@craftedxp/voice-js 0.3.1 → 0.4.0

package/dist/node.js

@@ -1,592 +1,703 @@
-"use strict";
-var __create = Object.create;
-var __defProp = Object.defineProperty;
-var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
-var __getOwnPropNames = Object.getOwnPropertyNames;
-var __getProtoOf = Object.getPrototypeOf;
-var __hasOwnProp = Object.prototype.hasOwnProperty;
+'use strict'
+var __create = Object.create
+var __defProp = Object.defineProperty
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor
+var __getOwnPropNames = Object.getOwnPropertyNames
+var __getProtoOf = Object.getPrototypeOf
+var __hasOwnProp = Object.prototype.hasOwnProperty
 var __export = (target, all) => {
-  for (var name in all)
-    __defProp(target, name, { get: all[name], enumerable: true });
-};
+  for (var name in all) __defProp(target, name, { get: all[name], enumerable: true })
+}
 var __copyProps = (to, from, except, desc) => {
-  if (from && typeof from === "object" || typeof from === "function") {
+  if ((from && typeof from === 'object') || typeof from === 'function') {
     for (let key of __getOwnPropNames(from))
       if (!__hasOwnProp.call(to, key) && key !== except)
-        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+        __defProp(to, key, {
+          get: () => from[key],
+          enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable,
+        })
   }
-  return to;
-};
-var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__getProtoOf(mod)) : {}, __copyProps(
-  // If the importer is in node compatibility mode or this is not an ESM
-  // file that has been converted to a CommonJS file using a Babel-
-  // compatible transform (i.e. "__esModule" has not been set), then set
-  // "default" to the CommonJS "module.exports" for node compatibility.
-  isNodeMode || !mod || !mod.__esModule ? __defProp(target, "default", { value: mod, enumerable: true }) : target,
-  mod
-));
-var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+  return to
+}
+var __toESM = (mod, isNodeMode, target) => (
+  (target = mod != null ? __create(__getProtoOf(mod)) : {}),
+  __copyProps(
+    // If the importer is in node compatibility mode or this is not an ESM
+    // file that has been converted to a CommonJS file using a Babel-
+    // compatible transform (i.e. "__esModule" has not been set), then set
+    // "default" to the CommonJS "module.exports" for node compatibility.
+    isNodeMode || !mod || !mod.__esModule
+      ? __defProp(target, 'default', { value: mod, enumerable: true })
+      : target,
+    mod,
+  )
+)
+var __toCommonJS = (mod) => __copyProps(__defProp({}, '__esModule', { value: true }), mod)
 
 // src/node.ts
-var node_exports = {};
+var node_exports = {}
 __export(node_exports, {
   buildWsUrl: () => buildWsUrl,
   configureVoiceClient: () => configureVoiceClient,
   createProtocolState: () => createProtocolState,
   createReconnectingWebSocket: () => createReconnectingWebSocket,
-  handleServerMessage: () => handleServerMessage
-});
-module.exports = __toCommonJS(node_exports);
+  handleServerMessage: () => handleServerMessage,
+})
+module.exports = __toCommonJS(node_exports)
 
 // src/config.ts
 function normalizeConfig(config) {
-  if (!config) throw new Error("configureVoiceClient: config is required");
-  if ("apiKey" in config) {
+  if (!config) throw new Error('configureVoiceClient: config is required')
+  if ('apiKey' in config) {
     throw new Error(
-      "configureVoiceClient: `apiKey` is no longer supported. Embedding sk_ in JS code ships server-grade credentials to every client. Pass `fetchToken: async ({ agentId }) => { /* call YOUR backend mint */ }` instead \u2014 see the @craftedxp/voice-js README for the migration recipe."
