@livekit/agents 1.0.47 → 1.1.0-dev.0

package/dist/inference/interruption/ws_transport.cjs

@@ -0,0 +1,329 @@
+"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 });
+};
+var __copyProps = (to, from, except, desc) => {
+  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 });
+  }
+  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);
+var ws_transport_exports = {};
+__export(ws_transport_exports, {
+  createWsTransport: () => createWsTransport
+});
+module.exports = __toCommonJS(ws_transport_exports);
+var import_web = require("stream/web");
+var import_ws = __toESM(require("ws"), 1);
+var import_zod = require("zod");
+var import_log = require("../../log.cjs");
+var import_utils = require("../utils.cjs");
+var import_defaults = require("./defaults.cjs");
+var import_interruption_cache_entry = require("./interruption_cache_entry.cjs");
+const MSG_SESSION_CREATE = "session.create";
+const MSG_SESSION_CLOSE = "session.close";
+const MSG_SESSION_CREATED = "session.created";
+const MSG_SESSION_CLOSED = "session.closed";
+const MSG_INTERRUPTION_DETECTED = "bargein_detected";
+const MSG_INFERENCE_DONE = "inference_done";
+const MSG_ERROR = "error";
+const wsMessageSchema = import_zod.z.discriminatedUnion("type", [
+  import_zod.z.object({
+    type: import_zod.z.literal(MSG_SESSION_CREATED)
+  }),
+  import_zod.z.object({
+    type: import_zod.z.literal(MSG_SESSION_CLOSED)
+  }),
+  import_zod.z.object({
+    type: import_zod.z.literal(MSG_INTERRUPTION_DETECTED),
+    created_at: import_zod.z.number(),
+    probabilities: import_zod.z.array(import_zod.z.number()).default([]),
+    prediction_duration: import_zod.z.number().default(0)
+  }),
+  import_zod.z.object({
+    type: import_zod.z.literal(MSG_INFERENCE_DONE),
+    created_at: import_zod.z.number(),
+    probabilities: import_zod.z.array(import_zod.z.number()).default([]),
+    prediction_duration: import_zod.z.number().default(0),
+    is_bargein: import_zod.z.boolean().optional()
+  }),
+  import_zod.z.object({
+    type: import_zod.z.literal(MSG_ERROR),
+    message: import_zod.z.string(),
+    code: import_zod.z.number().optional(),
+    session_id: import_zod.z.string().optional()
+  })
+]);
+async function connectWebSocket(options) {
+  const baseUrl = options.baseUrl.replace(/^http/, "ws");
+  const token = await (0, import_utils.createAccessToken)(options.apiKey, options.apiSecret);
+  const url = `${baseUrl}/bargein`;
+  const ws = new import_ws.default(url, {
+    headers: { Authorization: `Bearer ${token}` }
+  });
+  await new Promise((resolve, reject) => {
+    const timeout = setTimeout(() => {
+      ws.terminate();
+      reject(new Error("WebSocket connection timeout"));
+    }, options.timeout);
+    ws.once("open", () => {
+      clearTimeout(timeout);
+      resolve();
+    });
+    ws.once("error", (err) => {
+      clearTimeout(timeout);
+      ws.terminate();
+      reject(err);
+    });
+  });
+  return ws;
+}
+function createWsTransport(options, getState, setState, updateUserSpeakingSpan, onRequestSent, getAndResetNumRequests) {
+  const logger = (0, import_log.log)();
+  let ws = null;
+  let outputController = null;
+  function setupMessageHandler(socket) {
+    socket.on("message", (data) => {
+      try {
+        const message = wsMessageSchema.parse(JSON.parse(data.toString()));
+        handleMessage(message);
+      } catch {
+        logger.warn({ data: data.toString() }, "Failed to parse WebSocket message");
+      }
+    });
+    socket.on("error", (err) => {
+      logger.error({ err }, "WebSocket error");
+    });
+    socket.on("close", (code, reason) => {
+      logger.debug({ code, reason: reason.toString() }, "WebSocket closed");
+    });
+  }
+  async function ensureConnection() {
+    if (ws && ws.readyState === import_ws.default.OPEN) return;
+    const maxRetries = options.maxRetries ?? 3;
+    let lastError = null;
+    for (let attempt = 0; attempt <= maxRetries; attempt++) {
+      try {
+        ws = await connectWebSocket(options);
+        setupMessageHandler(ws);
+        const sessionCreateMsg = JSON.stringify({
+          type: MSG_SESSION_CREATE,
+          settings: {
+            sample_rate: options.sampleRate,
+            num_channels: 1,
+            threshold: options.threshold,
+            min_frames: options.minFrames,
+            encoding: "s16le"
+          }
+        });
+        ws.send(sessionCreateMsg);
+        return;
+      } catch (err) {
+        lastError = err instanceof Error ? err : new Error(String(err));
+        if (attempt < maxRetries) {
+          const delay = (0, import_defaults.intervalForRetry)(attempt);
