@agent-native/core 0.84.12 → 0.84.13

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (83) hide show
  1. package/corpus/README.md +1 -1
  2. package/corpus/core/CHANGELOG.md +11 -0
  3. package/corpus/core/package.json +1 -1
  4. package/corpus/core/src/agent/durable-background.ts +30 -0
  5. package/corpus/core/src/agent/production-agent.ts +34 -6
  6. package/corpus/core/src/agent/run-manager.ts +45 -2
  7. package/corpus/core/src/chat-threads/store.ts +2 -0
  8. package/corpus/core/src/cli/skills.ts +10 -6
  9. package/corpus/core/src/client/AssistantChat.tsx +40 -9
  10. package/corpus/core/src/client/agent-chat-adapter.ts +23 -2
  11. package/corpus/core/src/client/chat/run-recovery.tsx +52 -44
  12. package/corpus/core/src/client/chat/tool-call-display.tsx +7 -2
  13. package/corpus/core/src/server/agent-chat-plugin.ts +1 -0
  14. package/corpus/core/src/shared/streaming-text-smoothing.ts +10 -5
  15. package/corpus/templates/clips/app/components/library/recording-card.tsx +14 -4
  16. package/corpus/templates/clips/app/components/recorder/recorder-engine.ts +10 -12
  17. package/corpus/templates/clips/app/hooks/use-library.ts +7 -4
  18. package/corpus/templates/clips/app/lib/recording-status.ts +32 -0
  19. package/corpus/templates/clips/changelog/2026-07-01-dictation-keeps-listening-through-natural-pauses-and-gives-c.md +6 -0
  20. package/corpus/templates/clips/changelog/2026-07-01-redoing-a-paused-desktop-recording-no-longer-leaves-the-recording-controls-disabled.md +6 -0
  21. package/corpus/templates/clips/changelog/2026-07-01-stalled-clip-uploads-now-update-in-the-library-and-surface-as-failed.md +6 -0
  22. package/corpus/templates/clips/desktop/src/lib/recorder.ts +36 -18
  23. package/corpus/templates/clips/desktop/src/lib/voice-dictation.ts +99 -27
  24. package/corpus/templates/clips/desktop/src/overlays/flow-bar.tsx +4 -43
  25. package/corpus/templates/clips/desktop/src/overlays/toolbar.tsx +3 -0
  26. package/corpus/templates/clips/desktop/src/styles.css +5 -48
  27. package/corpus/templates/clips/desktop/src-tauri/src/clips/mod.rs +4 -5
  28. package/corpus/templates/design/.agents/skills/design-generation/SKILL.md +15 -6
  29. package/corpus/templates/design/AGENTS.md +8 -4
  30. package/corpus/templates/design/actions/create-design.ts +1 -0
  31. package/corpus/templates/design/actions/duplicate-design.ts +3 -1
  32. package/corpus/templates/design/actions/edit-design.ts +43 -13
  33. package/corpus/templates/design/actions/generate-design.ts +5 -1
  34. package/corpus/templates/design/actions/present-design-variants.ts +17 -12
  35. package/corpus/templates/design/app/hooks/use-question-flow.ts +2 -2
  36. package/corpus/templates/design/app/pages/DesignEditor.tsx +2 -2
  37. package/corpus/templates/design/changelog/2026-07-01-org-members-can-see-design-projects-created-in-their-workspace.md +6 -0
  38. package/corpus/templates/design/server/plugins/db.ts +9 -0
  39. package/dist/agent/durable-background.d.ts +3 -0
  40. package/dist/agent/durable-background.d.ts.map +1 -1
  41. package/dist/agent/durable-background.js +19 -0
  42. package/dist/agent/durable-background.js.map +1 -1
  43. package/dist/agent/production-agent.d.ts +1 -0
  44. package/dist/agent/production-agent.d.ts.map +1 -1
  45. package/dist/agent/production-agent.js +29 -6
  46. package/dist/agent/production-agent.js.map +1 -1
  47. package/dist/agent/run-manager.d.ts.map +1 -1
  48. package/dist/agent/run-manager.js +40 -2
  49. package/dist/agent/run-manager.js.map +1 -1
  50. package/dist/chat-threads/store.d.ts +1 -0
  51. package/dist/chat-threads/store.d.ts.map +1 -1
  52. package/dist/chat-threads/store.js +2 -0
  53. package/dist/chat-threads/store.js.map +1 -1
  54. package/dist/cli/skills.d.ts.map +1 -1
  55. package/dist/cli/skills.js +10 -6
  56. package/dist/cli/skills.js.map +1 -1
  57. package/dist/client/AssistantChat.d.ts +10 -0
  58. package/dist/client/AssistantChat.d.ts.map +1 -1
  59. package/dist/client/AssistantChat.js +26 -8
  60. package/dist/client/AssistantChat.js.map +1 -1
  61. package/dist/client/agent-chat-adapter.d.ts.map +1 -1
  62. package/dist/client/agent-chat-adapter.js +21 -2
  63. package/dist/client/agent-chat-adapter.js.map +1 -1
  64. package/dist/client/chat/run-recovery.d.ts.map +1 -1
  65. package/dist/client/chat/run-recovery.js +15 -10
  66. package/dist/client/chat/run-recovery.js.map +1 -1
  67. package/dist/client/chat/tool-call-display.d.ts +1 -0
  68. package/dist/client/chat/tool-call-display.d.ts.map +1 -1
  69. package/dist/client/chat/tool-call-display.js +3 -2
  70. package/dist/client/chat/tool-call-display.js.map +1 -1
  71. package/dist/collab/routes.d.ts +2 -2
  72. package/dist/notifications/routes.d.ts +2 -2
  73. package/dist/observability/routes.d.ts +3 -3
  74. package/dist/progress/routes.d.ts +1 -1
  75. package/dist/resources/handlers.d.ts +1 -1
  76. package/dist/server/agent-chat-plugin.d.ts.map +1 -1
  77. package/dist/server/agent-chat-plugin.js +1 -1
  78. package/dist/server/agent-chat-plugin.js.map +1 -1
  79. package/dist/server/agent-engine-api-key-route.d.ts +1 -1
  80. package/dist/shared/streaming-text-smoothing.d.ts.map +1 -1
  81. package/dist/shared/streaming-text-smoothing.js +8 -5
  82. package/dist/shared/streaming-text-smoothing.js.map +1 -1
  83. package/package.json +1 -1
@@ -1,6 +1,7 @@
1
1
  export const SMOOTH_STREAMING_COMMIT_INTERVAL_MS = 16;
2
2
  export const SMOOTH_STREAMING_LONG_TEXT_THRESHOLD_GRAPHEMES = 640;
3
3
  export const SMOOTH_STREAMING_LONG_TEXT_TAIL_GRAPHEMES = 180;
4
+ const SMOOTH_STREAMING_SPEED_MULTIPLIER = 2;
4
5
 
