Add cancel + progress UI, EngineToggle, and Docker deployment

src/worker/inference.worker.ts

@@ -19,6 +19,7 @@ import { resolveDefaultDtype } from "../lib/models";
 import {
   deviceAttempts,
   isUiToWorkerMessage,
+  TranscribeProgressTracker,
   type DeviceCapabilities,
   type DevicePreference,
   type InferenceDevice,
@@ -209,6 +210,12 @@ function load(message: LoadMessage): Promise<AutomaticSpeechRecognitionPipeline>
 // Transcription
 // ---------------------------------------------------------------------------
 
+/** Audio sample rate the whole app standardizes on (D-012). */
+const SAMPLE_RATE = 16_000;
+/** Chunked-pipeline window / overlap, in seconds (Whisper's usual 30/5). */
+const CHUNK_LENGTH_S = 30;
+const STRIDE_LENGTH_S = 5;
+
 async function transcribe(message: TranscribeMessage): Promise<void> {
   if (active === null) {
     post({
@@ -222,11 +229,39 @@ async function transcribe(message: TranscribeMessage): Promise<void> {
   const asr = await active.promise;
 
   const options: Record<string, unknown> = {
-    chunk_length_s: 30,
-    stride_length_s: 5,
+    chunk_length_s: CHUNK_LENGTH_S,
+    stride_length_s: STRIDE_LENGTH_S,
     return_timestamps: true,
   };
-  if (message.requestId === undefined) {
+  if (message.stream ?? message.requestId === undefined) {
+    // File-flow pass (id-less, or routed through BrowserEngine with an
+    // explicit stream flag, D-010): stream partial tokens and report audio
+    // progress. The streamer's chunk timestamps are window-relative; the
+    // tracker maps them to monotone absolute seconds (see protocol.ts).
+    // Live passes skip both — they resolve wholesale via "result" (D-016).
+    const totalSeconds = message.audio.length / SAMPLE_RATE;
+    const tracker = new TranscribeProgressTracker(
+      totalSeconds,
+      CHUNK_LENGTH_S,
+      CHUNK_LENGTH_S - 2 * STRIDE_LENGTH_S,
+    );
+    const postProgress = (processedSeconds: number | null) => {
+      if (processedSeconds === null) return;
+      post({
+        type: "transcribe-progress",
+        processedSeconds,
+        totalSeconds,
+        requestId: message.requestId,
+      });
+    };
+    // Immediate 0/total so the UI shows the audio length before the first
+    // (possibly slow) token arrives.
+    post({
+      type: "transcribe-progress",
+      processedSeconds: 0,
+      totalSeconds,
+      requestId: message.requestId,
+    });
     options.streamer = new WhisperTextStreamer(
       // The ASR pipeline for Whisper models always carries a WhisperTokenizer.
       asr.tokenizer as ConstructorParameters<typeof WhisperTextStreamer>[0],
@@ -234,6 +269,9 @@ async function transcribe(message: TranscribeMessage): Promise<void> {
         skip_prompt: true,
         callback_function: (text: string) =>
           post({ type: "partial", text, requestId: message.requestId }),
+        on_chunk_start: (time: number) => postProgress(tracker.timestamp(time)),
+        on_chunk_end: (time: number) => postProgress(tracker.timestamp(time)),
+        on_finalize: () => postProgress(tracker.finalizeWindow()),
       },
     );
   }