+          logger.debug(
+            { attempt, delay, err: lastError.message },
+            "WebSocket connection failed, retrying"
+          );
+          await new Promise((resolve) => setTimeout(resolve, delay));
+        }
+      }
+    }
+    throw lastError ?? new Error("Failed to connect to WebSocket after retries");
+  }
+  function handleMessage(message) {
+    const state = getState();
+    switch (message.type) {
+      case MSG_SESSION_CREATED:
+        logger.debug("WebSocket session created");
+        break;
+      case MSG_INTERRUPTION_DETECTED: {
+        const createdAt = message.created_at;
+        const overlapSpeechStartedAt = state.overlapSpeechStartedAt;
+        if (state.overlapSpeechStarted && overlapSpeechStartedAt !== void 0) {
+          const existing = state.cache.get(createdAt);
+          const totalDurationInS = (existing == null ? void 0 : existing.requestStartedAt) !== void 0 ? (performance.now() - existing.requestStartedAt) / 1e3 : (performance.now() - createdAt) / 1e3;
+          const entry = state.cache.setOrUpdate(
+            createdAt,
+            () => new import_interruption_cache_entry.InterruptionCacheEntry({ createdAt }),
+            {
+              speechInput: existing == null ? void 0 : existing.speechInput,
+              requestStartedAt: existing == null ? void 0 : existing.requestStartedAt,
+              totalDurationInS,
+              probabilities: message.probabilities,
+              isInterruption: true,
+              predictionDurationInS: message.prediction_duration,
+              detectionDelayInS: (Date.now() - overlapSpeechStartedAt) / 1e3
+            }
+          );
+          if (updateUserSpeakingSpan) {
+            updateUserSpeakingSpan(entry);
+          }
+          logger.debug(
+            {
+              totalDuration: entry.totalDurationInS,
+              predictionDuration: entry.predictionDurationInS,
+              detectionDelay: entry.detectionDelayInS,
+              probability: entry.probability
+            },
+            "interruption detected"
+          );
+          const event = {
+            type: "user_overlapping_speech",
+            timestamp: Date.now(),
+            isInterruption: true,
+            totalDurationInS: entry.totalDurationInS,
+            predictionDurationInS: entry.predictionDurationInS,
+            overlapStartedAt: overlapSpeechStartedAt,
+            speechInput: entry.speechInput,
+            probabilities: entry.probabilities,
+            detectionDelayInS: entry.detectionDelayInS,
+            probability: entry.probability,
+            numRequests: (getAndResetNumRequests == null ? void 0 : getAndResetNumRequests()) ?? 0
+          };
+          outputController == null ? void 0 : outputController.enqueue(event);
+          setState({ overlapSpeechStarted: false });
+        }
+        break;
+      }
+      case MSG_INFERENCE_DONE: {
+        const createdAt = message.created_at;
+        const overlapSpeechStartedAt = state.overlapSpeechStartedAt;
+        if (state.overlapSpeechStarted && overlapSpeechStartedAt !== void 0) {
+          const existing = state.cache.get(createdAt);
+          const totalDurationInS = (existing == null ? void 0 : existing.requestStartedAt) !== void 0 ? (performance.now() - existing.requestStartedAt) / 1e3 : (performance.now() - createdAt) / 1e3;
+          const entry = state.cache.setOrUpdate(
+            createdAt,
+            () => new import_interruption_cache_entry.InterruptionCacheEntry({ createdAt }),
+            {
+              speechInput: existing == null ? void 0 : existing.speechInput,
+              requestStartedAt: existing == null ? void 0 : existing.requestStartedAt,
+              totalDurationInS,
+              predictionDurationInS: message.prediction_duration,
+              probabilities: message.probabilities,
+              isInterruption: message.is_bargein ?? false,
+              detectionDelayInS: (Date.now() - overlapSpeechStartedAt) / 1e3
+            }
+          );
+          logger.debug(
+            {
+              totalDurationInS: entry.totalDurationInS,
+              predictionDurationInS: entry.predictionDurationInS
+            },
+            "interruption inference done"
+          );
+        }
+        break;
+      }
+      case MSG_SESSION_CLOSED:
+        logger.debug("WebSocket session closed");
+        break;
+      case MSG_ERROR:
+        outputController == null ? void 0 : outputController.error(
+          new Error(
+            `LiveKit Adaptive Interruption error${message.code !== void 0 ? ` (${message.code})` : ""}: ${message.message}`