5
6
  type SegmenterInstance = {
6
7
  segment(input: string): Iterable<{ segment: string }>;
@@ -137,7 +138,7 @@ export function smoothStreamingRevealCount({
137
138
  return 0;
138
139
  }
139
140
 
140
- const charactersPerSecond = inputDone
141
+ const baseCharactersPerSecond = inputDone
141
142
  ? backlog > 800
142
143
  ? 900
143
144
  : 420
@@ -148,8 +149,12 @@ export function smoothStreamingRevealCount({
148
149
  : backlog > 180
149
150
  ? 190
150
151
  : 95;
152
+ const charactersPerSecond =
153
+ baseCharactersPerSecond * SMOOTH_STREAMING_SPEED_MULTIPLIER;
151
154
 
152
- const maxBurst = inputDone ? 160 : backlog > 1400 ? 120 : 72;
155
+ const maxBurst =
156
+ (inputDone ? 160 : backlog > 1400 ? 120 : 72) *
157
+ SMOOTH_STREAMING_SPEED_MULTIPLIER;
153
158
  const count = Math.floor((elapsedMs / 1000) * charactersPerSecond);
154
159
 
155
160
  return Math.min(backlog, Math.max(1, count), maxBurst);
@@ -164,15 +169,15 @@ export function smoothStreamingPunctuationDelayMs(
164
169
  }
165
170
 
166
171
  if (grapheme === "\n") {
167
- return 80;
172
+ return 80 / SMOOTH_STREAMING_SPEED_MULTIPLIER;
168
173
  }
169
174
 
170
175
  if (/[.!?)]/.test(grapheme)) {
171
- return 70;
176
+ return 70 / SMOOTH_STREAMING_SPEED_MULTIPLIER;
172
177
  }
173
178
 
174
179
  if (/[,;:]/.test(grapheme)) {
175
- return 35;
180
+ return 35 / SMOOTH_STREAMING_SPEED_MULTIPLIER;
176
181
  }
177
182
 
