities: ["AUDIO"],
+      speechConfig: {
+        voiceConfig: {
+          prebuiltVoiceConfig: {
+            voiceName: this._voice
           }
-          this._receiveLoop();
-          resolve();
-        }).catch(reject);
-      });
-      ws.on("close", () => {
-        this._closed = true;
-      });
-      ws.on("error", (err) => {
-        if (!this._ws) {
-          reject(err);
         }
-        console.error("[GeminiRealtime] WebSocket error:", err.message);
-      });
+      },
+      inputAudioTranscription: {},
+      outputAudioTranscription: {}
+    };
+    if (this._systemPrompt) {
+      config["systemInstruction"] = this._systemPrompt;
+    }
+    const toolSchemas = this._buildToolSchemas();
+    if (toolSchemas.length > 0) {
+      config["tools"] = [{ functionDeclarations: toolSchemas }];
+    }
+    this._session = await client.live.connect({
+      model: this._model,
+      config,
+      callbacks: {
+        onmessage: (msg) => this._handleMessage(msg),
+        onerror: (err) => {
+          console.error("[GeminiRealtime] SDK error:", err);
+        },
+        onclose: (ev) => {
+          console.log(
+            `[GeminiRealtime] Connection closed: code=${ev?.code ?? "unknown"}`
+          );
+          this._closed = true;
+        }
+      }
     });
+    if (this._greeting) {
+      this._session.sendClientContent({
+        turns: [
+          {
+            role: "user",
+            parts: [{ text: "\uC778\uC0AC\uD574 \uC8FC\uC138\uC694." }]
+          }
+        ],
+        turnComplete: true
+      });
+    }
   }
   feedAudio(audio) {
-    if (this._ws && this._ws.readyState === 1 && !this._closed) {
+    if (this._session && !this._closed && !this._toolCallInProgress) {
       const pcm8k = ulawToPcm16(audio);
+      if (this._recorder) {
+        this._recorder.writeInbound(pcm8k);
+      }
       const pcm16k = resamplePcm16(pcm8k, 8e3, 16e3);
-      this._ws.send(
-        JSON.stringify({
-          realtimeInput: {
-            mediaChunks: [
-              {
-                mimeType: "audio/pcm;rate=16000",
-                data: pcm16k.toString("base64")
-              }
-            ]
-          }
-        })
-      );
+      this._session.sendRealtimeInput({
+        audio: {
+          data: Buffer.from(pcm16k).toString("base64"),
+          mimeType: "audio/pcm;rate=16000"
+        }
+      });
     }
   }
-  async stop() {
-    this._closed = true;
-    if (this._ws) {
-      this._ws.close();
-      this._ws = null;
+  async feedDtmf(digits) {
+    if (this._session) {
+      this._session.sendClientContent({
+        turns: [{ role: "user", parts: [{ text: `[DTMF \uC785\uB825: ${digits}]` }] }],
+        turnComplete: true
+      });
     }
   }
-  _sendSetup() {
-    if (!this._ws || this._ws.readyState !== 1) return;
-    const setupConfig = {
-      model: `models/${this._model}`,
-      generationConfig: {
-        responseModalities: ["AUDIO"],
-        speechConfig: {
-          voiceConfig: {
-            prebuiltVoiceConfig: {
-              voiceName: this._voice
-            }
-          }
-        },
-        ...this._generationConfig
-      },
-      realtimeInputConfig: {
-        automaticActivityDetection: {
-          disabled: false
-        }
+  async stop() {
+    this._closed = true;
+    if (this._session) {
+      try {
+        this._session.close();
+      } catch {
       }
-    };
-    if (this._systemPrompt) {
-      setupConfig["systemInstruction"] = {
-        parts: [{ text: this._systemPrompt }]
-      };
+      this._session = null;
     }
+  }
+  _buildToolSchemas() {
     const toolDefs = this._tools ? this._tools.toOpenAITools().map((t) => ({
       name: t.function.name,
       description: t.function.description,
@@ -1910,108 +2300,59 @@ var GeminiRealtime = class {
