@dxos/plugin-transcription 0.8.0 → 0.8.1-main.013e445

package/dist/lib/browser/chunk-N3VKF2J3.mjs

@@ -0,0 +1,415 @@
+import {
+  TRANSCRIPTION_URL,
+  TranscriptBlock
+} from "./chunk-YRW4MNRE.mjs";
+
+// packages/plugins/experimental/plugin-transcription/src/transcriber/media-stream-recorder.ts
+import { MediaRecorder, register } from "extendable-media-recorder";
+import { WaveFile } from "wavefile";
+import { synchronized } from "@dxos/async";
+import { invariant } from "@dxos/invariant";
+import { log } from "@dxos/log";
+import { trace } from "@dxos/tracing";
+function _ts_decorate(decorators, target, key, desc) {
+  var c = arguments.length, r = c < 3 ? target : desc === null ? desc = Object.getOwnPropertyDescriptor(target, key) : desc, d;
+  if (typeof Reflect === "object" && typeof Reflect.decorate === "function") r = Reflect.decorate(decorators, target, key, desc);
+  else for (var i = decorators.length - 1; i >= 0; i--) if (d = decorators[i]) r = (c < 3 ? d(r) : c > 3 ? d(target, key, r) : d(target, key)) || r;
+  return c > 3 && r && Object.defineProperty(target, key, r), r;
+}
+var __dxlog_file = "/home/runner/work/dxos/dxos/packages/plugins/experimental/plugin-transcription/src/transcriber/media-stream-recorder.ts";
+var initializingPromise;
+var initializeExtendableMediaRecorder = async () => {
+  const { connect } = await import("extendable-media-recorder-wav-encoder");
+  await register(await connect());
+};
+var MediaStreamRecorder = class {
+  constructor({ mediaStreamTrack, interval }) {
+    this._mediaRecorder = void 0;
+    this._header = void 0;
+    this._mediaStreamTrack = mediaStreamTrack;
+    this._interval = interval;
+  }
+  get wavConfig() {
+    const settings = this._mediaStreamTrack.getSettings();
+    return {
+      channels: settings.channelCount ?? 1,
+      sampleRate: settings.sampleRate ?? 16e3,
+      bitDepthCode: settings.sampleSize ? String(settings.sampleSize) : "16"
+    };
+  }
+  get mediaStreamTrack() {
+    return this._mediaStreamTrack;
+  }
+  setOnChunk(onChunk) {
+    this._onChunk = onChunk;
+  }
+  async start() {
+    await (initializingPromise ??= initializeExtendableMediaRecorder()).catch((err) => log.info("initializeExtendableMediaRecorder", {
+      err
+    }, {
+      F: __dxlog_file,
+      L: 62,
+      S: this,
+      C: (f, a) => f(...a)
+    }));
+    if (!this._mediaRecorder) {
+      const stream = new MediaStream([
+        this._mediaStreamTrack
+      ]);
+      this._mediaRecorder = new MediaRecorder(stream, {
+        mimeType: "audio/wav"
+      });
+    }
+    if (this._mediaRecorder.state === "recording") {
+      return;
+    }
+    invariant(this._onChunk, "MediaStreamRecorder: onChunk is not set", {
+      F: __dxlog_file,
+      L: 73,
+      S: this,
+      A: [
+        "this._onChunk",
+        "'MediaStreamRecorder: onChunk is not set'"
+      ]
+    });
+    this._mediaRecorder.ondataavailable = (event) => this._ondataavailable(event);
+    this._mediaRecorder.start(this._interval);
+  }
+  async stop() {
+    if (this._mediaRecorder?.state !== "recording") {
+      return;
+    }
+    this._mediaRecorder.stop();
+  }
+  async _ondataavailable(event) {
+    const blob = event.data;
+    const uint8Array = new Uint8Array(await blob.arrayBuffer());
+    const isHeader = uint8Array[0] === 82 && // R
+    uint8Array[1] === 73 && // I
+    uint8Array[2] === 70 && // F
+    uint8Array[3] === 70;
+    let wav;
+    if (isHeader) {
+      wav = new WaveFile(uint8Array);
+      this._header = uint8Array.slice(0, 44);
+    } else if (this._header) {
+      wav = new WaveFile(new Uint8Array([
+        ...this._header,
+        ...uint8Array
+      ]));
+    } else {
+      log.warn("MediaStreamRecorder: no header", void 0, {
+        F: __dxlog_file,
+        L: 104,
+        S: this,
+        C: (f, a) => f(...a)
+      });
+      return;
+    }
+    this._onChunk({
+      timestamp: Date.now(),
+      data: wav.getSamples()
+    });
+  }
+};
+_ts_decorate([
+  synchronized
+], MediaStreamRecorder.prototype, "_ondataavailable", null);
+MediaStreamRecorder = _ts_decorate([
+  trace.resource()
+], MediaStreamRecorder);
+
+// packages/plugins/experimental/plugin-transcription/src/transcriber/summarize.ts
+import { Message } from "@dxos/artifact";
+import { DEFAULT_EDGE_MODEL, MixedStreamParser } from "@dxos/assistant";
+import { QueueImpl } from "@dxos/echo-db";
+import { createStatic, isInstanceOf } from "@dxos/echo-schema";
+import { invariant as invariant2 } from "@dxos/invariant";
+import { DXN } from "@dxos/keys";
+import { log as log2 } from "@dxos/log";
+var __dxlog_file2 = "/home/runner/work/dxos/dxos/packages/plugins/experimental/plugin-transcription/src/transcriber/summarize.ts";
+var summarizeTranscript = async (edge, ai, transcript, context) => {
+  invariant2(transcript.queue, "No queue for transcript", {
+    F: __dxlog_file2,
+    L: 23,
+    S: void 0,
+    A: [
+      "transcript.queue",
+      "'No queue for transcript'"
+    ]
+  });
+  const queue = new QueueImpl(edge, DXN.parse(transcript.queue));
+  await queue.refresh();
+  const content = queue.items.filter((block) => isInstanceOf(TranscriptBlock, block)).map((block) => `${block.authorName}: ${block.segments.map((segment) => segment.text).join("\n")}`).join("\n\n");
+  const parser = new MixedStreamParser();
+  log2.info("summarizing transcript", {
+    blockCount: queue.items.length
+  }, {
+    F: __dxlog_file2,
+    L: 34,
+    S: void 0,
+    C: (f, a) => f(...a)
+  });
+  const output = await parser.parse(await ai.exec({
+    model: DEFAULT_EDGE_MODEL,
+    systemPrompt: SUMMARIZE_PROMPT,
+    history: [
+      createStatic(Message, {
+        role: "user",
+        content: [
+          {
+            type: "text",
+            text: content
+          }
+        ]
+      })
+    ]
+  }));
+  log2.info("transcript summary", {
+    output
+  }, {
+    F: __dxlog_file2,
+    L: 43,
+    S: void 0,
+    C: (f, a) => f(...a)
+  });
+  invariant2(output[0].content[0].type === "text", "Expected text content", {
+    F: __dxlog_file2,
+    L: 44,
+    S: void 0,
+    A: [
+      "output[0].content[0].type === 'text'",
+      "'Expected text content'"
+    ]
+  });
+  return output[0].content[0].text;
+};
+var SUMMARIZE_PROMPT = `
+  # Goal
+  Create a markdown summary of the transcript provided.
+
+  # Formatting
+  - Format the summary as a markdown document without extra comments like "Here is the summary of the transcript:".
+  - Use markdown formatting for headings and bullet points.
+  - Format the summary as a list of key points and takeaways.
+  - All names of people should be in bold.
+
+  # Note Taking
+  - Correlate items in the summary with the person of origin to build a coherent narrative.
+  - Include short quotes verbatim where appropriate. Especially when concerned with design decisions and problem descriptions.
+
+  # Tasks
+  At the end of the summary include tasks.
+  Extract only the tasks that are:
+  - Directly actionable.
+  - Clearly assigned to a person or team (or can easily be inferred).
+  - Strongly implied by the conversation and/or user note (no speculative tasks).
+  - Specific enough that someone reading them would know exactly what to do next.
+
+  Format all tasks as markdown checkboxes using the syntax:
+  - [ ] Task description.
+
+  Additional information can be included (indented).