178
183
  return 0;
@@ -28,6 +28,7 @@ import {
28
28
  } from "@/components/ui/dropdown-menu";
29
29
  import { isDefaultTitle } from "@/hooks/use-auto-title";
30
30
  import type { RecordingSummary } from "@/hooks/use-library";
31
+ import { isStaleRecordingUpload } from "@/lib/recording-status";
31
32
  import { isStorageSetupFailureReason } from "@/lib/storage-failures";
32
33
  import { cn } from "@/lib/utils";
33
34
 
@@ -100,6 +101,12 @@ export function RecordingCard({
100
101
  const waitingForStorage = isStorageSetupFailureReason(
101
102
  recording.failureReason,
102
103
  );
104
+ const staleUpload = isStaleRecordingUpload(recording);
105
+ const displayFailed = recording.status === "failed" || staleUpload;
106
+ const failureReason = staleUpload
107
+ ? (recording.failureReason ??
108
+ t("recordingPage.processingStuck", { status: recording.status }))
109
+ : (recording.failureReason ?? t("clipsFinalRaw.removeFailedClip"));
103
110
  const nativeUploadPaused =
104
111
  recording.status === "failed" &&
105
112
  /native recording|native fullscreen|screencapture|avconvert/i.test(
@@ -221,11 +228,15 @@ export function RecordingCard({
221
228
  {/* Status pill for non-ready recordings */}
222
229
  {recording.status !== "ready" && (
223
230
  <div className="absolute top-2 end-2 rounded-full bg-black/80 px-2 py-0.5 text-[10px] font-medium text-white uppercase tracking-wide">
224
- {waitingForStorage ? "storage" : recording.status}
231
+ {waitingForStorage
232
+ ? "storage"
233
+ : staleUpload
234
+ ? "failed"
235
+ : recording.status}
225
236
  </div>
226
237
  )}
227
238
 
228
- {(recording.status === "failed" || waitingForStorage) && (
239
+ {(displayFailed || waitingForStorage) && (
229
240
  <div
230
241
  className={cn(
231
242
  "absolute inset-x-2 bottom-2 rounded-md border bg-background/95 p-2 text-start shadow-sm backdrop-blur",
@@ -252,8 +263,7 @@ export function RecordingCard({
252
263
  ? t("clipsFinalRaw.connectStorageToFinish")
253
264
  : nativeUploadPaused
254
265
  ? t("clipsFinalRaw.retryFromClipsMenu")
255
- : (recording.failureReason ??
256
- t("clipsFinalRaw.removeFailedClip"))}
266
+ : failureReason}
257
267
  </div>
258
268
  </div>
259
269
  {!waitingForStorage && (
@@ -1145,6 +1145,9 @@ export class RecorderEngine {
1145
1145
  // mid-state — the UI spinner is wired to engine state and would
1146
1146
  // hang forever otherwise.
1147
1147
  const e = err instanceof Error ? err : new Error(String(err));
1148
+ if (e.name !== "AbortError") {
1149
+ this.rememberUploadFailure(e);
1150
+ }
1148
1151
  this.transition("error", { message: e.message });
1149
1152
  throw e;
1150
1153
  } finally {
@@ -1183,6 +1186,9 @@ export class RecorderEngine {
1183
1186
  return this.toFinalizeResult(result, meta);
1184
1187
  } catch (err) {
1185
1188
  const e = err instanceof Error ? err : new Error(String(err));
1189
+ if (e.name !== "AbortError") {
1190
+ this.rememberUploadFailure(e);
1191
+ }
1186
1192
  this.transition("error", { message: e.message });
1187
1193
  throw e;
1188
1194
  } finally {
@@ -1595,26 +1601,18 @@ export class RecorderEngine {
1595
1601
  const failure = err instanceof Error ? err : new Error(String(err));
1596
1602
  // User-initiated cancel — cancel() already runs the abortUrl path.
1597
1603
  if (failure.name === "AbortError") return;
1598
- await this.markUploadFailed(failure);
1604
+ this.rememberUploadFailure(failure);
1599
1605
  this.emitError(failure);
1600
1606
  }
1601
1607
  });
1602
1608
  }
1603
1609
 
1604
- private async markUploadFailed(err: Error): Promise<void> {
1610
+ private rememberUploadFailure(err: Error): void {
1605
1611
  if (!this.uploadFailure) {
1606
1612
  this.uploadFailure = err;
1607
1613
  }
1608
- if (!this.opts.abortUrl) return;
1609
- try {
1610
- await fetch(this.opts.abortUrl, {
1611
- method: "POST",
1612
- headers: { "Content-Type": "application/json" },
1613
- body: JSON.stringify({ reason: err.message }),
1614
- });
1615
- } catch {
1616
- // ignore — the stop path will surface the original upload error.
1617
- }
1614
+ // Do not call abortUrl for retryable upload failures. retryUpload() reuses
1615
+ // this recording id; cancel() owns the terminal server-side abort path.
1618
1616
  }
1619
1617
 