-    );
+      'configureVoiceClient: `apiKey` is no longer supported. Embedding sk_ in JS code ships server-grade credentials to every client. Pass `fetchToken: async ({ agentId }) => { /* call YOUR backend mint */ }` instead \u2014 see the @craftedxp/voice-js README for the migration recipe.',
+    )
   }
   if (!config.apiBase) {
-    throw new Error("configureVoiceClient: apiBase is required");
+    throw new Error('configureVoiceClient: apiBase is required')
   }
-  if (typeof config.fetchToken !== "function") {
-    throw new Error("configureVoiceClient: fetchToken must be a function");
+  if (typeof config.fetchToken !== 'function') {
+    throw new Error('configureVoiceClient: fetchToken must be a function')
   }
   return {
     ...config,
-    apiBase: config.apiBase.replace(/\/+$/, "")
-  };
+    apiBase: config.apiBase.replace(/\/+$/, ''),
+  }
 }
 function mergeStartCallContext(factory, call) {
-  const context = factory.defaultContext || call.context ? { ...factory.defaultContext ?? {}, ...call.context ?? {} } : void 0;
-  const metadata = factory.defaultMetadata || call.metadata ? { ...factory.defaultMetadata ?? {}, ...call.metadata ?? {} } : void 0;
-  return { context, metadata };
+  const context =
+    factory.defaultContext || call.context
+      ? { ...(factory.defaultContext ?? {}), ...(call.context ?? {}) }
+      : void 0
+  const metadata =
+    factory.defaultMetadata || call.metadata
+      ? { ...(factory.defaultMetadata ?? {}), ...(call.metadata ?? {}) }
+      : void 0
+  return { context, metadata }
 }
 
 // src/ReconnectingWebSocket.ts
-var READYSTATE_OPEN = 1;
-var READYSTATE_CLOSED = 3;
+var READYSTATE_OPEN = 1
+var READYSTATE_CLOSED = 3
 var createReconnectingWebSocket = (options, onEvent) => {
-  const maxRetries = options.maxRetries ?? 3;
-  const initialBackoff = options.initialBackoffMs ?? 500;
-  const maxBackoff = options.maxBackoffMs ?? 8e3;
-  let ws = null;
-  let intentionalClose = false;
-  let retries = 0;
-  let backoff = initialBackoff;
-  let reconnectTimer = null;
+  const maxRetries = options.maxRetries ?? 3
+  const initialBackoff = options.initialBackoffMs ?? 500
+  const maxBackoff = options.maxBackoffMs ?? 8e3
+  let ws = null
+  let intentionalClose = false
+  let retries = 0
+  let backoff = initialBackoff
+  let reconnectTimer = null
   const openOnce = () => {
-    ws = options.wsFactory(options.url);
-    ws.binaryType = "arraybuffer";
+    ws = options.wsFactory(options.url)
+    ws.binaryType = 'arraybuffer'
     ws.onopen = () => {
-      if (retries === 0) onEvent({ type: "open" });
-      else onEvent({ type: "reconnected" });
-      retries = 0;
-      backoff = initialBackoff;
-    };
+      if (retries === 0) onEvent({ type: 'open' })
+      else onEvent({ type: 'reconnected' })
+      retries = 0
+      backoff = initialBackoff
+    }
     ws.onmessage = (ev) => {
-      onEvent({ type: "message", data: ev.data });
-    };
+      onEvent({ type: 'message', data: ev.data })
+    }
     ws.onerror = () => {
-      onEvent({ type: "error", error: new Error("WebSocket error") });
-    };
+      onEvent({ type: 'error', error: new Error('WebSocket error') })
+    }
     ws.onclose = (ev) => {
-      ws = null;
-      const shouldRetry = !intentionalClose && retries < maxRetries;
+      ws = null
+      const shouldRetry = !intentionalClose && retries < maxRetries
       if (!shouldRetry) {
         onEvent({
-          type: "close",
+          type: 'close',
           code: ev.code,
           reason: ev.reason,
-          permanent: true
-        });
-        return;
+          permanent: true,
+        })
+        return
       }
       onEvent({
-        type: "close",
+        type: 'close',
         code: ev.code,
         reason: ev.reason,
-        permanent: false
-      });
-      retries++;
-      const delay = Math.min(backoff, maxBackoff);
-      backoff = Math.min(backoff * 2, maxBackoff);
-      reconnectTimer = setTimeout(openOnce, delay);
-    };
-  };
-  openOnce();
+        permanent: false,
+      })
+      retries++
+      const delay = Math.min(backoff, maxBackoff)
+      backoff = Math.min(backoff * 2, maxBackoff)
+      reconnectTimer = setTimeout(openOnce, delay)
+    }
+  }
+  openOnce()
   return {
     send: (data) => {
-      if (ws && ws.readyState === READYSTATE_OPEN) ws.send(data);
+      if (ws && ws.readyState === READYSTATE_OPEN) ws.send(data)
     },
-    close: (code = 1e3, reason = "client-requested") => {
-      intentionalClose = true;
+    close: (code = 1e3, reason = 'client-requested') => {
+      intentionalClose = true
       if (reconnectTimer) {