+          )
+        );
+        break;
+    }
+  }
+  function sendAudioData(audioSlice) {
+    if (!ws || ws.readyState !== import_ws.default.OPEN) {
+      throw new Error("WebSocket not connected");
+    }
+    const state = getState();
+    const createdAt = Math.floor(performance.now());
+    state.cache.set(
+      createdAt,
+      new import_interruption_cache_entry.InterruptionCacheEntry({
+        createdAt,
+        requestStartedAt: performance.now(),
+        speechInput: audioSlice
+      })
+    );
+    const header = new ArrayBuffer(8);
+    const view = new DataView(header);
+    view.setUint32(0, createdAt >>> 0, true);
+    view.setUint32(4, Math.floor(createdAt / 4294967296) >>> 0, true);
+    const audioBytes = new Uint8Array(
+      audioSlice.buffer,
+      audioSlice.byteOffset,
+      audioSlice.byteLength
+    );
+    const combined = new Uint8Array(8 + audioBytes.length);
+    combined.set(new Uint8Array(header), 0);
+    combined.set(audioBytes, 8);
+    try {
+      ws.send(combined);
+      onRequestSent == null ? void 0 : onRequestSent();
+    } catch (e) {
+      logger.error(e, `failed to send audio via websocket`);
+    }
+  }
+  function close() {
+    if ((ws == null ? void 0 : ws.readyState) === import_ws.default.OPEN) {
+      const closeMsg = JSON.stringify({ type: MSG_SESSION_CLOSE });
+      try {
+        ws.send(closeMsg);
+      } catch (e) {
+        logger.error(e, "failed to send close message");
+      }
+    }
+    ws == null ? void 0 : ws.close(1e3);
+    ws = null;
+  }
+  async function reconnect() {
+    close();
+  }
+  const transport = new import_web.TransformStream(
+    {
+      async start(controller) {
+        outputController = controller;
+        await ensureConnection();
+      },
+      transform(chunk, controller) {
+        if (!(chunk instanceof Int16Array)) {
+          controller.enqueue(chunk);
+          return;
+        }
+        const state = getState();
+        if (!state.overlapSpeechStartedAt || !state.overlapSpeechStarted) return;
+        try {
+          sendAudioData(chunk);
+        } catch (err) {
+          logger.error({ err }, "Failed to send audio data over WebSocket");
+        }
+      },
+      flush() {
+        close();
+      }
+    },
+    { highWaterMark: 2 },
+    { highWaterMark: 2 }
+  );
+  return { transport, reconnect };
+}
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  createWsTransport
+});
+//# sourceMappingURL=ws_transport.cjs.map

package/dist/inference/interruption/ws_transport.cjs.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/inference/interruption/ws_transport.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { TransformStream } from 'stream/web';\nimport WebSocket from 'ws';\nimport { z } from 'zod';\nimport { log } from '../../log.js';\nimport { createAccessToken } from '../utils.js';\nimport { intervalForRetry } from './defaults.js';\nimport { InterruptionCacheEntry } from './interruption_cache_entry.js';\nimport type { OverlappingSpeechEvent } from './types.js';\nimport type { BoundedCache } from './utils.js';\n\n// WebSocket message types\nconst MSG_SESSION_CREATE = 'session.create';\nconst MSG_SESSION_CLOSE = 'session.close';\nconst MSG_SESSION_CREATED = 'session.created';\nconst MSG_SESSION_CLOSED = 'session.closed';\nconst MSG_INTERRUPTION_DETECTED = 'bargein_detected';\nconst MSG_INFERENCE_DONE = 'inference_done';\nconst MSG_ERROR = 'error';\n\nexport interface WsTransportOptions {\n  baseUrl: string;\n  apiKey: string;\n  apiSecret: string;\n  sampleRate: number;\n  threshold: number;\n  minFrames: number;\n  timeout: number;\n  maxRetries?: number;\n}\n\nexport interface WsTransportState {\n  overlapSpeechStarted: boolean;\n  overlapSpeechStartedAt: number | undefined;\n  cache: BoundedCache<number, InterruptionCacheEntry>;\n}\n\nconst wsMessageSchema = z.discriminatedUnion('type', [\n  z.object({\n    type: z.literal(MSG_SESSION_CREATED),\n  }),\n  z.object({\n    type: z.literal(MSG_SESSION_CLOSED),\n  }),\n  z.object({\n    type: z.literal(MSG_INTERRUPTION_DETECTED),\n    created_at: z.number(),\n    probabilities: z.array(z.number()).default([]),\n    prediction_duration: z.number().default(0),\n  }),\n  z.object({\n    type: z.literal(MSG_INFERENCE_DONE),\n    created_at: z.number(),\n    probabilities: z.array(z.number()).default([]),\n    prediction_duration: z.number().default(0),\n    is_bargein: z.boolean().optional(),\n  }),\n  z.object({\n    type: z.literal(MSG_ERROR),\n    message: z.string(),\n    code: z.number().optional(),\n    session_id: z.string().optional(),\n  }),\n]);\n\ntype WsMessage = z.infer<typeof wsMessageSchema>;\n\n/**\n * Creates a WebSocket connection and waits for it to open.