1620
1618
  private async uploadBlobInSlices(
@@ -1,5 +1,7 @@
1
1
  import { useActionQuery, useActionMutation } from "@agent-native/core/client";
2
2
 
3
+ import { isLiveRecordingUpload } from "@/lib/recording-status";
4
+
3
5
  export interface RecordingSummary {
4
6
  id: string;
5
7
  title: string;
@@ -72,16 +74,17 @@ export function useRecordings(args: ListRecordingsArgs = {}) {
72
74
  recordings: Array.isArray(data?.recordings) ? data.recordings : [],
73
75
  };
74
76
  },
75
- // If any recording is still on a replaceable seed title, keep a 3s
76
- // refetch cadence so the skeleton in `recording-card` upgrades to
77
- // the real title promptly even if the refresh-signal poll is missed.
77
+ // Keep a short poll while uploads/processors are active so the library
78
+ // card does not get stuck if the global refresh signal is missed.
79
+ // Also poll for replaceable seed titles so the card upgrades promptly.
78
80
  refetchInterval: (q) => {
79
81
  const recs = (q.state.data as any)?.recordings as
80
82
  | RecordingSummary[]
81
83
  | undefined;
82
84
  if (!recs || recs.length === 0) return false;
85
+ const pendingUpload = recs.some((rec) => isLiveRecordingUpload(rec));
83
86
  const pendingTitle = recs.some(isAwaitingAutoTitle);
84
- return pendingTitle ? 3000 : false;
87
+ return pendingUpload || pendingTitle ? 3000 : false;
85
88
  },
86
89
  },
87
90
  );
@@ -0,0 +1,32 @@
1
+ export const STALE_RECORDING_UPLOAD_MS = 30 * 60 * 1000;
2
+
3
+ type RecordingStatusLike = {
4
+ status?: string | null;
5
+ updatedAt?: string | null;
6
+ };
7
+
8
+ export function isActiveRecordingUploadStatus(
9
+ status: string | null | undefined,
10
+ ): boolean {
11
+ return status === "uploading" || status === "processing";
12
+ }
13
+
14
+ export function isStaleRecordingUpload(
15
+ recording: RecordingStatusLike,
16
+ nowMs = Date.now(),
17
+ ): boolean {
18
+ if (!isActiveRecordingUploadStatus(recording.status)) return false;
19
+ const updatedAtMs = Date.parse(recording.updatedAt ?? "");
20
+ if (!Number.isFinite(updatedAtMs)) return false;
21
+ return nowMs - updatedAtMs >= STALE_RECORDING_UPLOAD_MS;
22
+ }
23
+
24
+ export function isLiveRecordingUpload(
25
+ recording: RecordingStatusLike,
26
+ nowMs = Date.now(),
27
+ ): boolean {
28
+ if (!isActiveRecordingUploadStatus(recording.status)) return false;
29
+ const updatedAtMs = Date.parse(recording.updatedAt ?? "");
30
+ if (!Number.isFinite(updatedAtMs)) return true;
31
+ return nowMs - updatedAtMs < STALE_RECORDING_UPLOAD_MS;
32
+ }
@@ -0,0 +1,6 @@
1
+ ---
2
+ type: fixed
3
+ date: 2026-07-01
4
+ ---
5
+
6
+ Dictation keeps listening through natural pauses and gives clearer start feedback.
@@ -0,0 +1,6 @@
1
+ ---
2
+ type: fixed
3
+ date: 2026-07-01
4
+ ---
5
+
6
+ Redoing a paused desktop recording no longer leaves the recording controls disabled.
@@ -0,0 +1,6 @@
1
+ ---
2
+ type: fixed
3
+ date: 2026-07-01
4
+ ---
5
+
6
+ Stalled clip uploads now update in the library and surface as failed instead of staying stuck as uploading.
@@ -1278,6 +1278,18 @@ async function trashRecording(
1278
1278
  }
1279
1279
  }
1280
1280
 