-        clearTimeout(reconnectTimer);
-        reconnectTimer = null;
+        clearTimeout(reconnectTimer)
+        reconnectTimer = null
       }
       try {
-        ws?.close(code, reason);
-      } catch {
-      }
+        ws?.close(code, reason)
+      } catch {}
     },
-    readyState: () => ws?.readyState ?? READYSTATE_CLOSED
-  };
-};
+    readyState: () => ws?.readyState ?? READYSTATE_CLOSED,
+  }
+}
 
 // src/protocol.ts
 var createProtocolState = () => ({
-  state: "idle",
+  state: 'idle',
   transcript: [],
   agentBubbleId: null,
   idCounter: 0,
-  endReason: null
-});
+  endReason: null,
+})
 var mapEndReason = (raw) => {
-  if (raw === "agent_ended") return "agent_ended";
-  if (raw === "caller_hung_up") return "user_hangup";
-  if (raw === "silence_timeout" || raw === "max_duration") return "timeout";
-  return "error";
-};
+  if (raw === 'agent_ended') return 'agent_ended'
+  if (raw === 'caller_hung_up') return 'user_hangup'
+  if (raw === 'silence_timeout' || raw === 'max_duration') return 'timeout'
+  return 'error'
+}
 function handleServerMessage(raw, state, cb) {
-  let msg;
+  let msg
   try {
-    msg = JSON.parse(raw);
+    msg = JSON.parse(raw)
   } catch {
-    return;
+    return
   }
   switch (msg.type) {
-    case "connected":
-      cb.onConnected();
-      setState(state, "listening", cb);
-      return;
-    case "transcript": {
-      const text = msg.text ?? "";
-      if (!text) return;
-      const isFinal = !!msg.isFinal;
-      if (!isFinal) setState(state, "user_speaking", cb);
-      upsertUserPartial(state, text, isFinal);
-      cb.onTranscript(state.transcript);
-      return;
-    }
-    case "agent_turn_start": {
-      const id = `m${state.idCounter++}`;
-      state.agentBubbleId = id;
-      state.transcript = [...state.transcript, { id, role: "agent", text: "" }];
-      cb.onTranscript(state.transcript);
-      cb.onAgentTurnStart();
-      setState(state, "agent_speaking", cb);
-      return;
-    }
-    case "agent_text": {
-      const delta = msg.text ?? "";
-      if (!delta || !state.agentBubbleId) return;
-      const id = state.agentBubbleId;
-      state.transcript = state.transcript.map(
-        (e) => e.id === id && e.role === "agent" ? { ...e, text: e.text + delta } : e
-      );
-      cb.onTranscript(state.transcript);
-      return;
-    }
-    case "agent_turn_end":
-      state.agentBubbleId = null;
-      setState(state, "listening", cb);
-      return;
-    case "interrupt":
-      cb.onInterrupt();
-      return;
-    case "agent_turn_abort": {
-      const committed = (msg.committedText ?? "").trim();
+    case 'connected':
+      cb.onConnected()
+      setState(state, 'listening', cb)
+      return
+    case 'transcript': {
+      const text = msg.text ?? ''
+      if (!text) return
+      const isFinal = !!msg.isFinal
+      if (!isFinal) setState(state, 'user_speaking', cb)
+      upsertUserPartial(state, text, isFinal)
+      cb.onTranscript(state.transcript)
+      return
+    }
+    case 'agent_turn_start': {
+      const id = `m${state.idCounter++}`
+      state.agentBubbleId = id
+      state.transcript = [...state.transcript, { id, role: 'agent', text: '' }]
+      cb.onTranscript(state.transcript)
+      const seq = typeof msg.seq === 'number' ? msg.seq : void 0
+      cb.onAgentTurnStart(seq)
+      setState(state, 'agent_speaking', cb)
+      return
+    }
+    case 'agent_text': {
+      const delta = msg.text ?? ''
+      if (!delta || !state.agentBubbleId) return
+      const id = state.agentBubbleId
+      state.transcript = state.transcript.map((e) =>
+        e.id === id && e.role === 'agent' ? { ...e, text: e.text + delta } : e,
+      )
+      cb.onTranscript(state.transcript)
+      return
+    }
+    case 'agent_turn_end': {
+      state.agentBubbleId = null
+      const seq = typeof msg.seq === 'number' ? msg.seq : void 0
+      cb.onAgentTurnEnd(seq)
+      setState(state, 'listening', cb)
+      return
+    }
+    case 'interrupt':
+      cb.onInterrupt()
+      return
+    case 'agent_turn_abort': {
+      const committed = (msg.committedText ?? '').trim()
       if (state.agentBubbleId) {
-        const id = state.agentBubbleId;
+        const id = state.agentBubbleId
         if (committed) {
-          state.transcript = state.transcript.map(
-            (e) => e.id === id && e.role === "agent" ? { ...e, text: committed, interrupted: true } : e
-          );
+          state.transcript = state.transcript.map((e) =>
+            e.id === id && e.role === 'agent' ? { ...e, text: committed, interrupted: true } : e,
+          )