\n */\nasync function connectWebSocket(options: WsTransportOptions): Promise<WebSocket> {\n  const baseUrl = options.baseUrl.replace(/^http/, 'ws');\n  const token = await createAccessToken(options.apiKey, options.apiSecret);\n  const url = `${baseUrl}/bargein`;\n\n  const ws = new WebSocket(url, {\n    headers: { Authorization: `Bearer ${token}` },\n  });\n\n  await new Promise<void>((resolve, reject) => {\n    const timeout = setTimeout(() => {\n      ws.terminate();\n      reject(new Error('WebSocket connection timeout'));\n    }, options.timeout);\n    ws.once('open', () => {\n      clearTimeout(timeout);\n      resolve();\n    });\n    ws.once('error', (err: Error) => {\n      clearTimeout(timeout);\n      ws.terminate();\n      reject(err);\n    });\n  });\n\n  return ws;\n}\n\nexport interface WsTransportResult {\n  transport: TransformStream<Int16Array | OverlappingSpeechEvent, OverlappingSpeechEvent>;\n  reconnect: () => Promise<void>;\n}\n\n/**\n * Creates a WebSocket transport TransformStream for interruption detection.\n *\n * This transport receives Int16Array audio slices and outputs InterruptionEvents.\n * It maintains a persistent WebSocket connection with automatic retry on failure.\n * Returns both the transport and a reconnect function for option updates.\n */\nexport function createWsTransport(\n  options: WsTransportOptions,\n  getState: () => WsTransportState,\n  setState: (partial: Partial<WsTransportState>) => void,\n  updateUserSpeakingSpan?: (entry: InterruptionCacheEntry) => void,\n  onRequestSent?: () => void,\n  getAndResetNumRequests?: () => number,\n): WsTransportResult {\n  const logger = log();\n  let ws: WebSocket | null = null;\n  let outputController: TransformStreamDefaultController<OverlappingSpeechEvent> | null = null;\n\n  function setupMessageHandler(socket: WebSocket): void {\n    socket.on('message', (data: WebSocket.Data) => {\n      try {\n        const message = wsMessageSchema.parse(JSON.parse(data.toString()));\n        handleMessage(message);\n      } catch {\n        logger.warn({ data: data.toString() }, 'Failed to parse WebSocket message');\n      }\n    });\n\n    socket.on('error', (err: Error) => {\n      logger.error({ err }, 'WebSocket error');\n    });\n\n    socket.on('close', (code: number, reason: Buffer) => {\n      logger.debug({ code, reason: reason.toString() }, 'WebSocket closed');\n    });\n  }\n\n  async function ensureConnection(): Promise<void> {\n    if (ws && ws.readyState === WebSocket.OPEN) return;\n\n    const maxRetries = options.maxRetries ?? 3;\n    let lastError: Error | null = null;\n\n    for (let attempt = 0; attempt <= maxRetries; attempt++) {\n      try {\n        ws = await connectWebSocket(options);\n        setupMessageHandler(ws);\n\n        // Send session.create message\n        const sessionCreateMsg = JSON.stringify({\n          type: MSG_SESSION_CREATE,\n          settings: {\n            sample_rate: options.sampleRate,\n            num_channels: 1,\n            threshold: options.threshold,\n            min_frames: options.minFrames,\n            encoding: 's16le',\n          },\n        });\n        ws.send(sessionCreateMsg);\n        return;\n      } catch (err) {\n        lastError = err instanceof Error ? err : new Error(String(err));\n        if (attempt < maxRetries) {\n          const delay = intervalForRetry(attempt);\n          logger.debug(\n            { attempt, delay, err: lastError.message },\n            'WebSocket connection failed, retrying',\n          );\n          await new Promise((resolve) => setTimeout(resolve, delay));\n        }\n      }\n    }\n\n    throw lastError ?? new Error('Failed to connect to WebSocket after retries');\n  }\n\n  function handleMessage(message: WsMessage): void {\n    const state = getState();\n\n    switch (message.type) {\n      case MSG_SESSION_CREATED:\n        logger.debug('WebSocket session created');\n        break;\n\n      case MSG_INTERRUPTION_DETECTED: {\n        const createdAt = message.created_at;\n        const overlapSpeechStartedAt = state.overlapSpeechStartedAt;\n        if (state.overlapSpeechStarted && overlapSpeechStartedAt !== undefined) {\n          const existing = state.cache.get(createdAt);\n\n          const totalDurationInS =\n            existing?.requestStartedAt !