1281
+ async function cleanupCancelledRemoteRecording(
1282
+ serverUrl: string,
1283
+ recordingId: string,
1284
+ ): Promise<void> {
1285
+ await abortRecordingUpload(
1286
+ serverUrl,
1287
+ recordingId,
1288
+ "Recording cancelled by user",
1289
+ );
1290
+ await trashRecording(serverUrl, recordingId);
1291
+ }
1292
+
1281
1293
  class CountdownCancelledError extends Error {
1282
1294
  constructor() {
1283
1295
  super("Recording cancelled during countdown");
@@ -2029,7 +2041,12 @@ async function startNativeFullscreenRecording(
2029
2041
  }
2030
2042
  stateUnlistens.forEach((u) => u());
2031
2043
  stateUnlistens = [];
2032
- await transcriptionCapture?.cancel().catch(() => {});
2044
+ void transcriptionCapture?.cancel().catch((err) => {
2045
+ console.warn(
2046
+ "[clips-recorder] native transcription cancel failed:",
2047
+ err,
2048
+ );
2049
+ });
2033
2050
  await localCameraExport?.cancel().catch(() => {});
2034
2051
  await invoke("native_fullscreen_recording_cancel").catch((err) =>
2035
2052
  console.warn(
@@ -2049,12 +2066,14 @@ async function startNativeFullscreenRecording(
2049
2066
  streamCleanups.forEach((cleanup) => cleanup());
2050
2067
  await invoke("hide_overlays").catch(() => {});
2051
2068
  if (!localOnly && id) {
2052
- await abortRecordingUpload(
2053
- params.serverUrl,
2054
- id,
2055
- "Recording cancelled by user",
2069
+ void cleanupCancelledRemoteRecording(params.serverUrl, id).catch(
2070
+ (err) => {
2071
+ console.warn(
2072
+ "[clips-recorder] cancelled recording cleanup failed:",
2073
+ err,
2074
+ );
2075
+ },
2056
2076
  );
2057
- await trashRecording(params.serverUrl, id);
2058
2077
  }
2059
2078
  })();
2060
2079
  return cancelPromise;
@@ -3244,7 +3263,9 @@ async function startRecordingInner(
3244
3263
  if (tickHandle) clearInterval(tickHandle);
3245
3264
  stateUnlistens.forEach((u) => u());
3246
3265
  stateUnlistens = [];
3247
- transcriptionCapture?.cancel().catch(() => {});
3266
+ void transcriptionCapture?.cancel().catch((err) => {
3267
+ console.warn("[clips-recorder] transcription cancel failed:", err);
3268
+ });
3248
3269
  // Remove MediaRecorder's data handler so any final `ondataavailable`
3249
3270
  // from the stop() below doesn't push a new Blob into `inflight`
3250
3271
  // after we've decided to discard everything.
@@ -3282,18 +3303,15 @@ async function startRecordingInner(
3282
3303
  inflight.clear();
3283
3304
  await invoke("hide_recording_chrome").catch(() => {});
3284
3305
  // Tell the server to abort the partial recording (drops chunks from
3285
- // application_state, flips the recording row to 'failed'). Fire and
3286
- // forget with a short-circuit on failure we don't want to keep the
3287
- // user waiting on a network call to a dev server that may be down.
3288
- try {
3289
- await abortRecordingUpload(
3290
- params.serverUrl,
3291
- id,
3292
- "Recording cancelled by user",
3306
+ // application_state, flips the recording row to 'failed'), then trash
3307
+ // it. This is best-effort background cleanup: redo/cancel must release
3308
+ // the desktop chrome immediately even if the server is slow or offline.
3309
+ if (id) {
3310
+ void cleanupCancelledRemoteRecording(params.serverUrl, id).catch(
3311
+ (err) => {
3312
+ console.warn("[clips-recorder] abort failed (non-fatal):", err);
3313
+ },
3293
3314
  );
3294
- await trashRecording(params.serverUrl, id);
3295
- } catch (err) {
3296
- console.warn("[clips-recorder] abort failed (non-fatal):", err);
3297
3315
  }
3298
3316
  await deleteBrowserRecordingBackup(id).catch((err) => {
3299
3317
  console.warn("[clips-recorder] local backup cleanup failed:", err);
@@ -102,6 +102,11 @@ interface VoiceSession {
102
102
  // Accumulated final transcript from interim webkit results, in case
103
103
  // the recognition session ends before we ask it to stop.
104
104
  browserTranscript: string;
105
+ // Native / Whisper engines can emit one final segment per pause. Keep
106
+ // committed segments separate from the current partial so long dictations
107
+ // survive natural pauses.
108
+ finalTranscriptParts: string[];
109
+ interimTranscript: string;
105
110
  // browser-only: monotonic timestamp of the most recent
106
111
  // recognition.onresult event. stop() reads this to decide whether
107
112
  // the user was actively speaking up to Fn release (worth a tail-
@@ -229,6 +234,73 @@ function setFlowState(state: FlowState): void {
229
234
  emit("voice:state-change", { state }).catch(() => {});
230
235
  }
231
236
 