         } else {
-          state.transcript = state.transcript.filter((e) => e.id !== id);
+          state.transcript = state.transcript.filter((e) => e.id !== id)
         }
-        cb.onTranscript(state.transcript);
+        cb.onTranscript(state.transcript)
       }
-      state.agentBubbleId = null;
-      return;
+      state.agentBubbleId = null
+      return
     }
-    case "tool_call":
+    case 'tool_call':
       state.transcript = [
         ...state.transcript,
         {
           id: `m${state.idCounter++}`,
-          role: "tool",
-          text: `\u2192 ${String(msg.tool ?? "?")}(${msg.args ? JSON.stringify(msg.args) : ""})`
-        }
-      ];
-      cb.onTranscript(state.transcript);
-      return;
-    case "tool_result":
+          role: 'tool',
+          text: `\u2192 ${String(msg.tool ?? '?')}(${msg.args ? JSON.stringify(msg.args) : ''})`,
+        },
+      ]
+      cb.onTranscript(state.transcript)
+      return
+    case 'tool_result':
       state.transcript = [
         ...state.transcript,
         {
           id: `m${state.idCounter++}`,
-          role: "tool",
-          text: `${msg.ok ? "\u2713" : "\u2717"} ${String(msg.tool ?? "?")}`
-        }
-      ];
-      cb.onTranscript(state.transcript);
-      return;
-    case "client_tool_call": {
-      const toolCallId = String(msg.toolCallId ?? "");
-      const name = String(msg.name ?? "");
-      const args = msg.args ?? {};
-      if (!toolCallId || !name) return;
-      cb.onClientToolCall({ toolCallId, name, args });
-      return;
-    }
-    case "call_end": {
-      const reasonRaw = String(msg.reason ?? "");
-      const reason = mapEndReason(reasonRaw);
-      state.endReason = reason;
+          role: 'tool',
+          text: `${msg.ok ? '\u2713' : '\u2717'} ${String(msg.tool ?? '?')}`,
+        },
+      ]
+      cb.onTranscript(state.transcript)
+      return
+    case 'client_tool_call': {
+      const toolCallId = String(msg.toolCallId ?? '')
+      const name = String(msg.name ?? '')
+      const args = msg.args ?? {}
+      if (!toolCallId || !name) return
+      cb.onClientToolCall({ toolCallId, name, args })
+      return
+    }
+    case 'call_end': {
+      const reasonRaw = String(msg.reason ?? '')
+      const reason = mapEndReason(reasonRaw)
+      state.endReason = reason
       state.transcript = [
         ...state.transcript,
         {
           id: `m${state.idCounter++}`,
-          role: "system",
-          text: `call ended${reasonRaw ? ` (${reasonRaw})` : ""}`
-        }
-      ];
-      cb.onTranscript(state.transcript);
-      cb.onCallEnd(reason);
-      return;
+          role: 'system',
+          text: `call ended${reasonRaw ? ` (${reasonRaw})` : ''}`,
+        },
+      ]
+      cb.onTranscript(state.transcript)
+      cb.onCallEnd(reason)
+      return
     }
-    case "error": {
-      const code = msg.code ?? "server_error";
-      const message = msg.message ?? "server error";
-      cb.onError({ code, message });
-      return;
+    case 'error': {
+      const code = msg.code ?? 'server_error'
+      const message = msg.message ?? 'server error'
+      cb.onError({ code, message })
+      return
     }
   }
 }
 var setState = (state, next, cb) => {
-  if (state.state === next) return;
-  state.state = next;
-  cb.onState(next);
-};
+  if (state.state === next) return
+  cb.onState(next)
+}
 var upsertUserPartial = (state, text, isFinal) => {
-  let idx = -1;
+  let idx = -1
   for (let i = state.transcript.length - 1; i >= 0; i--) {
-    const e = state.transcript[i];
-    if (e.role === "user" && e.committed === false) {
-      idx = i;
-      break;
+    const e = state.transcript[i]
+    if (e.role === 'user' && e.committed === false) {
+      idx = i
+      break
     }
   }
   if (idx === -1) {
     state.transcript = [
       ...state.transcript,
-      { id: `m${state.idCounter++}`, role: "user", text, committed: isFinal }
-    ];
-    return;
+      { id: `m${state.idCounter++}`, role: 'user', text, committed: isFinal },
+    ]
+    return
   }
-  const target = state.transcript[idx];
-  const next = [...state.transcript];
-  next[idx] = { ...target, text, committed: isFinal };
-  state.transcript = next;
-};
+  const target = state.transcript[idx]
+  const next = [...state.transcript]
+  next[idx] = { ...target, text, committed: isFinal }
+  state.transcript = next
+}
 function buildWsUrl(args) {
-  const base = new URL(args.apiBase);
-  const proto = base.protocol === "https:" ? "wss:" : "ws:";
-  const bargeQS = args.bargeIn === false ? "&barge=off" : "";