== undefined\n              ? (performance.now() - existing.requestStartedAt) / 1000\n              : (performance.now() - createdAt) / 1000;\n\n          const entry = state.cache.setOrUpdate(\n            createdAt,\n            () => new InterruptionCacheEntry({ createdAt }),\n            {\n              speechInput: existing?.speechInput,\n              requestStartedAt: existing?.requestStartedAt,\n              totalDurationInS,\n              probabilities: message.probabilities,\n              isInterruption: true,\n              predictionDurationInS: message.prediction_duration,\n              detectionDelayInS: (Date.now() - overlapSpeechStartedAt) / 1000,\n            },\n          );\n\n          if (updateUserSpeakingSpan) {\n            updateUserSpeakingSpan(entry);\n          }\n\n          logger.debug(\n            {\n              totalDuration: entry.totalDurationInS,\n              predictionDuration: entry.predictionDurationInS,\n              detectionDelay: entry.detectionDelayInS,\n              probability: entry.probability,\n            },\n            'interruption detected',\n          );\n\n          const event: OverlappingSpeechEvent = {\n            type: 'user_overlapping_speech',\n            timestamp: Date.now(),\n            isInterruption: true,\n            totalDurationInS: entry.totalDurationInS,\n            predictionDurationInS: entry.predictionDurationInS,\n            overlapStartedAt: overlapSpeechStartedAt,\n            speechInput: entry.speechInput,\n            probabilities: entry.probabilities,\n            detectionDelayInS: entry.detectionDelayInS,\n            probability: entry.probability,\n            numRequests: getAndResetNumRequests?.() ?? 0,\n          };\n\n          outputController?.enqueue(event);\n          setState({ overlapSpeechStarted: false });\n        }\n        break;\n      }\n\n      case MSG_INFERENCE_DONE: {\n        const createdAt = message.created_at;\n        const overlapSpeechStartedAt = state.overlapSpeechStartedAt;\n        if (state.overlapSpeechStarted && overlapSpeechStartedAt !== undefined) {\n          const existing = state.cache.get(createdAt);\n          const totalDurationInS =\n            existing?.requestStartedAt !== undefined\n              ? (performance.now() - existing.requestStartedAt) / 1000\n              : (performance.now() - createdAt) / 1000;\n          const entry = state.cache.setOrUpdate(\n            createdAt,\n            () => new InterruptionCacheEntry({ createdAt }),\n            {\n              speechInput: existing?.speechInput,\n              requestStartedAt: existing?.requestStartedAt,\n              totalDurationInS,\n              predictionDurationInS: message.prediction_duration,\n              probabilities: message.probabilities,\n              isInterruption: message.is_bargein ?? false,\n              detectionDelayInS: (Date.now() - overlapSpeechStartedAt) / 1000,\n            },\n          );\n\n          logger.debug(\n            {\n              totalDurationInS: entry.totalDurationInS,\n              predictionDurationInS: entry.predictionDurationInS,\n            },\n            'interruption inference done',\n          );\n        }\n        break;\n      }\n\n      case MSG_SESSION_CLOSED:\n        logger.debug('WebSocket session closed');\n        break;\n\n      case MSG_ERROR:\n        outputController?.error(\n          new Error(\n            `LiveKit Adaptive Interruption error${\n              message.code !== undefined ? ` (${message.code})` : ''\n            }: ${message.message}`,\n          ),\n        );\n        break;\n    }\n  }\n\n  function sendAudioData(audioSlice: Int16Array): void {\n    if (!ws || ws.readyState !== WebSocket.OPEN) {\n      throw new Error('WebSocket not connected');\n    }\n\n    const state = getState();\n    // Use truncated timestamp consistently for both cache key and header\n    // This ensures the server's response created_at matches our cache key\n    const createdAt = Math.floor(performance.now());\n\n    // Store the audio data in cache with truncated timestamp\n    state.cache.set(\n      createdAt,\n      new InterruptionCacheEntry({\n        createdAt,\n        requestStartedAt: performance.now(),\n        speechInput: audioSlice,\n      }),\n    );\n\n    // Create header: 8-byte little-endian uint64 timestamp (milliseconds as integer)\n    const header = new ArrayBuffer(8);\n    const view = new DataView(header);\n    view.setUint32(0, createdAt >>> 0, true);\n    view.setUint32(4, Math.floor(createdAt / 0x100000000) >>> 0, true);\n\n    // Combine header and audio data\n    const audioBytes = new Uint8Array(\n      audioSlice.buffer,\n      audioSlice.byteOffset,\n      audioSlice.byteLength,\n    );\n    const combined = new Uint8Array(8 + audioBytes.length);\n    combined.set(new Uint8Array(header), 0);\n    combined.set(audioBytes, 8);\n\n    try {\n      ws.send(combined);\n      onRequestSent?.