         t.function.parameters ?? { type: "object", properties: {} }
       )
     })) : [];
-    toolDefs.push(HANG_UP_TOOL2);
-    setupConfig["tools"] = [{ functionDeclarations: toolDefs }];
-    this._ws.send(JSON.stringify({ setup: setupConfig }));
-  }
-  _waitSetupComplete() {
-    return new Promise((resolve, reject) => {
-      if (!this._ws) {
-        reject(new Error("WebSocket not connected"));
-        return;
-      }
-      const onMessage = (data) => {
-        try {
-          const msg = JSON.parse(data.toString());
-          if ("setupComplete" in msg) {
-            this._ws?.removeListener("message", onMessage);
-            resolve();
-          }
-        } catch {
-        }
-      };
-      this._ws.on("message", onMessage);
-    });
-  }
-  _sendGreeting() {
-    if (!this._ws || this._ws.readyState !== 1) return;
-    this._ws.send(
-      JSON.stringify({
-        clientContent: {
-          turns: [
-            {
-              role: "user",
-              parts: [{ text: "\uC778\uC0AC\uD574 \uC8FC\uC138\uC694." }]
-            }
-          ],
-          turnComplete: true
-        }
-      })
-    );
-  }
-  _receiveLoop() {
-    if (!this._ws) return;
-    this._ws.on("message", (data) => {
-      try {
-        const msg = JSON.parse(data.toString());
-        this._handleMessage(msg);
-      } catch {
-      }
-    });
+    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;
   }
   _handleMessage(msg) {
     if (!this._call) return;
-    const serverContent = msg["serverContent"];
+    const serverContent = msg.serverContent;
     if (serverContent) {
-      const modelTurn = serverContent["modelTurn"];
+      const modelTurn = serverContent.modelTurn;
       if (modelTurn) {
-        const parts = modelTurn["parts"];
-        if (parts) {
-          for (const part of parts) {
-            const inlineData = part["inlineData"];
-            if (inlineData && inlineData["data"]) {
-              const mimeType = inlineData["mimeType"] ?? "";
-              if (mimeType.includes("audio")) {
-                this._handleAudioData(inlineData["data"]);
-              }
-            }
-            const text = part["text"];
-            if (text && this._call) {
-              this._call._emit("transcript", "assistant", text);
+        for (const part of modelTurn.parts ?? []) {
+          const inlineData = part.inlineData;
+          if (inlineData?.data) {
+            const mimeType = inlineData.mimeType ?? "";
+            if (mimeType.includes("audio")) {
+              this._handleAudioData(inlineData.data);
             }
           }
         }
       }
-      if (serverContent["turnComplete"]) {
+      if (serverContent.turnComplete) {
+        console.log("[GeminiRealtime] Turn complete");
         this._flushAudioRemainder();
       }
-      if (serverContent["interrupted"]) {
+      if (serverContent.interrupted) {
+        console.log("[GeminiRealtime] Barge-in detected");
         if (this._call) {
           this._call.clearAudio();
         }
         this._sentAudioChunks = 0;
         this._audioRemainder = Buffer.alloc(0);
       }
-    }
-    const inputTranscription = msg["inputTranscription"];
-    if (inputTranscription) {
-      const text = inputTranscription["text"];
-      if (text && this._call) {
-        this._call._emit("transcript", "user", text);
+      const inputText = serverContent.inputTranscription?.text;
+      if (inputText && this._call) {
+        console.log(`[GeminiRealtime] [TRANSCRIPT-USER] ${inputText}`);
+        this._call._emit("transcript", "user", inputText);
       }
-    }
-    const outputTranscription = msg["outputTranscription"];
-    if (outputTranscription) {
-      const text = outputTranscription["text"];
-      if (text && this._call) {
-        this._call._emit("transcript", "assistant", text);