237
+ let lastStartPingAt = 0;
238
+
239
+ function playStartPing(): void {
240
+ const now = Date.now();
241
+ if (now - lastStartPingAt < 250) return;
242
+ lastStartPingAt = now;
243
+ try {
244
+ const AudioCtx = window.AudioContext || (window as any).webkitAudioContext;
245
+ if (!AudioCtx) return;
246
+ const ctx = new AudioCtx();
247
+ const osc = ctx.createOscillator();
248
+ const gain = ctx.createGain();
249
+ const start = ctx.currentTime;
250
+ osc.type = "sine";
251
+ osc.frequency.setValueAtTime(880, start);
252
+ osc.frequency.exponentialRampToValueAtTime(1174.66, start + 0.08);
253
+ gain.gain.setValueAtTime(0.0001, start);
254
+ gain.gain.exponentialRampToValueAtTime(0.08, start + 0.015);
255
+ gain.gain.exponentialRampToValueAtTime(0.0001, start + 0.13);
256
+ osc.connect(gain);
257
+ gain.connect(ctx.destination);
258
+ osc.start(start);
259
+ osc.stop(start + 0.14);
260
+ window.setTimeout(() => {
261
+ ctx.close().catch(() => {});
262
+ }, 250);
263
+ } catch {
264
+ // Sound feedback is best-effort; the moving bars remain the visual cue.
265
+ }
266
+ }
267
+
268
+ function enterRecordingState(): void {
269
+ setFlowState("recording");
270
+ playStartPing();
271
+ }
272
+
273
+ function joinedTranscript(parts: string[], interim: string): string {
274
+ return [...parts, interim].filter(Boolean).join(" ").trim();
275
+ }
276
+
277
+ function appendFinalTranscript(session: VoiceSession, text: string): void {
278
+ const clean = text.trim();
279
+ if (!clean) return;
280
+ const committed = joinedTranscript(session.finalTranscriptParts, "");
281
+ if (!committed) {
282
+ session.finalTranscriptParts = [clean];
283
+ } else if (clean === committed || clean.startsWith(`${committed} `)) {
284
+ // Some recognizers send the whole dictation as their final result.
285
+ session.finalTranscriptParts = [clean];
286
+ } else if (session.finalTranscriptParts.at(-1) !== clean) {
287
+ session.finalTranscriptParts.push(clean);
288
+ }
289
+ session.interimTranscript = "";
290
+ session.browserTranscript = joinedTranscript(
291
+ session.finalTranscriptParts,
292
+ "",
293
+ );
294
+ }
295
+
296
+ function setInterimTranscript(session: VoiceSession, text: string): void {
297
+ session.interimTranscript = text.trim();
298
+ session.browserTranscript = joinedTranscript(
299
+ session.finalTranscriptParts,
300
+ session.interimTranscript,
301
+ );
302
+ }
303
+
232
304
  function stopMeter(session: VoiceSession): void {
233
305
  if (session.raf != null) {
234
306
  cancelAnimationFrame(session.raf);
@@ -698,7 +770,6 @@ export function installDesktopVoiceDictation(
698
770
  } catch (err) {
699
771
  console.warn("[voice-dictation] vocab learn-monitor failed:", err);
700
772
  }
701
- emit("voice:partial-transcript", { text }).catch(() => {});
702
773
  if (target.cleanupProvider) setFlowState("idle");
703
774
  return text;
704
775
  };
@@ -808,7 +879,7 @@ export function installDesktopVoiceDictation(
808
879
  abortPendingStart();
809
880
  return;
810
881
  }
811
- setFlowState("recording");
882
+ enterRecordingState();
812
883
  const mimeType = pickMimeType();
813
884
  const recorder = new MediaRecorder(stream, { mimeType });
814
885
  const next: VoiceSession = {
@@ -822,6 +893,8 @@ export function installDesktopVoiceDictation(
822
893
  mimeType: recorder.mimeType || mimeType,
823
894
  recognition: null,
824
895
  browserTranscript: "",
896
+ finalTranscriptParts: [],
897
+ interimTranscript: "",
825
898
  lastResultAt: 0,
826
899
  startedAt: Date.now(),
827
900
  stopping: false,
@@ -948,7 +1021,7 @@ export function installDesktopVoiceDictation(
948
1021
  abortPendingStart();
949
1022
  return;
950
1023
  }
951
- setFlowState("recording");
1024
+ enterRecordingState();
952
1025
  // Reset any prior partial transcript display in the flow-bar.
953
1026
  emit("voice:partial-transcript", { text: "" }).catch(() => {});
954
1027
  const next: VoiceSession = {
@@ -962,6 +1035,8 @@ export function installDesktopVoiceDictation(
962
1035
  mimeType: "",
963
1036
  recognition: null,
964
1037
  browserTranscript: "",
1038
+ finalTranscriptParts: [],
1039
+ interimTranscript: "",
965
1040
  lastResultAt: 0,
966
1041
  startedAt: Date.now(),
967
1042
  stopping: false,
@@ -1043,7 +1118,7 @@ export function installDesktopVoiceDictation(
1043
1118