();\n    } catch (e: unknown) {\n      logger.error(e, `failed to send audio via websocket`);\n    }\n  }\n\n  function close(): void {\n    if (ws?.readyState === WebSocket.OPEN) {\n      const closeMsg = JSON.stringify({ type: MSG_SESSION_CLOSE });\n      try {\n        ws.send(closeMsg);\n      } catch (e: unknown) {\n        logger.error(e, 'failed to send close message');\n      }\n    }\n    ws?.close(1000); // signal normal websocket closure\n    ws = null;\n  }\n\n  /**\n   * Reconnect the WebSocket with updated options.\n   * This is called when options are updated via updateOptions().\n   */\n  async function reconnect(): Promise<void> {\n    close();\n  }\n\n  const transport = new TransformStream<\n    Int16Array | OverlappingSpeechEvent,\n    OverlappingSpeechEvent\n  >(\n    {\n      async start(controller) {\n        outputController = controller;\n        await ensureConnection();\n      },\n\n      transform(chunk, controller) {\n        if (!(chunk instanceof Int16Array)) {\n          controller.enqueue(chunk);\n          return;\n        }\n\n        // Only forwards buffered audio while overlap speech is actively on.\n        const state = getState();\n        if (!state.overlapSpeechStartedAt || !state.overlapSpeechStarted) return;\n\n        try {\n          sendAudioData(chunk);\n        } catch (err) {\n          logger.error({ err }, 'Failed to send audio data over WebSocket');\n        }\n      },\n\n      flush() {\n        close();\n      },\n    },\n    { highWaterMark: 2 },\n    { highWaterMark: 2 },\n  );\n\n  return { transport, reconnect };\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,iBAAgC;AAChC,gBAAsB;AACtB,iBAAkB;AAClB,iBAAoB;AACpB,mBAAkC;AAClC,sBAAiC;AACjC,sCAAuC;AAKvC,MAAM,qBAAqB;AAC3B,MAAM,oBAAoB;AAC1B,MAAM,sBAAsB;AAC5B,MAAM,qBAAqB;AAC3B,MAAM,4BAA4B;AAClC,MAAM,qBAAqB;AAC3B,MAAM,YAAY;AAmBlB,MAAM,kBAAkB,aAAE,mBAAmB,QAAQ;AAAA,EACnD,aAAE,OAAO;AAAA,IACP,MAAM,aAAE,QAAQ,mBAAmB;AAAA,EACrC,CAAC;AAAA,EACD,aAAE,OAAO;AAAA,IACP,MAAM,aAAE,QAAQ,kBAAkB;AAAA,EACpC,CAAC;AAAA,EACD,aAAE,OAAO;AAAA,IACP,MAAM,aAAE,QAAQ,yBAAyB;AAAA,IACzC,YAAY,aAAE,OAAO;AAAA,IACrB,eAAe,aAAE,MAAM,aAAE,OAAO,CAAC,EAAE,QAAQ,CAAC,CAAC;AAAA,IAC7C,qBAAqB,aAAE,OAAO,EAAE,QAAQ,CAAC;AAAA,EAC3C,CAAC;AAAA,EACD,aAAE,OAAO;AAAA,IACP,MAAM,aAAE,QAAQ,kBAAkB;AAAA,IAClC,YAAY,aAAE,OAAO;AAAA,IACrB,eAAe,aAAE,MAAM,aAAE,OAAO,CAAC,EAAE,QAAQ,CAAC,CAAC;AAAA,IAC7C,qBAAqB,aAAE,OAAO,EAAE,QAAQ,CAAC;AAAA,IACzC,YAAY,aAAE,QAAQ,EAAE,SAAS;AAAA,EACnC,CAAC;AAAA,EACD,aAAE,OAAO;AAAA,IACP,MAAM,aAAE,QAAQ,SAAS;AAAA,IACzB,SAAS,aAAE,OAAO;AAAA,IAClB,MAAM,aAAE,OAAO,EAAE,SAAS;AAAA,IAC1B,YAAY,aAAE,OAAO,EAAE,SAAS;AAAA,EAClC,CAAC;AACH,CAAC;AAOD,eAAe,iBAAiB,SAAiD;AAC/E,QAAM,UAAU,QAAQ,QAAQ,QAAQ,SAAS,IAAI;AACrD,QAAM,QAAQ,UAAM,gCAAkB,QAAQ,QAAQ,QAAQ,SAAS;AACvE,QAAM,MAAM,GAAG,OAAO;AAEtB,QAAM,KAAK,IAAI,UAAAA,QAAU,KAAK;AAAA,IAC5B,SAAS,EAAE,eAAe,UAAU,KAAK,GAAG;AAAA,EAC9C,CAAC;AAED,QAAM,IAAI,QAAc,CAAC,SAAS,WAAW;AAC3C,UAAM,UAAU,WAAW,MAAM;AAC/B,SAAG,UAAU;AACb,aAAO,IAAI,MAAM,8BAA8B,CAAC;AAAA,IAClD,GAAG,QAAQ,OAAO;AAClB,OAAG,KAAK,QAAQ,MAAM;AACpB,mBAAa,OAAO;AACpB,cAAQ;AAAA,IACV,CAAC;AACD,OAAG,KAAK,SAAS,CAAC,QAAe;AAC/B,mBAAa,OAAO;AACpB,SAAG,UAAU;AACb,aAAO,GAAG;AAAA,IACZ,CAAC;AAAA,EACH,CAAC;AAED,SAAO;AACT;AAcO,SAAS,kBACd,SACA,UACA,UACA,wBACA,eACA,wBACmB;AACnB,QAAM,aAAS,gBAAI;AACnB,MAAI,KAAuB;AAC3B,MAAI,mBAAoF;AAExF,WAAS,oBAAoB,QAAyB;AACpD,WAAO,GAAG,WAAW,CAAC,SAAyB;AAC7C,UAAI;AACF,cAAM,UAAU,gBAAgB,MAAM,KAAK,MAAM,KAAK,SAAS,CAAC,CAAC;AACjE,sBAAc,OAAO;AAAA,MACvB,QAAQ;AACN,eAAO,KAAK,EAAE,MAAM,KAAK,SAAS,EAAE,GAAG,mCAAmC;AAAA,MAC5E;AAAA,IACF,CAAC;AAED,WAAO,GAAG,SAAS,CAAC,QAAe;AACjC,aAAO,MAAM,EAAE,IAAI,GAAG,iBAAiB;AAAA,IACzC,CAAC;AAED,WAAO,GAAG,SAAS,CAAC,MAAc,WAAmB;AACnD,aAAO,MAAM,EAAE,MAAM,QAAQ,OAAO,SAAS,EAAE,GAAG,kBAAkB;AAAA,IACtE,CAAC;AAAA,EACH;AAEA,iBAAe,mBAAkC;AAC/C,QAAI,MAAM,GAAG,eAAe,UAAAA,QAAU,KAAM;AAE5C,UAAM,aAAa,QAAQ,cAAc;AACzC,QAAI,YAA0B;AAE9B,aAAS,UAAU,GAAG,WAAW,YAAY,WAAW;AACtD,UAAI;AACF,aAAK,MAAM,iBAAiB,OAAO;AACnC,4BAAoB,EAAE;AAGtB,cAAM,mBAAmB,KAAK,UAAU;AAAA,UACtC,MAAM;AAAA,UACN,UAAU;AAAA,YACR,aAAa,QAAQ;AAAA,YACrB,cAAc;AAAA,YACd,WAAW,QAAQ;AAAA,YACnB,YAAY,QAAQ;AAAA,YACpB,UAAU;AAAA,UACZ;AAAA,QACF,CAAC;AACD,WAAG,KAAK,gBAAgB;AACxB;AAAA,MACF,SAAS,KAAK;AACZ,oBAAY,eAAe,QAAQ,MAAM,IAAI,MAAM,OAAO,GAAG,CAAC;AAC9D,YAAI,UAAU,YAAY;AACxB,gBAAM,YAAQ,kCAAiB,OAAO;AACtC,iBAAO;AAAA,YACL,EAAE,SAAS,OAAO,KAAK,UAAU,QAAQ;AAAA,YACzC;AAAA,UACF;AACA,gBAAM,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