+
+  If no actionable tasks are found, omit this tasks section.
+`;
+
+// packages/plugins/experimental/plugin-transcription/src/util/get-time-str.ts
+var getTimeStr = (timestamp) => new Date(timestamp ?? Date.now()).toLocaleTimeString("en-US", {
+  timeZone: "UTC",
+  hour12: true,
+  hour: "numeric",
+  minute: "2-digit",
+  second: "2-digit"
+});
+
+// packages/plugins/experimental/plugin-transcription/src/util/random-queue-dxn.ts
+import { ObjectId } from "@dxos/echo-schema";
+import { DXN as DXN2, QueueSubspaceTags, SpaceId } from "@dxos/keys";
+var randomQueueDxn = (spaceId) => new DXN2(DXN2.kind.QUEUE, [
+  QueueSubspaceTags.DATA,
+  spaceId ?? SpaceId.random(),
+  ObjectId.random()
+]);
+
+// packages/plugins/experimental/plugin-transcription/src/util/array.ts
+var mergeFloat64Arrays = (arrays) => {
+  const newArray = new Float64Array(arrays.reduce((acc, array) => array.length + acc, 0));
+  let offset = 0;
+  for (const array of arrays) {
+    newArray.set(array, offset);
+    offset += array.length;
+  }
+  return newArray;
+};
+
+// packages/plugins/experimental/plugin-transcription/src/transcriber/transcriber.ts
+import { WaveFile as WaveFile2 } from "wavefile";
+import { DeferredTask, Trigger } from "@dxos/async";
+import { LifecycleState, Resource } from "@dxos/context";
+import { log as log3 } from "@dxos/log";
+import { trace as trace2 } from "@dxos/tracing";
+function _ts_decorate2(decorators, target, key, desc) {
+  var c = arguments.length, r = c < 3 ? target : desc === null ? desc = Object.getOwnPropertyDescriptor(target, key) : desc, d;
+  if (typeof Reflect === "object" && typeof Reflect.decorate === "function") r = Reflect.decorate(decorators, target, key, desc);
+  else for (var i = decorators.length - 1; i >= 0; i--) if (d = decorators[i]) r = (c < 3 ? d(r) : c > 3 ? d(target, key, r) : d(target, key)) || r;
+  return c > 3 && r && Object.defineProperty(target, key, r), r;
+}
+var __dxlog_file3 = "/home/runner/work/dxos/dxos/packages/plugins/experimental/plugin-transcription/src/transcriber/transcriber.ts";
+var Transcriber = class extends Resource {
+  constructor({ config, recorder, onSegments }) {
+    super();
+    this._audioChunks = [];
+    this._lastUsedTimestamp = 0;
+    this._openTrigger = new Trigger({
+      autoReset: true
+    });
+    this._recording = false;
+    this._transcribeTask = void 0;
+    this._config = config;
+    this._recorder = recorder;
+    this._onSegments = onSegments;
+  }
+  async _open(ctx) {
+    this._recorder.setOnChunk((chunk) => this._saveAudioChunk(chunk));
+    await this._recorder.start();
+    this._transcribeTask = new DeferredTask(ctx, async () => this._transcribe());
+    this._openTrigger.wake();
+  }
+  async _close() {
+    this._recording = false;
+    this._transcribeTask = void 0;
+    await this._recorder.stop();
+  }
+  startChunksRecording() {
+    if (this._lifecycleState !== LifecycleState.OPEN) {
+      return;
+    }
+    this._recording = true;
+  }
+  stopChunksRecording() {
+    if (this._lifecycleState !== LifecycleState.OPEN || !this._recording) {
+      return;
+    }
+    this._recording = false;
+    this._transcribeTask?.schedule();
+  }
+  _saveAudioChunk(chunk) {
+    log3("saving audio chunk", {
+      chunk
+    }, {
+      F: __dxlog_file3,
+      L: 124,
+      S: this,
+      C: (f, a) => f(...a)
+    });
+    this._audioChunks.push(chunk);
+    if (!this._recording && (!this._transcribeTask || !this._transcribeTask.scheduled) && this._audioChunks.length >= this._config.prefixBufferChunksAmount) {
+      this._audioChunks = this._dropOldChunks();
+    } else if (this._audioChunks.length >= this._config.transcribeAfterChunksAmount && this._recording) {
+      this._transcribeTask?.schedule();
+    }
+  }
+  async _transcribe() {
+    log3("transcribing", {
+      chunks: this._audioChunks.length
+    }, {
+      F: __dxlog_file3,
+      L: 140,
+      S: this,
+      C: (f, a) => f(...a)
+    });
+    const chunks = this._audioChunks;
+    this._audioChunks = this._dropOldChunks();
+    if (chunks.length === 0) {
+      return;
+    }
+    const audio = await this._mergeAudioChunks(chunks);
+    const segments = await this._fetchTranscription(audio);
+    if (!Array.isArray(segments) || segments.length === 0) {
+      return;
+    }
+    const alignedSegments = this._alignSegments(segments, chunks);
+    if (alignedSegments.length > 0) {
+      await this._onSegments(alignedSegments);
+    }
+  }
+  _dropOldChunks() {
+    return this._config.prefixBufferChunksAmount > 0 ? this._audioChunks.slice(-this._config.prefixBufferChunksAmount) : [];
+  }
+  async _mergeAudioChunks(chunks) {
+    const file = new WaveFile2();
+    const wavConfig = this._recorder.wavConfig;
+    file.fromScratch(wavConfig.channels, wavConfig.sampleRate, wavConfig.bitDepthCode, mergeFloat64Arrays(chunks.map(({ data }) => data)));
+    return file.toBase64();
+  }
+  async _fetchTranscription(audio) {
+    if (audio.length === 0) {
+      this._audioChunks = [];
+      throw new Error("No audio to send for transcribing");
+    }
+    const response = await fetch(`${TRANSCRIPTION_URL}/transcribe`, {
+      method: "POST",
+      body: JSON.stringify({
+        audio
+      }),
+      headers: {
+        "Content-Type": "application/json"
+      }
+    });
+    if (!response.ok) {
+      this._audioChunks = [];
+      throw new Error("Failed to transcribe");
+    }
+    const { segments } = await response.json();
+    log3.info("transcription response", {
+      segments,
+      string: segments.map((segments2) => segments2.text).join(" ")
+    }, {
+      F: __dxlog_file3,
+      L: 206,
+      S: this,
+      C: (f, a) => f(...a)
+    });
+    return segments;
+  }
+  _alignSegments(segments, originalChunks) {
+    const zeroTimestamp = originalChunks.at(0).timestamp;
+    const filteredSegments = segments.filter((segment) => zeroTimestamp + segment.end * 1e3 >= this._lastUsedTimestamp).map((segment) => {