-  return `${proto}//${base.host}/v1/agents/${encodeURIComponent(args.agentId)}/call?token=${encodeURIComponent(args.token)}${bargeQS}`;
+  const base = new URL(args.apiBase)
+  const proto = base.protocol === 'https:' ? 'wss:' : 'ws:'
+  const bargeQS = args.bargeIn === false ? '&barge=off' : ''
+  return `${proto}//${base.host}/v1/agents/${encodeURIComponent(args.agentId)}/call?token=${encodeURIComponent(args.token)}${bargeQS}`
 }
 
 // src/clientTools.ts
-var NAME_RE = /^[a-zA-Z_][a-zA-Z0-9_]*$/;
-var MAX_TOOLS = 64;
-var MAX_USAGE = 500;
-var MAX_TIMEOUT_MS = 3e4;
+var NAME_RE = /^[a-zA-Z_][a-zA-Z0-9_]*$/
+var MAX_TOOLS = 64
+var MAX_USAGE = 500
+var MAX_TIMEOUT_MS = 3e4
 var validateClientToolMap = (tools) => {
-  if (tools === void 0) return;
-  if (typeof tools !== "object" || tools === null || Array.isArray(tools)) {
-    throw new Error("clientTools must be an object keyed by tool name");
+  if (tools === void 0) return
+  if (typeof tools !== 'object' || tools === null || Array.isArray(tools)) {
+    throw new Error('clientTools must be an object keyed by tool name')
   }
-  const entries = Object.entries(tools);
+  const entries = Object.entries(tools)
   if (entries.length > MAX_TOOLS) {
-    throw new Error(`clientTools may declare at most 64 tools (got ${entries.length})`);
+    throw new Error(`clientTools may declare at most 64 tools (got ${entries.length})`)
   }
   for (const [name, def] of entries) {
     if (!NAME_RE.test(name)) {
       throw new Error(
-        `clientTools["${name}"]: name must be a valid identifier (^[a-zA-Z_][a-zA-Z0-9_]*$)`
-      );
+        `clientTools["${name}"]: name must be a valid identifier (^[a-zA-Z_][a-zA-Z0-9_]*$)`,
+      )
     }
-    if (!def || typeof def !== "object") {
-      throw new Error(`clientTools["${name}"]: must be an object`);
+    if (!def || typeof def !== 'object') {
+      throw new Error(`clientTools["${name}"]: must be an object`)
     }
-    if (typeof def.description !== "string" || def.description.length === 0) {
-      throw new Error(`clientTools["${name}"]: must have a description`);
+    if (typeof def.description !== 'string' || def.description.length === 0) {
+      throw new Error(`clientTools["${name}"]: must have a description`)
     }
-    if (typeof def.handler !== "function") {
-      throw new Error(`clientTools["${name}"]: must have a handler function`);
+    if (typeof def.handler !== 'function') {
+      throw new Error(`clientTools["${name}"]: must have a handler function`)
     }
     if (def.usage !== void 0 && def.usage.length > MAX_USAGE) {
-      throw new Error(`clientTools["${name}"]: usage must be \u2264500 chars`);
+      throw new Error(`clientTools["${name}"]: usage must be \u2264500 chars`)
     }
-    if (def.timeoutMs !== void 0 && (!Number.isFinite(def.timeoutMs) || def.timeoutMs <= 0 || def.timeoutMs > MAX_TIMEOUT_MS)) {
-      throw new Error(`clientTools["${name}"]: timeoutMs must be in (0, 30000]`);
+    if (
+      def.timeoutMs !== void 0 &&
+      (!Number.isFinite(def.timeoutMs) || def.timeoutMs <= 0 || def.timeoutMs > MAX_TIMEOUT_MS)
+    ) {
+      throw new Error(`clientTools["${name}"]: timeoutMs must be in (0, 30000]`)
     }
   }
-};
+}
 var buildRegisterFrame = (tools) => ({
-  type: "client_tools_register",
+  type: 'client_tools_register',
   tools: Object.entries(tools).map(([name, def]) => ({
     name,
     description: def.description,
     parameters: def.parameters,
-    ...def.usage !== void 0 ? { usage: def.usage } : {},
-    ...def.timeoutMs !== void 0 ? { timeoutMs: def.timeoutMs } : {}
-  }))
-});
+    ...(def.usage !== void 0 ? { usage: def.usage } : {}),
+    ...(def.timeoutMs !== void 0 ? { timeoutMs: def.timeoutMs } : {}),
+  })),
+})
 var dispatchClientToolCall = (send, tools, frame) => {
   const safeSend = (payload) => {
     try {
-      send(payload);
-    } catch {
-    }
-  };
-  const tool = tools[frame.name];
+      send(payload)
+    } catch {}
+  }
+  const tool = tools[frame.name]
   if (!tool) {
     safeSend({
-      type: "client_tool_result",
+      type: 'client_tool_result',
       toolCallId: frame.toolCallId,
-      error: `No handler for ${frame.name}`
-    });
-    return;
+      error: `No handler for ${frame.name}`,
+    })
+    return
   }
   void (async () => {
     try {
-      const out = await tool.handler(frame.args);
+      const out = await tool.handler(frame.args)
       safeSend({
-        type: "client_tool_result",
+        type: 'client_tool_result',
         toolCallId: frame.toolCallId,
-        result: typeof out === "string" ? out : JSON.stringify(out)
-      });
+        result: typeof out === 'string' ? out : JSON.stringify(out),