  abortPendingStart();
1044
1119
  return;
1045
1120
  }
1046
- setFlowState("recording");
1121
+ enterRecordingState();
1047
1122
  emit("voice:partial-transcript", { text: "" }).catch(() => {});
1048
1123
  const next: VoiceSession = {
1049
1124
  kind: "whisper",
@@ -1056,6 +1131,8 @@ export function installDesktopVoiceDictation(
1056
1131
  mimeType: "",
1057
1132
  recognition: null,
1058
1133
  browserTranscript: "",
1134
+ finalTranscriptParts: [],
1135
+ interimTranscript: "",
1059
1136
  lastResultAt: 0,
1060
1137
  startedAt: Date.now(),
1061
1138
  stopping: false,
@@ -1112,7 +1189,7 @@ export function installDesktopVoiceDictation(
1112
1189
  abortPendingStart();
1113
1190
  return;
1114
1191
  }
1115
- setFlowState("recording");
1192
+ enterRecordingState();
1116
1193
  // Reset any prior partial transcript display in the flow-bar.
1117
1194
  emit("voice:partial-transcript", { text: "" }).catch(() => {});
1118
1195
  const recognition = new Ctor();
@@ -1131,6 +1208,8 @@ export function installDesktopVoiceDictation(
1131
1208
  mimeType: "",
1132
1209
  recognition,
1133
1210
  browserTranscript: "",
1211
+ finalTranscriptParts: [],
1212
+ interimTranscript: "",
1134
1213
  lastResultAt: 0,
1135
1214
  startedAt: Date.now(),
1136
1215
  stopping: false,
@@ -1203,10 +1282,6 @@ export function installDesktopVoiceDictation(
1203
1282
  // the tail because Web Speech only marks a segment as `isFinal`
1204
1283
  // after a confidence-threshold pass.
1205
1284
  next.browserTranscript = (finalSoFar + interim).trim();
1206
- // Stream the live transcript to the flow-bar.
1207
- emit("voice:partial-transcript", {
1208
- text: next.browserTranscript,
1209
- }).catch(() => {});
1210
1285
  };
1211
1286
  recognition.onerror = (ev) => {
1212
1287
  if (ev.error !== "no-speech" && ev.error !== "aborted") {
@@ -1732,31 +1807,28 @@ export function installDesktopVoiceDictation(
1732
1807
  if (!current || (current.kind !== "native" && current.kind !== "whisper"))
1733
1808
  return;
1734
1809
  if (current.cancelled || current.stopping) return;
1735
- current.browserTranscript = text.trim();
1810
+ setInterimTranscript(current, text);
1736
1811
  })
1737
1812
  .then((u) => unlistens.push(u))
1738
1813
  .catch(() => {});
1739
1814
  onFinalTranscript(({ text }) => {
1740
- // Final transcripts are ONLY emitted by Rust after `endAudio()`,
1741
- // which we call in stop(). At that point the session has been
1742
- // moved from `session` to `lingeringSession`, so route there
1743
- // exclusively. Do NOT fall back to the active `session`: a
1744
- // late-arriving final from the previous session would otherwise
1745
- // overwrite the new session's transcript with stale text.
1746
- const current =
1747
- lingeringSession &&
1748
- (lingeringSession.kind === "native" ||
1749
- lingeringSession.kind === "whisper")
1750
- ? lingeringSession
1815
+ // Whisper emits final segments throughout a long dictation whenever the
1816
+ // speaker pauses. Native speech emits its final after stop, when the
1817
+ // stopped session is parked in lingeringSession. Native final events do not
1818
+ // carry a session id, so never fall back to an active native session here:
1819
+ // a late final from the previous stop could otherwise be appended to a new
1820
+ // dictation.
1821
+ const current = lingeringSession
1822
+ ? lingeringSession
1823
+ : session && session.kind === "whisper" && !session.stopping
1824
+ ? session
1751
1825
  : null;
1752
1826
  if (!current) return;
1753
1827
  if (current.cancelled) return;
1754
- // Final beats partial — overwrite so a `complete_voice_dictation`
1755
- // from a late stop() picks up the better text.
1756
- current.browserTranscript = text.trim();
1757
- // If stop() is waiting on this event before lingering, trigger the
1758
- // finalize sequence now (paste → 1s linger → dismiss).
1759
- current.onNativeFinalize?.();
1828
+ appendFinalTranscript(current, text);
1829
+ if (current === lingeringSession) {
1830
+ current.onNativeFinalize?.();
1831
+ }
1760
1832
  })
1761
1833
  .then((u) => unlistens.push(u))
1762
1834
  .catch(() => {});
@@ -3,11 +3,7 @@ import { invoke } from "@tauri-apps/api/core";
3
3
  import { emit, listen } from "@tauri-apps/api/event";
4
4
  import { useEffect, useRef, useState } from "react";
5
5
 