+      const words = segment.words.filter((word) => zeroTimestamp + word.start * 1e3 >= this._lastUsedTimestamp);
+      return {
+        ...segment,
+        words,
+        text: words.map((word) => word.word).join(""),
+        start: words.at(0)?.start ?? segment.end
+      };
+    }).filter((segment) => segment.words.length > 0);
+    if (filteredSegments.length === 0) {
+      return [];
+    }
+    this._lastUsedTimestamp = zeroTimestamp + filteredSegments.at(-1).end * 1e3;
+    return filteredSegments.map((segment) => ({
+      started: new Date(zeroTimestamp + segment.start * 1e3),
+      text: segment.text
+    }));
+  }
+};
+_ts_decorate2([
+  trace2.span({
+    showInBrowserTimeline: true
+  })
+], Transcriber.prototype, "_mergeAudioChunks", null);
+_ts_decorate2([
+  trace2.span({
+    showInBrowserTimeline: true
+  })
+], Transcriber.prototype, "_fetchTranscription", null);
+
+export {
+  MediaStreamRecorder,
+  summarizeTranscript,
+  getTimeStr,
+  randomQueueDxn,
+  Transcriber
+};
+//# sourceMappingURL=chunk-N3VKF2J3.mjs.map

package/dist/lib/browser/chunk-N3VKF2J3.mjs.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../../src/transcriber/media-stream-recorder.ts", "../../../src/transcriber/summarize.ts", "../../../src/util/get-time-str.ts", "../../../src/util/random-queue-dxn.ts", "../../../src/util/array.ts", "../../../src/transcriber/transcriber.ts"],
+  "sourcesContent": ["//\n// Copyright 2025 DXOS.org\n//\n\nimport { type IBlobEvent, type IMediaRecorder, MediaRecorder, register } from 'extendable-media-recorder';\nimport { WaveFile } from 'wavefile';\n\nimport { synchronized } from '@dxos/async';\nimport { invariant } from '@dxos/invariant';\nimport { log } from '@dxos/log';\nimport { trace } from '@dxos/tracing';\n\nimport { type WavConfig, type AudioChunk, type AudioRecorder } from './audio-recorder';\n\nlet initializingPromise: Promise<void> | undefined;\n\nconst initializeExtendableMediaRecorder = async () => {\n  const { connect } = await import('extendable-media-recorder-wav-encoder');\n  await register(await connect());\n};\n\n/**\n * Recorder that uses the MediaContext API and AudioNode API to record audio.\n *\n * It records MediaStream using https://www.npmjs.com/package/extendable-media-recorder.\n */\n@trace.resource()\nexport class MediaStreamRecorder implements AudioRecorder {\n  private readonly _mediaStreamTrack: MediaStreamTrack;\n  /**\n   * Default MediaRecorder implementation do not support wav encoding.\n   */\n  private readonly _interval: number;\n  private _mediaRecorder?: IMediaRecorder = undefined;\n  private _header?: Uint8Array = undefined;\n  private _onChunk?: (chunk: AudioChunk) => void;\n\n  constructor({ mediaStreamTrack, interval }: { mediaStreamTrack: MediaStreamTrack; interval: number }) {\n    this._mediaStreamTrack = mediaStreamTrack;\n    this._interval = interval;\n  }\n\n  get wavConfig(): WavConfig {\n    const settings = this._mediaStreamTrack.getSettings();\n    return {\n      channels: settings.channelCount ?? 1,\n      sampleRate: settings.sampleRate ?? 16000,\n      bitDepthCode: settings.sampleSize ? String(settings.sampleSize) : '16',\n    };\n  }\n\n  get mediaStreamTrack() {\n    return this._mediaStreamTrack;\n  }\n\n  setOnChunk(onChunk: (chunk: AudioChunk) => void): void {\n    this._onChunk = onChunk;\n  }\n\n  async start() {\n    await (initializingPromise ??= initializeExtendableMediaRecorder()).catch((err) =>\n      log.info('initializeExtendableMediaRecorder', { err }),\n    );\n\n    if (!this._mediaRecorder) {\n      const stream = new MediaStream([this._mediaStreamTrack]);\n      this._mediaRecorder = new MediaRecorder(stream, { mimeType: 'audio/wav' });\n    }\n\n    if (this._mediaRecorder.state === 'recording') {\n      return;\n    }\n    invariant(this._onChunk, 'MediaStreamRecorder: onChunk is not set');\n    this._mediaRecorder.ondataavailable = (event) => this._ondataavailable(event);\n    this._mediaRecorder.start(this._interval);\n  }\n\n  async stop() {\n    if (this._mediaRecorder?.state !== 'recording') {\n      return;\n    }\n    this._mediaRecorder.stop();\n  }\n\n  @synchronized\n  private async _ondataavailable(event: IBlobEvent) {\n    const blob = event.data;\n    const uint8Array = new Uint8Array(await blob.arrayBuffer());\n\n    // First chunk from the MediaRecorder has a header.\n    const isHeader =\n      uint8Array[0] === 0x52 && // R\n      uint8Array[1] === 0x49 && // I\n      uint8Array[2] === 0x46 && // F\n      uint8Array[3] === 0x46; // F\n    let wav: WaveFile;\n\n    if (isHeader) {\n      wav = new WaveFile(uint8Array);\n      this._header = uint8Array.slice(0, 44);\n    } else if (this._header) {\n      wav = new WaveFile(new Uint8Array([...this._header, ...uint8Array]));\n    } else {\n      log.warn('MediaStreamRecorder: no header');\n      return;\n    }\n\n    this._onChunk!({\n      timestamp: Date.now(),\n      data: wav.getSamples(),\n    });\n  }\n}\n", "//\n// Copyright 2024 DXOS.org\n//\n\nimport { Message } from '@dxos/artifact';\nimport { DEFAULT_EDGE_MODEL, MixedStreamParser, type AIServiceClient } from '@dxos/assistant';\nimport { QueueImpl } from '@dxos/echo-db';\nimport { createStatic, isInstanceOf } from '@dxos/echo-schema';\nimport { invariant } from '@dxos/invariant';\nimport { DXN } from '@dxos/keys';\nimport { log } from '@dxos/log';\nimport { type EdgeHttpClient } from '@dxos/react-edge-client';\n\nimport { TranscriptBlock, type TranscriptType } from '../types';\n\nexport const summarizeTranscript = async (\n  edge: EdgeHttpClient,\n  ai: AIServiceClient,\n  transcript: TranscriptType,\n  // TODO(wittjosiah): Add additional context into the prompt.