+      })
     } catch (err) {
       safeSend({
-        type: "client_tool_result",
+        type: 'client_tool_result',
         toolCallId: frame.toolCallId,
-        error: err instanceof Error ? err.message : String(err)
-      });
+        error: err instanceof Error ? err.message : String(err),
+      })
+    }
+  })()
+}
+
+// src/ClientMarksBuffer.ts
+var createClientMarksBuffer = (args) => {
+  const now = args.now ?? (() => performance.now())
+  let pendingFirstOutboundAt = null
+  const inFlight = /* @__PURE__ */ new Map()
+  const tryEmit = (seq) => {
+    const slot = inFlight.get(seq)
+    if (!slot) return
+    if (!slot.ended) return
+    const marks = {}
+    if (slot.firstOutboundAt !== null && slot.firstAudibleAt !== null) {
+      marks.client_mic_to_first_audible_ms = slot.firstAudibleAt - slot.firstOutboundAt
+    }
+    args.send({
+      type: 'client_marks',
+      seq,
+      marks,
+      clientNow: Date.now(),
+    })
+    inFlight.delete(seq)
+  }
+  const markFirstOutboundAudio = () => {
+    if (pendingFirstOutboundAt !== null) return
+    pendingFirstOutboundAt = now()
+  }
+  const markFirstAudibleOutput = () => {
+    let target
+    for (const slot of inFlight.values()) {
+      if (!slot.ended) {
+        target = slot
+      }
+    }
+    if (!target) return
+    if (target.firstAudibleAt !== null) return
+    target.firstAudibleAt = now()
+  }
+  const onAgentTurnStart = (seq) => {
+    inFlight.set(seq, {
+      firstOutboundAt: pendingFirstOutboundAt,
+      firstAudibleAt: null,
+      ended: false,
+    })
+    pendingFirstOutboundAt = null
+  }
+  const onAgentTurnEnd = (seq) => {
+    const slot = inFlight.get(seq)
+    if (!slot) {
+      args.send({ type: 'client_marks', seq, marks: {}, clientNow: Date.now() })
+      return
+    }
+    slot.ended = true
+    tryEmit(seq)
+  }
+  const flush = () => {
+    for (const seq of [...inFlight.keys()]) {
+      const slot = inFlight.get(seq)
+      slot.ended = true
+      tryEmit(seq)
     }
-  })();
-};
+    pendingFirstOutboundAt = null
+  }
+  return {
+    markFirstOutboundAudio,
+    markFirstAudibleOutput,
+    onAgentTurnStart,
+    onAgentTurnEnd,
+    flush,
+  }
+}
 
 // src/NodeVoiceClient.ts
 var NodeVoiceClient = class {
   constructor(args) {
-    this.rws = null;
-    this.muted = false;
-    this.startedAt = null;
-    this.endedFired = false;
-    this.lastError = null;
+    this.rws = null
+    this.muted = false
+    this.startedAt = null
+    this.endedFired = false
+    this.lastError = null
     this.end = () => {
-      this.teardown("user_hangup");
-    };
+      this.teardown('user_hangup')
+    }
     this.mute = () => {
-      this.muted = true;
-    };
+      this.muted = true
+    }
     this.unmute = () => {
-      this.muted = false;
-    };
+      this.muted = false
+    }
     // ---------------------------------------------------------------
     // Node-only raw audio surface
     // ---------------------------------------------------------------
     this.sendAudioChunk = (pcm) => {
-      if (!this.rws) return false;
+      if (!this.rws) return false
+      this.marks.markFirstOutboundAudio()
       if (this.muted) {
-        const len = ArrayBuffer.isView(pcm) ? pcm.byteLength : pcm.byteLength;
-        this.rws.send(new ArrayBuffer(len));
-        return true;
+        const len = ArrayBuffer.isView(pcm) ? pcm.byteLength : pcm.byteLength
+        this.rws.send(new ArrayBuffer(len))
+        return true
       }
-      this.rws.send(pcm);
-      return true;
-    };
+      this.rws.send(pcm)
+      return true
+    }
     // ---------------------------------------------------------------
     // Internal
     // ---------------------------------------------------------------
     this.setState = (next) => {
-      if (this.proto.state === next) return;
-      this.proto.state = next;
-      this.args.options.onStateChange?.(next);
-    };
+      if (this.proto.state === next) return
+      this.proto.state = next
+      this.args.options.onStateChange?.(next)
+    }
     this.sendClientToolsRegister = () => {
-      const frame = buildRegisterFrame(this.args.options.clientTools ?? {});
-      this.rws?.send(JSON.stringify(frame));
-    };
+      const frame = buildRegisterFrame(this.args.options.clientTools ?? {})
+      this.rws?.send(JSON.stringify(frame))
+    }
     this.emitError = (err) => {
-      this.lastError = err;
-      this.args.options.onError?.(err);
-    };
+      this.lastError = err
+      this.args.options.onError?.(err)
+    }
     this.handleSocketEvent = (ev) => {