,KAAK,CAAC;AAAA,QAC3D;AAAA,MACF;AAAA,IACF;AAEA,UAAM,aAAa,IAAI,MAAM,8CAA8C;AAAA,EAC7E;AAEA,WAAS,cAAc,SAA0B;AAC/C,UAAM,QAAQ,SAAS;AAEvB,YAAQ,QAAQ,MAAM;AAAA,MACpB,KAAK;AACH,eAAO,MAAM,2BAA2B;AACxC;AAAA,MAEF,KAAK,2BAA2B;AAC9B,cAAM,YAAY,QAAQ;AAC1B,cAAM,yBAAyB,MAAM;AACrC,YAAI,MAAM,wBAAwB,2BAA2B,QAAW;AACtE,gBAAM,WAAW,MAAM,MAAM,IAAI,SAAS;AAE1C,gBAAM,oBACJ,qCAAU,sBAAqB,UAC1B,YAAY,IAAI,IAAI,SAAS,oBAAoB,OACjD,YAAY,IAAI,IAAI,aAAa;AAExC,gBAAM,QAAQ,MAAM,MAAM;AAAA,YACxB;AAAA,YACA,MAAM,IAAI,uDAAuB,EAAE,UAAU,CAAC;AAAA,YAC9C;AAAA,cACE,aAAa,qCAAU;AAAA,cACvB,kBAAkB,qCAAU;AAAA,cAC5B;AAAA,cACA,eAAe,QAAQ;AAAA,cACvB,gBAAgB;AAAA,cAChB,uBAAuB,QAAQ;AAAA,cAC/B,oBAAoB,KAAK,IAAI,IAAI,0BAA0B;AAAA,YAC7D;AAAA,UACF;AAEA,cAAI,wBAAwB;AAC1B,mCAAuB,KAAK;AAAA,UAC9B;AAEA,iBAAO;AAAA,YACL;AAAA,cACE,eAAe,MAAM;AAAA,cACrB,oBAAoB,MAAM;AAAA,cAC1B,gBAAgB,MAAM;AAAA,cACtB,aAAa,MAAM;AAAA,YACrB;AAAA,YACA;AAAA,UACF;AAEA,gBAAM,QAAgC;AAAA,YACpC,MAAM;AAAA,YACN,WAAW,KAAK,IAAI;AAAA,YACpB,gBAAgB;AAAA,YAChB,kBAAkB,MAAM;AAAA,YACxB,uBAAuB,MAAM;AAAA,YAC7B,kBAAkB;AAAA,YAClB,aAAa,MAAM;AAAA,YACnB,eAAe,MAAM;AAAA,YACrB,mBAAmB,MAAM;AAAA,YACzB,aAAa,MAAM;AAAA,YACnB,cAAa,uEAA8B;AAAA,UAC7C;AAEA,+DAAkB,QAAQ;AAC1B,mBAAS,EAAE,sBAAsB,MAAM,CAAC;AAAA,QAC1C;AACA;AAAA,MACF;AAAA,MAEA,KAAK,oBAAoB;AACvB,cAAM,YAAY,QAAQ;AAC1B,cAAM,yBAAyB,MAAM;AACrC,YAAI,MAAM,wBAAwB,2BAA2B,QAAW;AACtE,gBAAM,WAAW,MAAM,MAAM,IAAI,SAAS;AAC1C,gBAAM,oBACJ,qCAAU,sBAAqB,UAC1B,YAAY,IAAI,IAAI,SAAS,oBAAoB,OACjD,YAAY,IAAI,IAAI,aAAa;AACxC,gBAAM,QAAQ,MAAM,MAAM;AAAA,YACxB;AAAA,YACA,MAAM,IAAI,uDAAuB,EAAE,UAAU,CAAC;AAAA,YAC9C;AAAA,cACE,aAAa,qCAAU;AAAA,cACvB,kBAAkB,qCAAU;AAAA,cAC5B;AAAA,cACA,uBAAuB,QAAQ;AAAA,cAC/B,eAAe,QAAQ;AAAA,cACvB,gBAAgB,QAAQ,cAAc;AAAA,cACtC,oBAAoB,KAAK,IAAI,IAAI,0BAA0B;AAAA,YAC7D;AAAA,UACF;AAEA,iBAAO;AAAA,YACL;AAAA,cACE,kBAAkB,MAAM;AAAA,cACxB,uBAAuB,MAAM;AAAA,YAC/B;AAAA,YACA;AAAA,UACF;AAAA,QACF;AACA;AAAA,MACF;AAAA,MAEA,KAAK;AACH,eAAO,MAAM,0BAA0B;AACvC;AAAA,MAEF,KAAK;AACH,6DAAkB;AAAA,UAChB,IAAI;AAAA,YACF,sCACE,QAAQ,SAAS,SAAY,KAAK,QAAQ,IAAI,MAAM,EACtD,KAAK,QAAQ,OAAO;AAAA,UACtB;AAAA;AAEF;AAAA,IACJ;AAAA,EACF;AAEA,WAAS,cAAc,YAA8B;AACnD,QAAI,CAAC,MAAM,GAAG,eAAe,UAAAA,QAAU,MAAM;AAC3C,YAAM,IAAI,MAAM,yBAAyB;AAAA,IAC3C;AAEA,UAAM,QAAQ,SAAS;AAGvB,UAAM,YAAY,KAAK,MAAM,YAAY,IAAI,CAAC;AAG9C,UAAM,MAAM;AAAA,MACV;AAAA,MACA,IAAI,uDAAuB;AAAA,QACzB;AAAA,QACA,kBAAkB,YAAY,IAAI;AAAA,QAClC,aAAa;AAAA,MACf,CAAC;AAAA,IACH;AAGA,UAAM,SAAS,IAAI,YAAY,CAAC;AAChC,UAAM,OAAO,IAAI,SAAS,MAAM;AAChC,SAAK,UAAU,GAAG,cAAc,GAAG,IAAI;AACvC,SAAK,UAAU,GAAG,KAAK,MAAM,YAAY,UAAW,MAAM,GAAG,IAAI;AAGjE,UAAM,aAAa,IAAI;AAAA,MACrB,WAAW;AAAA,MACX,WAAW;AAAA,MACX,WAAW;AAAA,IACb;AACA,UAAM,WAAW,IAAI,WAAW,IAAI,WAAW,MAAM;AACrD,aAAS,IAAI,IAAI,WAAW,MAAM,GAAG,CAAC;AACtC,aAAS,IAAI,YAAY,CAAC;AAE1B,QAAI;AACF,SAAG,KAAK,QAAQ;AAChB;AAAA,IACF,SAAS,GAAY;AACnB,aAAO,MAAM,GAAG,oCAAoC;AAAA,IACtD;AAAA,EACF;AAEA,WAAS,QAAc;AACrB,SAAI,yBAAI,gBAAe,UAAAA,QAAU,MAAM;AACrC,YAAM,WAAW,KAAK,UAAU,EAAE,MAAM,kBAAkB,CAAC;AAC3D,UAAI;AACF,WAAG,KAAK,QAAQ;AAAA,MAClB,SAAS,GAAY;AACnB,eAAO,MAAM,GAAG,8BAA8B;AAAA,MAChD;AAAA,IACF;AACA,6BAAI,MAAM;AACV,SAAK;AAAA,EACP;AAMA,iBAAe,YAA2B;AACxC,UAAM;AAAA,EACR;AAEA,QAAM,YAAY,IAAI;AAAA,IAIpB;AAAA,MACE,MAAM,MAAM,YAAY;AACtB,2BAAmB;AACnB,cAAM,iBAAiB;AAAA,MACzB;AAAA,MAEA,UAAU,OAAO,YAAY;AAC3B,YAAI,EAAE,iBAAiB,aAAa;AAClC,qBAAW,QAAQ,KAAK;AACxB;AAAA,QACF;AAGA,cAAM,QAAQ,SAAS;AACvB,YAAI,CAAC,MAAM,0BAA0B,CAAC,MAAM,qBAAsB;AAElE,YAAI;AACF,wBAAc,KAAK;AAAA,QACrB,SAAS,KAAK;AACZ,iBAAO,MAAM,EAAE,IAAI,GAAG,0CAA0C;AAAA,QAClE;AAAA,MACF;AAAA,MAEA,QAAQ;AACN,cAAM;AAAA,MACR;AAAA,IACF;AAAA,IACA,EAAE,eAAe,EAAE;AAAA,IACnB,EAAE,eAAe,EAAE;AAAA,EACrB;AAEA,SAAO,EAAE,WAAW,UAAU;AAChC;","names":["WebSocket"]}

package/dist/inference/interruption/ws_transport.d.cts

@@ -0,0 +1,33 @@
+/// <reference types="node" resolution-mode="require"/>
+import { TransformStream } from 'stream/web';
+import { InterruptionCacheEntry } from './interruption_cache_entry.js';
+import type { OverlappingSpeechEvent } from './types.js';
+import type { BoundedCache } from './utils.js';
+export interface WsTransportOptions {
+    baseUrl: string;
+    apiKey: string;
+    apiSecret: string;
+    sampleRate: number;
+    threshold: number;
+    minFrames: number;
+    timeout: number;
+    maxRetries?: number;
+}
+export interface WsTransportState {
+    overlapSpeechStarted: boolean;
+    overlapSpeechStartedAt: number | undefined;
+    cache: BoundedCache<number, InterruptionCacheEntry>;
+}
+export interface WsTransportResult {
+    transport: TransformStream<Int16Array | OverlappingSpeechEvent, OverlappingSpeechEvent>;
+    reconnect: () => Promise<void>;
+}
+/**
+ * Creates a WebSocket transport TransformStream for interruption detection.