\n  context?: string,\n): Promise<string> => {\n  invariant(transcript.queue, 'No queue for transcript');\n\n  const queue = new QueueImpl(edge, DXN.parse(transcript.queue));\n  await queue.refresh();\n  const content = queue.items\n    .filter((block) => isInstanceOf(TranscriptBlock, block))\n    .map((block) => `${block.authorName}: ${block.segments.map((segment) => segment.text).join('\\n')}`)\n    .join('\\n\\n');\n\n  const parser = new MixedStreamParser();\n\n  log.info('summarizing transcript', { blockCount: queue.items.length });\n  const output = await parser.parse(\n    await ai.exec({\n      model: DEFAULT_EDGE_MODEL,\n      systemPrompt: SUMMARIZE_PROMPT,\n      history: [createStatic(Message, { role: 'user', content: [{ type: 'text', text: content }] })],\n    }),\n  );\n\n  log.info('transcript summary', { output });\n  invariant(output[0].content[0].type === 'text', 'Expected text content');\n  return output[0].content[0].text;\n};\n\n// TODO(dmaretskyi): Add example to set consistent structure for the summary.\nconst SUMMARIZE_PROMPT = `\n  # Goal\n  Create a markdown summary of the transcript provided.\n\n  # Formatting\n  - Format the summary as a markdown document without extra comments like \"Here is the summary of the transcript:\".\n  - Use markdown formatting for headings and bullet points.\n  - Format the summary as a list of key points and takeaways.\n  - All names of people should be in bold.\n\n  # Note Taking\n  - Correlate items in the summary with the person of origin to build a coherent narrative.\n  - Include short quotes verbatim where appropriate. Especially when concerned with design decisions and problem descriptions.\n\n  # Tasks\n  At the end of the summary include tasks.\n  Extract only the tasks that are:\n  - Directly actionable.\n  - Clearly assigned to a person or team (or can easily be inferred).\n  - Strongly implied by the conversation and/or user note (no speculative tasks).\n  - Specific enough that someone reading them would know exactly what to do next.\n\n  Format all tasks as markdown checkboxes using the syntax:\n  - [ ] Task description.\n\n  Additional information can be included (indented).\n\n  If no actionable tasks are found, omit this tasks section.\n`;\n", "//\n// Copyright 2025 DXOS.org\n//\n\nexport const getTimeStr = (timestamp?: number) =>\n  new Date(timestamp ?? Date.now()).toLocaleTimeString('en-US', {\n    timeZone: 'UTC',\n    hour12: true,\n    hour: 'numeric',\n    minute: '2-digit',\n    second: '2-digit',\n  });\n", "//\n// Copyright 2025 DXOS.org\n//\n\nimport { ObjectId } from '@dxos/echo-schema';\nimport { DXN, QueueSubspaceTags, SpaceId } from '@dxos/keys';\n\nexport const randomQueueDxn = (spaceId?: SpaceId) =>\n  new DXN(DXN.kind.QUEUE, [QueueSubspaceTags.DATA, spaceId ?? SpaceId.random(), ObjectId.random()]);\n", "//\n// Copyright 2025 DXOS.org\n//\n\nexport const mergeFloat64Arrays = (arrays: Float64Array[]) => {\n  const newArray = new Float64Array(arrays.reduce((acc, array) => array.length + acc, 0));\n  let offset = 0;\n  for (const array of arrays) {\n    newArray.set(array, offset);\n    offset += array.length;\n  }\n\n  return newArray;\n};\n", "//\n// Copyright 2025 DXOS.org\n//\n\nimport { WaveFile } from 'wavefile';\n\nimport { DeferredTask, Trigger } from '@dxos/async';\nimport { type Context, LifecycleState, Resource } from '@dxos/context';\nimport { log } from '@dxos/log';\nimport { trace } from '@dxos/tracing';\n\nimport { type AudioRecorder, type AudioChunk } from './audio-recorder';\nimport { TRANSCRIPTION_URL, type TranscriptSegment } from '../types';\nimport { mergeFloat64Arrays } from '../util';\n\ntype WhisperWord = {\n  word: string;\n\n  /**\n   * Time in seconds from the start of the audio.\n   */\n  start: number;\n\n  /**\n   * Time in seconds from the start of the audio.\n   */\n  end: number;\n};\n\ntype WhisperSegment = {\n  text: string;\n\n  /**\n   * Time in seconds from the start of the audio.\n   */\n  start: number;\n\n  /**\n   * Time in seconds from the start of the audio.\n   */\n  end: number;\n\n  /**\n   * Probability of no speech in the segment.\n   */\n  no_speech_prob: number;\n\n  words: WhisperWord[];\n};\n\nexport type TranscribeConfig = {\n  /**\n   * Number of chunks to transcribe automatically after.\n   */\n  transcribeAfterChunksAmount: number;\n\n  /**\n   * Number of chunks to save before the user starts speaking.\n   * This is needed to catch the beginning of the user's speech.\n   */\n  prefixBufferChunksAmount: number;\n};\n\nexport type TranscriberParams = {\n  config: TranscribeConfig;\n  recorder: AudioRecorder;\n  onSegments: (segments: TranscriptSegment[]) => Promise<void>;\n};\n\n/**\n * Handles transcription of audio chunks.\n * If user is not speaking, the last `minChunksAmount` chunks are saved and transcribed.\n * If user is speaking, the chunks are added to the buffer until the user is done talking.\n */\nexport class Transcriber extends Resource {\n  private _audioChunks: AudioChunk[] = [];\n  private _lastUsedTimestamp = 0;\n\n  private readonly _openTrigger = new Trigger({ autoReset: true });\n\n  private readonly _config: TranscribeConfig;\n  private readonly _recorder: AudioRecorder;\n  private readonly _onSegments: TranscriberParams['onSegments'];\n\n  private _recording = false;\n  private _transcribeTask?: DeferredTask = undefined;\n\n  constructor({ config, recorder, onSegments }: TranscriberParams) {\n    super();\n    this._config = config;\n    this._recorder = recorder;\n    this._onSegments = onSegments;\n  }\n\n  protected override async _open(ctx: Context) {\n    this._recorder.setOnChunk((chunk) => this._saveAudioChunk(chunk));\n    await this._recorder.start();\n    this._transcribeTask = new DeferredTask(ctx, async () => this._transcribe());\n    this._openTrigger.wake();\n  }\n\n  protected override async _close() {\n    this._recording = false;\n    this._transcribeTask = undefined;\n    await this._recorder.stop();\n  }\n\n  startChunksRecording() {\n    if (this._lifecycleState !