6
- import {
7
- onAudioLevel,
8
- onFinalTranscript,
9
- onPartialTranscript,
10
- } from "../lib/transcription-engine";
6
+ import { onAudioLevel } from "../lib/transcription-engine";
11
7
 
12
8
  type FlowState = "idle" | "recording" | "processing" | "complete" | "error";
13
9
 
@@ -28,7 +24,6 @@ export function FlowBar() {
28
24
  // "idle" caused the bar to flash an "EN" language pill that never went
29
25
  // away if the start event was missed.
30
26
  const [state, setState] = useState<FlowState>("recording");
31
- const [partialTranscript, setPartialTranscript] = useState("");
32
27
  const canvasRef = useRef<HTMLCanvasElement | null>(null);
33
28
  const levelRef = useRef(0);
34
29
  const rafRef = useRef<number | null>(null);
@@ -63,23 +58,6 @@ export function FlowBar() {
63
58
  }),
64
59
  );
65
60
 
66
- trackListen(
67
- onPartialTranscript(({ text }) => {
68
- // Live transcript as the user speaks — rendered above the pill.
69
- // Empty payload clears the display (sent at session start/end).
70
- setPartialTranscript(text);
71
- }),
72
- );
73
-
74
- trackListen(
75
- onFinalTranscript(({ text }) => {
76
- // Final result from the recognizer (only fires after stop is
77
- // requested). Show it on the bar — the last word lingers there
78
- // for ~1s before voice-dictation.ts dismisses everything.
79
- if (text) setPartialTranscript(text);
80
- }),
81
- );
82
-
83
61
  return () => {
84
62
  stopped = true;
85
63
  unlistens.forEach((u) => {
@@ -166,29 +144,12 @@ export function FlowBar() {
166
144
  }, 250);
167
145
  };
168
146
 
169
- // The transcript chip is independent of the pill — it can linger on
170
- // its own after Fn release while the pill dismisses snappily. Voice-
171
- // dictation.ts emits an empty payload to clear it once the linger
172
- // window expires.
173
- const showTranscript = partialTranscript.length > 0;
174
-
175
147
  return (
176
148
  <div className="flow-bar-root">
177
- {showTranscript && (
178
- <div className="flow-bar-transcript">
179
- {/* <bdi> + unicode-bidi: plaintext (in CSS) keeps Latin text
180
- in its natural LTR order while the parent's direction:rtl
181
- clips overflow from the visual left. Without this, the
182
- last few characters of the newest words were being pushed
183
- past the visible right edge by bidi reordering. */}
184
- <bdi>{partialTranscript}</bdi>
185
- </div>
186
- )}
187
149
  {/* Pill is ALWAYS mounted — when state goes idle we fade the
188
- opacity to 0 (see CSS) instead of removing it from the DOM,
189
- so the transcript chip above doesn't reflow when the pill
190
- "goes away". Inner content keeps its last frame rendered
191
- during the fade so the canvas doesn't pop. */}
150
+ opacity to 0 (see CSS) instead of removing it from the DOM.
151
+ Inner content keeps its last frame rendered during the fade
152
+ so the canvas doesn't pop. */}
192
153
  <div className={`flow-bar flow-bar-${state}`}>
193
154
  {(state === "recording" || state === "idle") && (
194
155
  <div className="flow-bar-recording">
@@ -98,8 +98,11 @@ export function Toolbar() {
98
98
  trackListen(
99
99
  listen<boolean>("clips:toolbar-enabled", (ev) => {
100
100
  setEnabled(!!ev.payload);
101
+ setPendingAction(null);
101
102
  if (!ev.payload) {
102
103
  setDiskSpaceLevel("ok");
104
+ setPaused(false);
105
+ setElapsed(0);
103
106
  }
104
107
  }),
105
108
  );