+      const outputText = serverContent.outputTranscription?.text;
+      if (outputText && this._call) {
+        console.log(`[GeminiRealtime] [TRANSCRIPT-ASSISTANT] ${outputText}`);
+        this._call._emit("transcript", "assistant", outputText);
       }
     }
-    const toolCall = msg["toolCall"];
-    if (toolCall) {
-      this._handleToolCall(toolCall);
+    if (msg.toolCall) {
+      this._handleToolCall(msg.toolCall);
+    }
+    const toolCancellation = msg["toolCallCancellation"];
+    if (toolCancellation) {
+      console.log(
+        `[GeminiRealtime] Tool call cancelled: ${(toolCancellation.ids ?? []).join(", ")}`
+      );
     }
-    if (msg["toolCallCancellation"]) ;
   }
   _handleAudioData(b64Data) {
     if (!this._call) return;
@@ -2024,9 +2365,9 @@ var GeminiRealtime = class {
     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) {
-      this._call.sendAudio(combined.subarray(off, off + chunkSize));
-      this._sentAudioChunks++;
+    if (fullEnd > 0) {
+      this._call.sendAudio(combined.subarray(0, fullEnd));
+      this._sentAudioChunks += fullEnd / chunkSize;
     }
     this._audioRemainder = combined.subarray(fullEnd);
   }
@@ -2042,19 +2383,66 @@ var GeminiRealtime = class {
     }
   }
   async _handleToolCall(toolCall) {
-    const functionCalls = toolCall["functionCalls"];
+    const functionCalls = toolCall.functionCalls;
     if (!functionCalls) return;
+    this._toolCallInProgress = true;
+    console.log(
+      `[GeminiRealtime] toolCall: ${functionCalls.map((fc) => fc.name).join(", ")}`
+    );
     const responses = [];
     for (const fc of functionCalls) {
-      const name = fc["name"];
-      const fcId = fc["id"] ?? "";
-      const args = fc["args"] ?? {};
+      const name = fc.name ?? "";
+      const fcId = fc.id ?? "";
+      const args = fc.args ?? {};
+      console.log(`[GeminiRealtime] Tool call: ${name}(${JSON.stringify(args)})`);
       if (name === "hang_up") {
+        console.log("[GeminiRealtime] hang_up: ending call");
         if (this._call) {
-          this._call.hangup();
+          await this._call.hangup();
         }
         return;
       }
+      if (name === "collect_dtmf") {
+        if (this._call) {
+          let result;
+          try {
+            console.log(
+              `[GeminiRealtime] collect_dtmf: waiting for digits (maxDigits=${args["max_digits"] ?? 4}, timeout=${args["timeout"] ?? 5})`
+            );
+            result = await this._call.collectDtmf({
+              maxDigits: args["max_digits"] ?? 4,
+              finishOnKey: args["finish_on_key"] ?? "#",
+              timeout: args["timeout"] ?? 5
+            });
+            console.log(`[GeminiRealtime] DTMF collected: ${result || "(empty)"}`);
+          } catch (err) {
+            console.error(`[GeminiRealtime] collect_dtmf error:`, err);
+            result = `Error: ${err}`;
+          }
+          responses.push({
+            id: fcId,
+            name,
+            response: { result: result || "(\uD0C0\uC784\uC544\uC6C3 - \uC785\uB825 \uC5C6\uC74C)" }
+          });
+        }
+        continue;
+      }
+      if (name === "send_dtmf") {
+        if (this._call) {
+          let result;
+          try {
+            console.log(`[GeminiRealtime] send_dtmf: digits="${args["digits"] ?? ""}"`);
+            await this._call.sendDtmfSequence(args["digits"] ?? "");
+            result = "sent";
+            console.log(`[GeminiRealtime] send_dtmf: sent`);
+          } catch (err) {
+            console.error(`[GeminiRealtime] send_dtmf error:`, err);
+            result = `Error: ${err}`;
+          }
+          responses.push({ id: fcId, name, response: { result } });
+        }
+        continue;
+      }
       if (!this._tools || !this._tools.has(name)) {