== LifecycleState.OPEN) {\n      return;\n    }\n    this._recording = true;\n  }\n\n  stopChunksRecording() {\n    if (this._lifecycleState !== LifecycleState.OPEN || !this._recording) {\n      return;\n    }\n    this._recording = false;\n    this._transcribeTask?.schedule();\n  }\n\n  private _saveAudioChunk(chunk: AudioChunk) {\n    log('saving audio chunk', { chunk });\n    this._audioChunks.push(chunk);\n\n    // Clean the buffer if the user is not speaking and the transcription task is not scheduled.\n    if (\n      !this._recording &&\n      (!this._transcribeTask || !this._transcribeTask.scheduled) &&\n      this._audioChunks.length >= this._config.prefixBufferChunksAmount\n    ) {\n      this._audioChunks = this._dropOldChunks();\n    } else if (this._audioChunks.length >= this._config.transcribeAfterChunksAmount && this._recording) {\n      this._transcribeTask?.schedule();\n    }\n  }\n\n  private async _transcribe() {\n    log('transcribing', { chunks: this._audioChunks.length });\n    const chunks = this._audioChunks;\n    this._audioChunks = this._dropOldChunks();\n    if (chunks.length === 0) {\n      return;\n    }\n\n    const audio = await this._mergeAudioChunks(chunks);\n    const segments = await this._fetchTranscription(audio);\n    if (!Array.isArray(segments) || segments.length === 0) {\n      return;\n    }\n\n    const alignedSegments = this._alignSegments(segments, chunks);\n    if (alignedSegments.length > 0) {\n      await this._onSegments(alignedSegments);\n    }\n  }\n\n  private _dropOldChunks() {\n    return this._config.prefixBufferChunksAmount > 0\n      ? this._audioChunks.slice(-this._config.prefixBufferChunksAmount)\n      : [];\n  }\n\n  @trace.span({ showInBrowserTimeline: true })\n  private async _mergeAudioChunks(chunks: AudioChunk[]) {\n    const file = new WaveFile();\n    const wavConfig = this._recorder.wavConfig;\n\n    file.fromScratch(\n      wavConfig.channels,\n      wavConfig.sampleRate,\n      wavConfig.bitDepthCode,\n      mergeFloat64Arrays(chunks.map(({ data }) => data)),\n    );\n\n    return file.toBase64();\n  }\n\n  @trace.span({ showInBrowserTimeline: true })\n  private async _fetchTranscription(audio: string) {\n    if (audio.length === 0) {\n      this._audioChunks = [];\n      throw new Error('No audio to send for transcribing');\n    }\n\n    const response = await fetch(`${TRANSCRIPTION_URL}/transcribe`, {\n      method: 'POST',\n      body: JSON.stringify({ audio }),\n      headers: {\n        'Content-Type': 'application/json',\n      },\n    });\n\n    if (!response.ok) {\n      this._audioChunks = [];\n      throw new Error('Failed to transcribe');\n    }\n\n    const {\n      segments,\n    }: {\n      segments: WhisperSegment[];\n    } = await response.json();\n\n    log.info('transcription response', {\n      segments,\n      string: segments.map((segments) => segments.text).join(' '),\n    });\n\n    return segments;\n  }\n\n  private _alignSegments(segments: WhisperSegment[], originalChunks: AudioChunk[]): TranscriptSegment[] {\n    // Absolute zero for all relative timestamps in the segments.\n    const zeroTimestamp = originalChunks.at(0)!.timestamp;\n\n    const filteredSegments = segments\n      // Use segments that end after the last used timestamp.\n      // Segments could overlap, so we need to filter words that starts after the last used timestamp.\n      .filter((segment) => zeroTimestamp + segment.end * 1_000 >= this._lastUsedTimestamp)\n      // Use words that starts after the last used timestamp.\n      .map((segment) => {\n        const words = segment.words.filter((word) => zeroTimestamp + word.start * 1_000 >= this._lastUsedTimestamp);\n        return {\n          ...segment,\n          words,\n          text: words.map((word) => word.word).join(''),\n          start: words.at(0)?.start ?? segment.end,\n        };\n      })\n      .filter((segment) => segment.words.length > 0);\n\n    if (filteredSegments.length === 0) {\n      return [];\n    }\n\n    // Update last timestamp.\n    this._lastUsedTimestamp = zeroTimestamp + filteredSegments.at(-1)!.end * 1_000;\n\n    // Add absolute timestamp to each segment.\n    return filteredSegments.map((segment) => ({\n      started: new Date(zeroTimestamp + segment.start * 1_000),\n      text: segment.text,\n    }));\n  }\n}\n"],
+  "mappings": ";;;;;;AAIA,SAA+CA,eAAeC,gBAAgB;AAC9E,SAASC,gBAAgB;AAEzB,SAASC,oBAAoB;AAC7B,SAASC,iBAAiB;AAC1B,SAASC,WAAW;AACpB,SAASC,aAAa;;;;;;;;AAItB,IAAIC;AAEJ,IAAMC,oCAAoC,YAAA;AACxC,QAAM,EAAEC,QAAO,IAAK,MAAM,OAAO,uCAAA;AACjC,QAAMR,SAAS,MAAMQ,QAAAA,CAAAA;AACvB;AAQO,IAAMC,sBAAN,MAAMA;EAUXC,YAAY,EAAEC,kBAAkBC,SAAQ,GAA8D;AAJ9FC,0BAAkCC;AAClCC,mBAAuBD;AAI7B,SAAKE,oBAAoBL;AACzB,SAAKM,YAAYL;EACnB;EAEA,IAAIM,YAAuB;AACzB,UAAMC,WAAW,KAAKH,kBAAkBI,YAAW;AACnD,WAAO;MACLC,UAAUF,SAASG,gBAAgB;MACnCC,YAAYJ,SAASI,cAAc;MACnCC,cAAcL,SAASM,aAAaC,OAAOP,SAASM,UAAU,IAAI;IACpE;EACF;EAEA,IAAId,mBAAmB;AACrB,WAAO,KAAKK;EACd;EAEAW,WAAWC,SAA4C;AACrD,SAAKC,WAAWD;EAClB;EAEA,MAAME,QAAQ;AACZ,WAAOxB,wBAAwBC,kCAAAA,GAAqCwB,MAAM,CAACC,QACzE5B,IAAI6B,KAAK,qCAAqC;MAAED;IAAI,GAAA;;;;;;AAGtD,QAAI,CAAC,KAAKnB,gBAAgB;AACxB,YAAMqB,SAAS,IAAIC,YAAY;QAAC,KAAKnB;OAAkB;AACvD,WAAKH,iBAAiB,IAAId,cAAcmC,QAAQ;QAAEE,UAAU;MAAY,CAAA;IAC1E;AAEA,QAAI,KAAKvB,eAAewB,UAAU,aAAa;AAC7C;IACF;AACAlC,cAAU,KAAK0B,UAAU,2CAAA;;;;;;;;;AACzB,SAAKhB,eAAeyB,kBAAkB,CAACC,UAAU,KAAKC,iBAAiBD,KAAAA;AACvE,SAAK1B,eAAeiB,MAAM,KAAKb,SAAS;EAC1C;EAEA,MAAMwB,OAAO;AACX,QAAI,KAAK5B,gBAAgBwB,UAAU,aAAa;AAC9C;IACF;AACA,SAAKxB,eAAe4B,KAAI;EAC1B;EAEA,MACcD,iBAAiBD,OAAmB;AAChD,UAAMG,OAAOH,MAAMI;AACnB,UAAMC,aAAa,IAAIC,WAAW,MAAMH,KAAKI,YAAW,CAAA;AAGxD,UAAMC,WACJH,WAAW,CAAA,MAAO;IAClBA,WAAW,CAAA,MAAO;IAClBA,WAAW,CAAA,MAAO;IAClBA,WAAW,CAAA,MAAO;AACpB,QAAII;AAEJ,QAAID,UAAU;AACZC,YAAM,IAAI/C,SAAS2C,UAAAA;AACnB,WAAK7B,UAAU6B,WAAWK,MAAM,GAAG,EAAA;IACrC,WAAW,KAAKlC,SAAS;AACvBiC,YAAM,IAAI/C,SAAS,IAAI4C,WAAW;WAAI,KAAK9B;WAAY6B;OAAW,CAAA;IACpE,OAAO;AACLxC,UAAI8C,KAAK,kCAAA,QAAA;;;;;;AACT;IACF;AAEA,SAAKrB,SAAU;MACbsB,