+ *
+ * This transport receives Int16Array audio slices and outputs InterruptionEvents.
+ * It maintains a persistent WebSocket connection with automatic retry on failure.
+ * Returns both the transport and a reconnect function for option updates.
+ */
+export declare function createWsTransport(options: WsTransportOptions, getState: () => WsTransportState, setState: (partial: Partial<WsTransportState>) => void, updateUserSpeakingSpan?: (entry: InterruptionCacheEntry) => void, onRequestSent?: () => void, getAndResetNumRequests?: () => number): WsTransportResult;
+//# sourceMappingURL=ws_transport.d.ts.map

package/dist/inference/interruption/ws_transport.d.ts

@@ -0,0 +1,33 @@
+/// <reference types="node" resolution-mode="require"/>
+import { TransformStream } from 'stream/web';
+import { InterruptionCacheEntry } from './interruption_cache_entry.js';
+import type { OverlappingSpeechEvent } from './types.js';
+import type { BoundedCache } from './utils.js';
+export interface WsTransportOptions {
+    baseUrl: string;
+    apiKey: string;
+    apiSecret: string;
+    sampleRate: number;
+    threshold: number;
+    minFrames: number;
+    timeout: number;
+    maxRetries?: number;
+}
+export interface WsTransportState {
+    overlapSpeechStarted: boolean;
+    overlapSpeechStartedAt: number | undefined;
+    cache: BoundedCache<number, InterruptionCacheEntry>;
+}
+export interface WsTransportResult {
+    transport: TransformStream<Int16Array | OverlappingSpeechEvent, OverlappingSpeechEvent>;
+    reconnect: () => Promise<void>;
+}
+/**
+ * Creates a WebSocket transport TransformStream for interruption detection.
+ *
+ * This transport receives Int16Array audio slices and outputs InterruptionEvents.
+ * It maintains a persistent WebSocket connection with automatic retry on failure.
+ * Returns both the transport and a reconnect function for option updates.
+ */
+export declare function createWsTransport(options: WsTransportOptions, getState: () => WsTransportState, setState: (partial: Partial<WsTransportState>) => void, updateUserSpeakingSpan?: (entry: InterruptionCacheEntry) => void, onRequestSent?: () => void, getAndResetNumRequests?: () => number): WsTransportResult;
+//# sourceMappingURL=ws_transport.d.ts.map

package/dist/inference/interruption/ws_transport.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"ws_transport.d.ts","sourceRoot":"","sources":["../../../src/inference/interruption/ws_transport.ts"],"names":[],"mappings":";AAGA,OAAO,EAAE,eAAe,EAAE,MAAM,YAAY,CAAC;AAM7C,OAAO,EAAE,sBAAsB,EAAE,MAAM,+BAA+B,CAAC;AACvE,OAAO,KAAK,EAAE,sBAAsB,EAAE,MAAM,YAAY,CAAC;AACzD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,YAAY,CAAC;AAW/C,MAAM,WAAW,kBAAkB;IACjC,OAAO,EAAE,MAAM,CAAC;IAChB,MAAM,EAAE,MAAM,CAAC;IACf,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;IACnB,SAAS,EAAE,MAAM,CAAC;IAClB,SAAS,EAAE,MAAM,CAAC;IAClB,OAAO,EAAE,MAAM,CAAC;IAChB,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB;AAED,MAAM,WAAW,gBAAgB;IAC/B,oBAAoB,EAAE,OAAO,CAAC;IAC9B,sBAAsB,EAAE,MAAM,GAAG,SAAS,CAAC;IAC3C,KAAK,EAAE,YAAY,CAAC,MAAM,EAAE,sBAAsB,CAAC,CAAC;CACrD;AA+DD,MAAM,WAAW,iBAAiB;IAChC,SAAS,EAAE,eAAe,CAAC,UAAU,GAAG,sBAAsB,EAAE,sBAAsB,CAAC,CAAC;IACxF,SAAS,EAAE,MAAM,OAAO,CAAC,IAAI,CAAC,CAAC;CAChC;AAED;;;;;;GAMG;AACH,wBAAgB,iBAAiB,CAC/B,OAAO,EAAE,kBAAkB,EAC3B,QAAQ,EAAE,MAAM,gBAAgB,EAChC,QAAQ,EAAE,CAAC,OAAO,EAAE,OAAO,CAAC,gBAAgB,CAAC,KAAK,IAAI,EACtD,sBAAsB,CAAC,EAAE,CAAC,KAAK,EAAE,sBAAsB,KAAK,IAAI,EAChE,aAAa,CAAC,EAAE,MAAM,IAAI,EAC1B,sBAAsB,CAAC,EAAE,MAAM,MAAM,GACpC,iBAAiB,CA0RnB"}