       switch (ev.type) {
-        case "open":
-          break;
-        case "reconnected":
-          this.proto.transcript = [];
-          this.proto.agentBubbleId = null;
-          this.args.options.onTranscript?.(this.proto.transcript);
-          this.setState("listening");
-          break;
-        case "message":
-          if (typeof ev.data === "string") {
+        case 'open':
+          break
+        case 'reconnected':
+          this.proto.transcript = []
+          this.proto.agentBubbleId = null
+          this.args.options.onTranscript?.(this.proto.transcript)
+          this.setState('listening')
+          break
+        case 'message':
+          if (typeof ev.data === 'string') {
             handleServerMessage(ev.data, this.proto, {
               onState: this.setState,
               onTranscript: (entries) => this.args.options.onTranscript?.(entries),
               onError: this.emitError,
               onInterrupt: () => this.args.options.onInterrupt?.(),
-              onAgentTurnStart: () => this.args.options.onAgentTurnStart?.(),
+              onAgentTurnStart: (seq) => {
+                if (typeof seq === 'number') this.marks.onAgentTurnStart(seq)
+                this.args.options.onAgentTurnStart?.()
+              },
+              onAgentTurnEnd: (seq) => {
+                if (typeof seq === 'number') this.marks.onAgentTurnEnd(seq)
+              },
               onCallEnd: (reason) => this.teardown(reason),
               onConnected: () => this.sendClientToolsRegister(),
-              onClientToolCall: (frame) => dispatchClientToolCall(
-                (f) => this.rws?.send(JSON.stringify(f)),
-                this.args.options.clientTools ?? {},
-                frame
-              )
-            });
+              onClientToolCall: (frame) =>
+                dispatchClientToolCall(
+                  (f) => this.rws?.send(JSON.stringify(f)),
+                  this.args.options.clientTools ?? {},
+                  frame,
+                ),
+            })
           } else {
-            this.args.options.onAudioChunk?.(ev.data);
+            this.marks.markFirstAudibleOutput()
+            this.args.options.onAudioChunk?.(ev.data)
           }
-          break;
-        case "close":
+          break
+        case 'close':
           if (ev.permanent) {
-            const reason = this.proto.endReason ?? (this.lastError ? "error" : "user_hangup");
-            this.teardown(reason);
+            const reason = this.proto.endReason ?? (this.lastError ? 'error' : 'user_hangup')
+            this.teardown(reason)
           }
-          break;
-        case "error":
-          this.emitError({ code: "socket_error", message: ev.error.message });
-          break;
+          break
+        case 'error':
+          this.emitError({ code: 'socket_error', message: ev.error.message })
+          break
       }
-    };
+    }
     this.teardown = (reason) => {
       try {
-        this.rws?.close(1e3, reason);
-      } catch {
-      }
-      this.rws = null;
-      this.setState("ended");
-      this.fireEndOnce(reason);
-    };
+        this.marks.flush()
+      } catch {}
+      try {
+        this.rws?.close(1e3, reason)
+      } catch {}
+      this.rws = null
+      this.setState('ended')
+      this.fireEndOnce(reason)
+    }
     this.fireEndOnce = (reason) => {
-      if (this.endedFired) return;
-      this.endedFired = true;
-      const startedAt = this.startedAt ?? Date.now();
+      if (this.endedFired) return
+      this.endedFired = true
+      const startedAt = this.startedAt ?? Date.now()
       this.args.options.onEnd?.({
         reason,
-        errorCode: reason === "error" ? this.lastError?.code : void 0,
-        durationMs: Date.now() - startedAt
-      });
-    };
-    this.args = args;
-    this.proto = createProtocolState();
-    validateClientToolMap(args.options.clientTools);
+        errorCode: reason === 'error' ? this.lastError?.code : void 0,
+        durationMs: Date.now() - startedAt,
+      })
+    }
+    this.args = args
+    this.proto = createProtocolState()
+    validateClientToolMap(args.options.clientTools)
+    this.marks = createClientMarksBuffer({
+      send: (frame) => {
+        try {
+          this.rws?.send(JSON.stringify(frame))
+        } catch {}
+      },
+    })
   }
   // ---------------------------------------------------------------
   // Call interface
   // ---------------------------------------------------------------
   get state() {
-    return this.proto.state;
+    return this.proto.state
   }
   get transcript() {
-    return this.proto.transcript.slice();