WAAWC,KAAKC,IAAG;MACnBV,MAAMK,IAAIM,WAAU;IACtB,CAAA;EACF;AACF;;EA5BGpD;GAzDUO,oBAAAA,WAAAA,oBAAAA,IAAAA;AAAAA,sBAAAA,aAAAA;EADZJ,MAAMkD,SAAQ;GACF9C,mBAAAA;;;ACvBb,SAAS+C,eAAe;AACxB,SAASC,oBAAoBC,yBAA+C;AAC5E,SAASC,iBAAiB;AAC1B,SAASC,cAAcC,oBAAoB;AAC3C,SAASC,aAAAA,kBAAiB;AAC1B,SAASC,WAAW;AACpB,SAASC,OAAAA,YAAW;;AAKb,IAAMC,sBAAsB,OACjCC,MACAC,IACAC,YAEAC,YAAAA;AAEAC,EAAAA,WAAUF,WAAWG,OAAO,2BAAA;;;;;;;;;AAE5B,QAAMA,QAAQ,IAAIC,UAAUN,MAAMO,IAAIC,MAAMN,WAAWG,KAAK,CAAA;AAC5D,QAAMA,MAAMI,QAAO;AACnB,QAAMC,UAAUL,MAAMM,MACnBC,OAAO,CAACC,UAAUC,aAAaC,iBAAiBF,KAAAA,CAAAA,EAChDG,IAAI,CAACH,UAAU,GAAGA,MAAMI,UAAU,KAAKJ,MAAMK,SAASF,IAAI,CAACG,YAAYA,QAAQC,IAAI,EAAEC,KAAK,IAAA,CAAA,EAAO,EACjGA,KAAK,MAAA;AAER,QAAMC,SAAS,IAAIC,kBAAAA;AAEnBC,EAAAA,KAAIC,KAAK,0BAA0B;IAAEC,YAAYrB,MAAMM,MAAMgB;EAAO,GAAA;;;;;;AACpE,QAAMC,SAAS,MAAMN,OAAOd,MAC1B,MAAMP,GAAG4B,KAAK;IACZC,OAAOC;IACPC,cAAcC;IACdC,SAAS;MAACC,aAAaC,SAAS;QAAEC,MAAM;QAAQ3B,SAAS;UAAC;YAAE4B,MAAM;YAAQlB,MAAMV;UAAQ;;MAAG,CAAA;;EAC7F,CAAA,CAAA;AAGFc,EAAAA,KAAIC,KAAK,sBAAsB;IAAEG;EAAO,GAAA;;;;;;AACxCxB,EAAAA,WAAUwB,OAAO,CAAA,EAAGlB,QAAQ,CAAA,EAAG4B,SAAS,QAAQ,yBAAA;;;;;;;;;AAChD,SAAOV,OAAO,CAAA,EAAGlB,QAAQ,CAAA,EAAGU;AAC9B;AAGA,IAAMa,mBAAmB;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AC5ClB,IAAMM,aAAa,CAACC,cACzB,IAAIC,KAAKD,aAAaC,KAAKC,IAAG,CAAA,EAAIC,mBAAmB,SAAS;EAC5DC,UAAU;EACVC,QAAQ;EACRC,MAAM;EACNC,QAAQ;EACRC,QAAQ;AACV,CAAA;;;ACPF,SAASC,gBAAgB;AACzB,SAASC,OAAAA,MAAKC,mBAAmBC,eAAe;AAEzC,IAAMC,iBAAiB,CAACC,YAC7B,IAAIC,KAAIA,KAAIC,KAAKC,OAAO;EAACC,kBAAkBC;EAAML,WAAWM,QAAQC,OAAM;EAAIC,SAASD,OAAM;CAAG;;;ACJ3F,IAAME,qBAAqB,CAACC,WAAAA;AACjC,QAAMC,WAAW,IAAIC,aAAaF,OAAOG,OAAO,CAACC,KAAKC,UAAUA,MAAMC,SAASF,KAAK,CAAA,CAAA;AACpF,MAAIG,SAAS;AACb,aAAWF,SAASL,QAAQ;AAC1BC,aAASO,IAAIH,OAAOE,MAAAA;AACpBA,cAAUF,MAAMC;EAClB;AAEA,SAAOL;AACT;;;ACTA,SAASQ,YAAAA,iBAAgB;AAEzB,SAASC,cAAcC,eAAe;AACtC,SAAuBC,gBAAgBC,gBAAgB;AACvD,SAASC,OAAAA,YAAW;AACpB,SAASC,SAAAA,cAAa;;;;;;;;AAiEf,IAAMC,cAAN,cAA0BC,SAAAA;EAa/BC,YAAY,EAAEC,QAAQC,UAAUC,WAAU,GAAuB;AAC/D,UAAK;AAbCC,wBAA6B,CAAA;AAC7BC,8BAAqB;AAEZC,wBAAe,IAAIC,QAAQ;MAAEC,WAAW;IAAK,CAAA;AAMtDC,sBAAa;AACbC,2BAAiCC;AAIvC,SAAKC,UAAUX;AACf,SAAKY,YAAYX;AACjB,SAAKY,cAAcX;EACrB;EAEA,MAAyBY,MAAMC,KAAc;AAC3C,SAAKH,UAAUI,WAAW,CAACC,UAAU,KAAKC,gBAAgBD,KAAAA,CAAAA;AAC1D,UAAM,KAAKL,UAAUO,MAAK;AAC1B,SAAKV,kBAAkB,IAAIW,aAAaL,KAAK,YAAY,KAAKM,YAAW,CAAA;AACzE,SAAKhB,aAAaiB,KAAI;EACxB;EAEA,MAAyBC,SAAS;AAChC,SAAKf,aAAa;AAClB,SAAKC,kBAAkBC;AACvB,UAAM,KAAKE,UAAUY,KAAI;EAC3B;EAEAC,uBAAuB;AACrB,QAAI,KAAKC,oBAAoBC,eAAeC,MAAM;AAChD;IACF;AACA,SAAKpB,aAAa;EACpB;EAEAqB,sBAAsB;AACpB,QAAI,KAAKH,oBAAoBC,eAAeC,QAAQ,CAAC,KAAKpB,YAAY;AACpE;IACF;AACA,SAAKA,aAAa;AAClB,SAAKC,iBAAiBqB,SAAAA;EACxB;EAEQZ,gBAAgBD,OAAmB;AACzCc,IAAAA,KAAI,sBAAsB;MAAEd;IAAM,GAAA;;;;;;AAClC,SAAKd,aAAa6B,KAAKf,KAAAA;AAGvB,QACE,CAAC,KAAKT,eACL,CAAC,KAAKC,mBAAmB,CAAC,KAAKA,gBAAgBwB,cAChD,KAAK9B,aAAa+B,UAAU,KAAKvB,QAAQwB,0BACzC;AACA,WAAKhC,eAAe,KAAKiC,eAAc;IACzC,WAAW,KAAKjC,aAAa+B,UAAU,KAAKvB,QAAQ0B,+BAA+B,KAAK7B,YAAY;AAClG,WAAKC,iBAAiBqB,SAAAA;IACxB;EACF;EAEA,MAAcT,cAAc;AAC1BU,IAAAA,KAAI,gBAAgB;MAAEO,QAAQ,KAAKnC,aAAa+B;IAAO,GAAA;;;;;;AACvD,UAAMI,SAAS,KAAKnC;AACpB,SAAKA,eAAe,KAAKiC,eAAc;AACvC,QAAIE,OAAOJ,WAAW,GAAG;AACvB;IACF;AAEA,UAAMK,QAAQ,MAAM,KAAKC,kBAAkBF,MAAAA;AAC3C,UAAMG,WAAW,MAAM,KAAKC,oBAAoBH,KAAAA;AAChD,QAAI,CAACI,MAAMC,QAAQH,QAAAA,KAAaA,SAASP,WAAW,GAAG;AACrD;IACF;AAEA,UAAMW,kBAAkB,KAAKC,eAAeL,UAAUH,MAAAA;AACtD,QAAIO,gBAAgBX,SAAS,GAAG;AAC9B,YAAM,KAAKrB,YAAYgC,eAAAA;IACzB;EACF;EAEQT,iBAAiB;AACvB,WAAO,KAAKzB,QAAQwB,2BAA2B,IAC3C,KAAKhC,aAAa4C,MAAM,CAAC,KAAKpC,QAAQwB,wBAAwB,IAC9D,CAAA;EACN;EAEA,MACcK,kBAAkBF,QAAsB;AACpD,UAAMU,OAAO,IAAIC,UAAAA;AACjB,UAAMC,YAAY,KAAKtC,UAAUsC;AAEjCF,SAAKG,YACHD,UAAUE,UACVF,UAAUG,YACVH,UAAUI,cACVC,mBAAmBjB,OAAOkB,IAAI,CAAC,EAAEC,KAAI,MAAOA,IAAAA,CAAAA,CAAAA;AAG9C,WAAOT,KAAKU,SAAQ;EACtB;EAEA,MACchB,oBAAoBH,OAAe;AAC/C,QAAIA,MAAML,WAAW,GAAG;AACtB,WAAK/B,eAAe,CAAA;AACpB,YAAM,IAAIwD,MAAM,mCAAA;IAClB;AAEA,UAAMC,WAAW,MAAMC,MAAM,GAAGC,iBAAAA,eAAgC;MAC9DC,QAAQ;MACRC,MAAMC,KAAKC,UAAU;QAAE3B;MAAM,CAAA;MAC7B4B,SAAS;QACP,gBAAgB;MAClB;IACF,CAAA;AAEA,QAAI,CAACP,SAASQ,IAAI;AAChB,WAAKjE,eAAe,CAAA;AACpB,YAAM,IAAIwD,MAAM,sBAAA;IAClB;AAEA,UAAM,EACJlB,SAAQ,IAGN,MAAMmB,SAASS,KAAI;AAEvBtC,IAAAA,KAAIuC,KAAK,0BAA0B;MACjC7B;MACA8B,QAAQ9B,SAASe,IAAI,CAACf,cAAaA,UAAS+B,IAAI,EAAEC,KAAK,GAAA;IACzD,GAAA;;;;;;AAEA,WAAOhC;EACT;EAEQK,eAAeL,UAA4BiC,gBAAmD;AAEpG,UAAMC,gBAAgBD,eAAeE,GAAG,CAAA,EAAIC;AAE5C,UAAMC,mBAAmBrC,SAGtBsC,OAAO,CAACC,YAAYL,gBAAgBK,QAAQC,MAAM,OAAS,KAAK7E,kBAAkB,EAElFoD,IAAI,CAACwB,YAAAA;AACJ,YAAME,QAAQF,QAAQE,MAAMH,OAAO,CAACI,SAASR,gBAAgBQ,KAAKhE,QAAQ,OAAS,KAAKf,kBAAkB;AAC1G,aAAO;QACL,GAAG4E;QACHE;QACAV,MAAMU,MAAM1B,IAAI,CAAC2B,SAASA,KAAKA,IAAI,EAAEV,KAAK,EAAA;QAC1CtD,OAAO+D,MAAMN,GAAG,CAAA,GAAIzD,SAAS6D,QAAQC;MACvC;IACF,CAAA,EACCF,OAAO,CAACC,YAAYA,QAAQE,MAAMhD,SAAS,CAAA;AAE9C,QAAI4C,iBAAiB5C,WAAW,GAAG;AACjC,aAAO,CAAA;IACT;AAGA,SAAK9B,qBAAqBuE,gBAAgBG,iBAAiBF,GAAG,EAAC,EAAIK,MAAM;AAGzE,WAAOH,iBAAiBtB,IAAI,CAACwB,aAAa;MACxCI,SAAS,IAAIC,KAAKV,gBAAgBK,QAAQ7D,QAAQ,GAAA;MAClDqD,MAAMQ,QAAQR;IAChB,EAAA;EACF;AACF;;EAlFGc,OAAMC,KAAK;IAAEC,uBAAuB;EAAK,CAAA;GA1F/B3F,YAAAA,WAAAA,qBAAAA,IAAAA;;EAyGVyF,OAAMC,KAAK;IAAEC,uBAAuB;EAAK,CAAA;GAzG/B3F,YAAAA,WAAAA,uBAAAA,IAAAA;",