         console.error(`[GeminiRealtime] Unknown tool: ${name}`);
         responses.push({ id: fcId, name, response: { error: `Unknown tool: ${name}` } });
@@ -2062,13 +2450,15 @@ var GeminiRealtime = class {
       }
       try {
         const result = await this._tools.call(name, args);
+        const resultStr = typeof result === "string" ? result : JSON.stringify(result);
+        console.log(`[GeminiRealtime] Tool result: ${name} -> ${resultStr.substring(0, 200)}`);
         responses.push({
           id: fcId,
           name,
-          response: { result: typeof result === "string" ? result : JSON.stringify(result) }
+          response: { result: resultStr }
         });
       } catch (err) {
-        console.error(`[GeminiRealtime] Tool call error for ${name}:`, err);
+        console.error(`[GeminiRealtime] Tool call failed: ${name}:`, err);
         responses.push({
           id: fcId,
           name,
@@ -2076,19 +2466,41 @@ var GeminiRealtime = class {
         });
       }
     }
-    if (this._ws && this._ws.readyState === 1) {
-      this._ws.send(
-        JSON.stringify({
-          toolResponse: {
-            functionResponses: responses
-          }
-        })
-      );
+    if (responses.length > 0 && this._session) {
+      console.log(`[GeminiRealtime] Sending ${responses.length} tool response(s)`);
+      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"]
+    }
+  }
+};
+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;
@@ -2107,6 +2519,7 @@ var PipelineSession = class {
   _audioBuffer = [];
   _running = false;
   _speaking = false;
+  _builtinTools = null;
   constructor(options) {
     this._stt = options.stt;
     this._llm = options.llm;
@@ -2127,6 +2540,9 @@ var PipelineSession = class {
   setRecorder(recorder) {
     this._recorder = recorder;
   }
+  setBuiltinTools(tools) {
+    this._builtinTools = tools;
+  }
   async start(callSession, tools) {
     this._callSession = callSession;
     this._tools = tools ?? null;
@@ -2152,6 +2568,13 @@ var PipelineSession = class {
       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._audioBuffer = [];
@@ -2193,11 +2616,37 @@ var PipelineSession = class {
     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: this._tools ?? void 0,
+      tools: effectiveTools,
       temperature: this._temperature,
       maxTokens: this._maxTokens
     });
@@ -2216,10 +2665,38 @@ var PipelineSession = class {
     }
   }
   async _handleToolCall(chunk) {
-    if (!chunk.toolCall || !this._tools) return;
+    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",
@@ -2232,9 +2709,10 @@ var PipelineSession = class {
         tool_call_id: id,
         name
       });
+      const effectiveTools = this._buildEffectiveTools();
       let followUpText = "";
       const followUpStream = this._llm.generate(this._conversation, {
-        tools: this._tools ?? void 0,
+        tools: effectiveTools,
         temperature: this._temperature,
         maxTokens: this._maxTokens
       });
@@ -3058,6 +3536,6 @@ function mcpServerHTTP(options) {
   };
 }
 
-export { AnthropicLLM, AudioRecorder, CallSession, ClawOpsAgent, DECODE_TABLE, DeepSeekLLM, DeepgramSTT, ElevenLabsTTS, FireworksLLM, GeminiLLM, GeminiRealtime, GroqLLM, MCPClient, MistralLLM, OllamaLLM, OpenAICompatLLM, OpenAILLM, OpenAIRealtime, PerplexityLLM, PipelineSession, TogetherLLM, ToolRegistry, XaiLLM, functionTool, getTracingConfig, mcpServerHTTP, mcpServerStdio, pcm16ToUlaw, resamplePcm16, resetTracingConfig, setTracingConfig, ulawToPcm16, zodToToolParams };
+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, functionTool, getTracingConfig, mcpServerHTTP, mcpServerStdio, pcm16ToUlaw, resamplePcm16, resetTracingConfig, setTracingConfig, ulawToPcm16, zodToToolParams };
 //# sourceMappingURL=index.js.map
 //# sourceMappingURL=index.js.map