+    return this.proto.transcript.slice()
   }
   get isMuted() {
-    return this.muted;
+    return this.muted
   }
   // ---------------------------------------------------------------
   // Lifecycle
   // ---------------------------------------------------------------
   async start() {
-    this.setState("connecting");
-    this.startedAt = Date.now();
+    this.setState('connecting')
+    this.startedAt = Date.now()
     const url = buildWsUrl({
       apiBase: this.args.config.apiBase,
       agentId: this.args.options.agentId,
       token: this.args.token,
-      bargeIn: this.args.options.bargeIn
-    });
+      bargeIn: this.args.options.bargeIn,
+    })
     this.rws = createReconnectingWebSocket(
       {
         url,
         wsFactory: this.args.wsFactory,
-        maxRetries: 3
+        maxRetries: 3,
       },
-      (ev) => this.handleSocketEvent(ev)
-    );
+      (ev) => this.handleSocketEvent(ev),
+    )
   }
-};
+}
 
 // src/node.ts
-var cachedWsCtor = null;
+var cachedWsCtor = null
 var loadWsCtor = async () => {
-  if (cachedWsCtor) return cachedWsCtor;
+  if (cachedWsCtor) return cachedWsCtor
   try {
-    const mod = await import("ws");
-    const ctor = mod.WebSocket ?? mod.default;
+    const mod = await import('ws')
+    const ctor = mod.WebSocket ?? mod.default
     if (!ctor) {
-      throw new Error("imported `ws` but neither default nor named WebSocket export was found");
+      throw new Error('imported `ws` but neither default nor named WebSocket export was found')
     }
-    cachedWsCtor = ctor;
-    return ctor;
+    cachedWsCtor = ctor
+    return ctor
   } catch (err) {
     throw new Error(
-      "@craftedxp/voice-js (node): missing optional peer `ws`. Install it with `npm install ws` (ws is declared as `peerDependenciesMeta.optional` so npm doesn't install it automatically). Original: " + (err instanceof Error ? err.message : String(err))
-    );
+      "@craftedxp/voice-js (node): missing optional peer `ws`. Install it with `npm install ws` (ws is declared as `peerDependenciesMeta.optional` so npm doesn't install it automatically). Original: " +
+        (err instanceof Error ? err.message : String(err)),
+    )
   }
-};
+}
 var NodeVoiceFactory = class {
   constructor(config) {
     this.startCall = async (options) => {
       if (!options.agentId) {
-        throw new Error("startCall: agentId is required");
+        throw new Error('startCall: agentId is required')
       }
-      const WsCtor = await loadWsCtor();
-      const wsFactory = (url) => new WsCtor(url);
-      const { context, metadata } = mergeStartCallContext(this.config, options);
+      const WsCtor = await loadWsCtor()
+      const wsFactory = (url) => new WsCtor(url)
+      const { context, metadata } = mergeStartCallContext(this.config, options)
       const fetchArgs = {
         agentId: options.agentId,
         userId: options.userId,
         context,
-        metadata
-      };
-      let token;
+        metadata,
+      }
+      let token
       if (options.token) {
-        token = options.token;
+        token = options.token
       } else {
-        token = await this.config.fetchToken(fetchArgs);
+        const r = await this.config.fetchToken(fetchArgs)
+        if (!r) {
+          throw new Error('configureVoiceClient.fetchToken returned empty token')
+        }
+        token = typeof r === 'string' ? r : r.token
         if (!token) {
-          throw new Error("configureVoiceClient.fetchToken returned empty token");
+          throw new Error('configureVoiceClient.fetchToken returned an object without `token`')
         }
       }
       const client = new NodeVoiceClient({
         config: this.config,
         options: { ...options, context, metadata },
         token,
-        wsFactory
-      });
-      await client.start();
-      return client;
-    };
-    this.config = config;
+        wsFactory,
+      })
+      await client.start()
+      return client
+    }
+    this.config = config
   }
-};
+}
 function configureVoiceClient(config) {
-  return new NodeVoiceFactory(normalizeConfig(config));
+  return new NodeVoiceFactory(normalizeConfig(config))
 }
 // Annotate the CommonJS export names for ESM import in node:
-0 && (module.exports = {
-  buildWsUrl,
-  configureVoiceClient,
-  createProtocolState,
-  createReconnectingWebSocket,
-  handleServerMessage
-});
-//# sourceMappingURL=node.js.map
+0 &&
+  (module.exports = {
+    buildWsUrl,
+    configureVoiceClient,
+    createProtocolState,
+    createReconnectingWebSocket,
+    handleServerMessage,
+  })
+//# sourceMappingURL=node.js.map