+  "names": ["MediaRecorder", "register", "WaveFile", "synchronized", "invariant", "log", "trace", "initializingPromise", "initializeExtendableMediaRecorder", "connect", "MediaStreamRecorder", "constructor", "mediaStreamTrack", "interval", "_mediaRecorder", "undefined", "_header", "_mediaStreamTrack", "_interval", "wavConfig", "settings", "getSettings", "channels", "channelCount", "sampleRate", "bitDepthCode", "sampleSize", "String", "setOnChunk", "onChunk", "_onChunk", "start", "catch", "err", "info", "stream", "MediaStream", "mimeType", "state", "ondataavailable", "event", "_ondataavailable", "stop", "blob", "data", "uint8Array", "Uint8Array", "arrayBuffer", "isHeader", "wav", "slice", "warn", "timestamp", "Date", "now", "getSamples", "resource", "Message", "DEFAULT_EDGE_MODEL", "MixedStreamParser", "QueueImpl", "createStatic", "isInstanceOf", "invariant", "DXN", "log", "summarizeTranscript", "edge", "ai", "transcript", "context", "invariant", "queue", "QueueImpl", "DXN", "parse", "refresh", "content", "items", "filter", "block", "isInstanceOf", "TranscriptBlock", "map", "authorName", "segments", "segment", "text", "join", "parser", "MixedStreamParser", "log", "info", "blockCount", "length", "output", "exec", "model", "DEFAULT_EDGE_MODEL", "systemPrompt", "SUMMARIZE_PROMPT", "history", "createStatic", "Message", "role", "type", "getTimeStr", "timestamp", "Date", "now", "toLocaleTimeString", "timeZone", "hour12", "hour", "minute", "second", "ObjectId", "DXN", "QueueSubspaceTags", "SpaceId", "randomQueueDxn", "spaceId", "DXN", "kind", "QUEUE", "QueueSubspaceTags", "DATA", "SpaceId", "random", "ObjectId", "mergeFloat64Arrays", "arrays", "newArray", "Float64Array", "reduce", "acc", "array", "length", "offset", "set", "WaveFile", "DeferredTask", "Trigger", "LifecycleState", "Resource", "log", "trace", "Transcriber", "Resource", "constructor", "config", "recorder", "onSegments", "_audioChunks", "_lastUsedTimestamp", "_openTrigger", "Trigger", "autoReset", "_recording", "_transcribeTask", "undefined", "_config", "_recorder", "_onSegments", "_open", "ctx", "setOnChunk", "chunk", "_saveAudioChunk", "start", "DeferredTask", "_transcribe", "wake", "_close", "stop", "startChunksRecording", "_lifecycleState", "LifecycleState", "OPEN", "stopChunksRecording", "schedule", "log", "push", "scheduled", "length", "prefixBufferChunksAmount", "_dropOldChunks", "transcribeAfterChunksAmount", "chunks", "audio", "_mergeAudioChunks", "segments", "_fetchTranscription", "Array", "isArray", "alignedSegments", "_alignSegments", "slice", "file", "WaveFile", "wavConfig", "fromScratch", "channels", "sampleRate", "bitDepthCode", "mergeFloat64Arrays", "map", "data", "toBase64", "Error", "response", "fetch", "TRANSCRIPTION_URL", "method", "body", "JSON", "stringify", "headers", "ok", "json", "info", "string", "text", "join", "originalChunks", "zeroTimestamp", "at", "timestamp", "filteredSegments", "filter", "segment", "end", "words", "word", "started", "Date", "trace", "span", "showInBrowserTimeline"]
+}

package/dist/lib/browser/chunk-YEVVZZTC.mjs

@@ -0,0 +1,19 @@
+// packages/plugins/experimental/plugin-transcription/src/meta.ts
+var TRANSCRIPTION_PLUGIN = "dxos.org/plugin/transcription";
+var meta = {
+  id: TRANSCRIPTION_PLUGIN,
+  name: "Transcription",
+  description: "Voice transcription.",
+  icon: "ph--microphone--regular",
+  source: "https://github.com/dxos/dxos/tree/main/packages/plugins/experimental/plugin-transcription",
+  tags: [
+    "experimental"
+  ],
+  screenshots: []
+};
+
+export {
+  TRANSCRIPTION_PLUGIN,
+  meta
+};
+//# sourceMappingURL=chunk-YEVVZZTC.mjs.map

package/dist/lib/browser/chunk-YEVVZZTC.mjs.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../../src/meta.ts"],
+  "sourcesContent": ["//\n// Copyright 2023 DXOS.org\n//\n\nimport { type PluginMeta } from '@dxos/app-framework';\n\nexport const TRANSCRIPTION_PLUGIN = 'dxos.org/plugin/transcription';\n\nexport const meta: PluginMeta = {\n  id: TRANSCRIPTION_PLUGIN,\n  name: 'Transcription',\n  description: 'Voice transcription.',\n  icon: 'ph--microphone--regular',\n  source: 'https://github.com/dxos/dxos/tree/main/packages/plugins/experimental/plugin-transcription',\n  tags: ['experimental'],\n  screenshots: [],\n};\n"],
+  "mappings": ";AAMO,IAAMA,uBAAuB;AAE7B,IAAMC,OAAmB;EAC9BC,IAAIF;EACJG,MAAM;EACNC,aAAa;EACbC,MAAM;EACNC,QAAQ;EACRC,MAAM;IAAC;;EACPC,aAAa,CAAA;AACf;",
+  "names": ["TRANSCRIPTION_PLUGIN", "meta", "id", "name", "description", "icon", "source", "tags", "screenshots"]
+}

package/dist/lib/browser/{chunk-QIHUPVQD.mjs → chunk-YRW4MNRE.mjs}

@@ -1,5 +1,9 @@
+import {
+  TRANSCRIPTION_PLUGIN
+} from "./chunk-YEVVZZTC.mjs";
+
 // packages/plugins/experimental/plugin-transcription/src/types/schema.ts
-import { EchoObject, ObjectId, S, TypedObject } from "@dxos/echo-schema";
+import { EchoObject, S, TypedObject } from "@dxos/echo-schema";
 var TranscriptSchema = S.Struct({
   name: S.String,
   // TODO(burdon): Use string for dates?
@@ -8,6 +12,7 @@ var TranscriptSchema = S.Struct({
   /**
   * Queue containing TranscriptBlock objects.
   */
+  // TODO(wittjosiah): Should be a ref to a queue.
   queue: S.optional(S.String)
 });
 var TranscriptType = class extends TypedObject({
@@ -24,52 +29,46 @@ var TranscriptSegment = S.Struct({
   text: S.String
 });
 var TranscriptBlock = S.Struct({
-  id: ObjectId,
-  author: S.optional(S.String),
+  id: S.String,
+  authorName: S.optional(S.String),
+  authorHue: S.optional(S.String),
   segments: S.Array(TranscriptSegment)
 }).pipe(EchoObject("dxos.org/type/TranscriptBlock", "0.1.0"));
 
 // packages/plugins/experimental/plugin-transcription/src/types/types.ts
 import { S as S2, isInstanceOf } from "@dxos/echo-schema";
 import { isReactiveObject } from "@dxos/react-client/echo";
-
-// packages/plugins/experimental/plugin-transcription/src/meta.ts
-var TRANSCRIPTION_PLUGIN = "dxos.org/plugin/transcription";
-var meta = {
-  id: TRANSCRIPTION_PLUGIN,
-  name: "Transcription",
-  description: "Voice transcription.",
-  icon: "ph--microphone--regular",
-  source: "https://github.com/dxos/dxos/tree/main/packages/plugins/experimental/plugin-transcription",
-  tags: [
-    "experimental"
-  ],
-  screenshots: []
-};
-
-// packages/plugins/experimental/plugin-transcription/src/types/types.ts
 var TRANSCRIPTION_URL = "https://calls-service.dxos.workers.dev";
 var TranscriptionAction;
 (function(TranscriptionAction2) {
   const TRANSCRIPTION_ACTION = `${TRANSCRIPTION_PLUGIN}/action`;
   class Create extends S2.TaggedClass()(`${TRANSCRIPTION_ACTION}/create`, {
     input: S2.Struct({
-      name: S2.optional(S2.String)
+      name: S2.optional(S2.String),
+      // TODO(wittjosiah): SpaceId.
+      spaceId: S2.String
     }),
     output: S2.Struct({
-      object: TranscriptSchema
+      object: TranscriptType
     })
   }) {
   }
   TranscriptionAction2.Create = Create;
+  class Summarize extends S2.TaggedClass()(`${TRANSCRIPTION_ACTION}/summarize`, {
+    input: S2.Struct({
+      transcript: TranscriptType,
+      context: S2.optional(S2.String)
+    }),
+    output: S2.String
+  }) {
+  }
+  TranscriptionAction2.Summarize = Summarize;
 })(TranscriptionAction || (TranscriptionAction = {}));
 var isTranscript = (object) => {
   return isReactiveObject(object) && isInstanceOf(TranscriptType, object);
 };
 
 export {
-  TRANSCRIPTION_PLUGIN,
-  meta,
   TranscriptSchema,
   TranscriptType,
   TranscriptBlock,
@@ -77,4 +76,4 @@ export {
   TranscriptionAction,
   isTranscript
 };
-//# sourceMappingURL=chunk-QIHUPVQD.mjs.map
+//# sourceMappingURL=chunk-YRW4MNRE.mjs.map

package/dist/lib/browser/chunk-YRW4MNRE.mjs.map

@@ -0,0 +1,7 @@
+{
+  "version": 3,
+  "sources": ["../../../src/types/schema.ts", "../../../src/types/types.ts"],
+  "sourcesContent": ["//\n// Copyright 2025 DXOS.org\n//\n\nimport { EchoObject, S, TypedObject } from '@dxos/echo-schema';\n\n/**\n * Root transcript object created when the user starts a transcription.\n */\nexport const TranscriptSchema = S.Struct({\n  name: S.String,\n\n  // TODO(burdon): Use string for dates?\n  started: S.optional(S.Date),\n  ended: S.optional(S.Date),\n\n  /**\n   * Queue containing TranscriptBlock objects.\n   */\n  // TODO(wittjosiah): Should be a ref to a queue.\n  queue: S.optional(S.String),\n});\n\nexport class TranscriptType extends TypedObject({\n  typename: 'dxos.org/type/Transcript',\n  version: '0.1.0',\n})(TranscriptSchema.fields) {}\n\n// TODO(burdon): Do these need to be kept in sync with EDGE?\n\n/**\n * First message in queue.\n * Contains metadata for the recording and transcript.\n */\nconst TranscriptHeader = S.Struct({\n  started: S.optional(S.Date),\n});\n\nexport type TranscriptHeader = S.Schema.Type<typeof TranscriptHeader>;\n\n/**\n * Transcription fragment.\n */\nconst TranscriptSegment = S.Struct({\n  // TODO(burdon): TS from service is not Unix TS (x1000).\n  started: S.Date,\n  text: S.String,\n});\n\nexport type TranscriptSegment = S.Schema.Type<typeof TranscriptSegment>;\n\n/**\n * Transcription block (from single speaker).\n */\nexport const TranscriptBlock = S.Struct({\n  id: S.String,\n  authorName: S.optional(S.String), // TODO(burdon): IdentityDid.\n  authorHue: S.optional(S.String),\n  segments: S.Array(TranscriptSegment),\n}).pipe(EchoObject('dxos.org/type/TranscriptBlock', '0.1.0'));\n\nexport type TranscriptBlock = S.Schema.Type<typeof TranscriptBlock>;\n", "//\n// Copyright 2025 DXOS.org\n//\n\nimport { S, isInstanceOf } from '@dxos/echo-schema';\nimport { isReactiveObject } from '@dxos/react-client/echo';\n\nimport { TranscriptType } from './schema';\nimport { TRANSCRIPTION_PLUGIN } from '../meta';\n\n// TODO(burdon): Move to separate proto.\n\n/**\n * Endpoint to the calls service.\n */\n// TODO(burdon): Reconfigure/move.\nexport const TRANSCRIPTION_URL = 'https://calls-service.dxos.workers.dev';\n\nexport namespace TranscriptionAction {\n  const TRANSCRIPTION_ACTION = `${TRANSCRIPTION_PLUGIN}/action`;\n\n  export class Create extends S.TaggedClass<Create>()(`${TRANSCRIPTION_ACTION}/create`, {\n    input: S.Struct({\n      name: S.optional(S.String),\n      // TODO(wittjosiah): SpaceId.\n      spaceId: S.String,\n    }),\n    output: S.Struct({\n      object: TranscriptType,\n    }),\n  }) {}\n\n  export class Summarize extends S.TaggedClass<Summarize>()(`${TRANSCRIPTION_ACTION}/summarize`, {\n    input: S.Struct({\n      transcript: TranscriptType,\n      context: S.optional(S.String),\n    }),\n    output: S.String,\n  }) {}\n}\n\nexport const isTranscript = (object: unknown): object is typeof TranscriptType => {\n  return isReactiveObject(object) && isInstanceOf(TranscriptType, object);\n};\n"],
+  "mappings": ";;;;;AAIA,SAASA,YAAYC,GAAGC,mBAAmB;AAKpC,IAAMC,mBAAmBC,EAAEC,OAAO;EACvCC,MAAMF,EAAEG;;EAGRC,SAASJ,EAAEK,SAASL,EAAEM,IAAI;EAC1BC,OAAOP,EAAEK,SAASL,EAAEM,IAAI;;;;;EAMxBE,OAAOR,EAAEK,SAASL,EAAEG,MAAM;AAC5B,CAAA;AAEO,IAAMM,iBAAN,cAA6BC,YAAY;EAC9CC,UAAU;EACVC,SAAS;AACX,CAAA,EAAGb,iBAAiBc,MAAM,EAAA;AAAG;AAQ7B,IAAMC,mBAAmBd,EAAEC,OAAO;EAChCG,SAASJ,EAAEK,SAASL,EAAEM,IAAI;AAC5B,CAAA;AAOA,IAAMS,oBAAoBf,EAAEC,OAAO;;EAEjCG,SAASJ,EAAEM;EACXU,MAAMhB,EAAEG;AACV,CAAA;AAOO,IAAMc,kBAAkBjB,EAAEC,OAAO;EACtCiB,IAAIlB,EAAEG;EACNgB,YAAYnB,EAAEK,SAASL,EAAEG,MAAM;EAC/BiB,WAAWpB,EAAEK,SAASL,EAAEG,MAAM;EAC9BkB,UAAUrB,EAAEsB,MAAMP,iBAAAA;AACpB,CAAA,EAAGQ,KAAKC,WAAW,iCAAiC,OAAA,CAAA;;;ACvDpD,SAASC,KAAAA,IAAGC,oBAAoB;AAChC,SAASC,wBAAwB;AAW1B,IAAMC,oBAAoB;;UAEhBC,sBAAAA;AACf,QAAMC,uBAAuB,GAAGC,oBAAAA;EAEzB,MAAMC,eAAeC,GAAEC,YAAW,EAAW,GAAGJ,oBAAAA,WAA+B;IACpFK,OAAOF,GAAEG,OAAO;MACdC,MAAMJ,GAAEK,SAASL,GAAEM,MAAM;;MAEzBC,SAASP,GAAEM;IACb,CAAA;IACAE,QAAQR,GAAEG,OAAO;MACfM,QAAQC;IACV,CAAA;EACF,CAAA,EAAA;EAAI;uBATSX,SAAAA;EAWN,MAAMY,kBAAkBX,GAAEC,YAAW,EAAc,GAAGJ,oBAAAA,cAAkC;IAC7FK,OAAOF,GAAEG,OAAO;MACdS,YAAYF;MACZG,SAASb,GAAEK,SAASL,GAAEM,MAAM;IAC9B,CAAA;IACAE,QAAQR,GAAEM;EACZ,CAAA,EAAA;EAAI;uBANSK,YAAAA;AAOf,GArBiBf,wBAAAA,sBAAAA,CAAAA,EAAAA;AAuBV,IAAMkB,eAAe,CAACL,WAAAA;AAC3B,SAAOM,iBAAiBN,MAAAA,KAAWO,aAAaN,gBAAgBD,MAAAA;AAClE;",
+  "names": ["EchoObject", "S", "TypedObject", "TranscriptSchema", "S", "Struct", "name", "String", "started", "optional", "Date", "ended", "queue", "TranscriptType", "TypedObject", "typename", "version", "fields", "TranscriptHeader", "TranscriptSegment", "text", "TranscriptBlock", "id", "authorName", "authorHue", "segments", "Array", "pipe", "EchoObject", "S", "isInstanceOf", "isReactiveObject", "TRANSCRIPTION_URL", "TranscriptionAction", "TRANSCRIPTION_ACTION", "TRANSCRIPTION_PLUGIN", "Create", "S", "TaggedClass", "input", "Struct", "name", "optional", "String", "spaceId", "output", "object", "TranscriptType", "Summarize", "transcript", "context", "isTranscript", "isReactiveObject", "isInstanceOf"]
+}