Fix a bunch of microphones app reliability issues
<details> <summary>Prompts Used</summary> User: Review the Microphones app, focusing on error handling and reliability. Opus 5.5: > I've finished reading the Microphones code and its native pieces. The 2026-10-14 commit `b50898e` fixed a lot, and the main problem left is the same in most places: the app reports "Listening" or "captions on" but never checks that audio is actually arriving. When something fails partway, the session usually stops working with no visible sign. > > ## High: silent loss of function or data > > **1. In stock mode, other mic users can take the mic away for good.** > - `stopRawCapture()` tears down the single shared audio tap no matter who started it (`app/native/voice-control.ts:257`). > - The EvenHub mic router calls it whenever a mic-using EvenHub app goes to the background or closes (`app/apps/evenhub/mic-router.ts:88`). That kills Microphones' audio. > - Transcribe takes the mic without setting `voiceActivity` (`voice-control.ts:269`), so Microphones is never told to get it back. > - If capture starts while the session isn't ready, the native controller only records the request, so no audio ever flows. > - `checkExtendedCapture` (`mic-session.ts:456`) only covers 4-mic mode. Its comment says the stock path "is resumed by the voice-control bridge", but that only covers the session-dropped-while-active case. > - In 4-mic mode the problem runs the other way: Transcribe takes over the shared Kotlin listener (`GlassesSessionCore.kt:470`) while the glasses keep sending 4-mic frames, so Transcribe gets audio in the wrong format. > - **Fix:** track who holds the raw tap (like `captureHolders`), add a stock-mode branch to the 3 s check, and have Transcribe set the same "mic in use" signal as the assistant. > > **2. One bad utterance ends captions for the rest of the session.** > - The caption worker has no per-utterance error handling. Any exception from `recognize()` or `embed()` stops the worker for good (`CaptionEngineCore.kt:334`). > - `started` stays true, so TypeScript still thinks captions are running and recording carries on. The only sign is a "Captions failed:" status string. > - **Fix:** catch errors per utterance in `onUtterance` (`:367`). Report a fatal worker exit through its own callback so the session can restart the engine or mark captions inactive. > > **3. A crash or low-memory kill during recording leaves files that are never shown or deleted.** > - The WAV header is written with a data size of 0 (`FaceclawWavRecorder.kt:26`) and only corrected in `finish()`. > - The file is only linked to its conversation in `stopRecording` (`mic-session.ts:1200`), and the conversation row is never closed. > - Nothing cleans up `mic-recordings/`, so after a crash you get the same never-expired orphan files the last commit fixed for the captions-off case. > - **Fix:** link the file to its conversation when recording starts. In the boot-time cleanup: close conversations left open (`ended_at IS NULL`), repair the WAV header from the file length, and delete files no conversation points to. > > **4. When the transcoder's success callback fails, it deletes both copies of the recording.** > - `FaceclawAudioTranscoder.kt:66-83` deletes the WAV before calling `onDone`. If `onDone` throws, the catch block deletes the M4A too, and the database still points at the deleted WAV. > - Separately, the JS listener is referenced only from Java (`mic-session.ts:1208`). Under `markingMode: 'none'` that's the situation `mic-control.ts:74` warns about. > - I haven't confirmed the garbage-collection risk. `asr-model.ts` does the same thing and its downloads seem to work. But the delete order is wrong either way. > - `rediarize.ts` and `mic-models.ts` hold their listeners the same way. > > ## Medium > > 5. **The last utterance is lost when you close the app or turn captions off.** When the caption engine stops, it flushes the utterance in progress and posts it to the main thread. By the time it arrives, `stop()` has already run `finishSessionRow()` (`mic-session.ts:328`), so the line is never saved and is missing from the session insights. If decoding takes longer than the 2 s wait, the line can land in the *next* session with wrong times. Also, `stop()` can block the UI for up to 2 s. > 6. **The beam filter can mute 4-mic captions permanently.** The direction and signal strength only update on frames with a non-zero angle or strength (`:685`), and they never go stale. If you turn off "Direction detection", the last direction stays frozen. If it was outside the beam, `deliverExtendedAudio` drops all audio from then on (`:751-757`). Stock mode avoids this because it updates the strength on every frame. **Fix:** timestamp the fused direction and treat it as unknown after `TEMPLE_STALE_MS`. > 7. **The 4-mic health check looks at a phone-side flag, not at whether audio is arriving.** > - The check only re-arms when `isAudioCaptureActive()` is false. > - A configure command can be skipped while the session isn't ready even though `send()` returns true (`mic-control.ts:155`, `GlassesSessionCore.kt:890`). > - A temple can reboot on its own, or `apply()` can refuse to arm when the developer setting is turned off mid-session. > - Each of these leaves forwarding "active" with no frames arriving. > - **Fix:** you already track `templeFrameAtMs`. Re-send the config when an enabled temple has gone quiet, and show that in the status instead of "Listening (4-mic array)". > 8. **A failed start is permanent.** `start()` returns before installing any timers (`:280`). A failed mode switch mid-session (`:376`) leaves `forwardingProxy` null, so the health check exits at `:457` every time. **Fix:** stay in a "waiting for glasses" state and let the timer retry. > 9. **The caption feed can flip between left and right on every frame.** The choice at `:744` doesn't take the current side into account, and `strongerChanged` is just a `!==` check. Near the 1.5× threshold, each flip splices about 25 ms from the other temple (which runs on its own clock) into captions and the recording, either repeating or skipping audio. > > ## Lower > - **Unguarded speaker call (`:1034`):** the speaker rename for a pending introduction runs before the line is added. A database error drops the line and leaves the introduction pending, so the next 2 minutes of lines from unnamed speakers fail the same way. Add the line first, then do the extra steps each with its own error handling. > - **Speaker list grows forever:** every unknown voice adds a permanent "Speaker N" with a voiceprint. Retention never prunes them. `assign()` reloads and decodes every profile on every utterance, on the main thread. > - **Translation:** there's no timeout, so lines can wait indefinitely while a model download is stuck offline. The detected language is only saved after translation finishes (`:1132`). The "hysteresis" comment at `:1120` describes logic that doesn't exist: `conversationLang` is written but never read. > - **Frame fields ignored:** the sequence number (but `TempleAudioPipeline` assumes consecutive frames), the truncated flag, and `sampleRateHz`. The UI offers 8–48 kHz while the code assumes 16 kHz. > - **Duplicated model path:** `moonshineModelDir()` (`:1337`) copies `asr-model.ts`'s directory name. A model version bump there would quietly drop captions to "no ASR model". > - **Silent recording gaps:** `WavRecorder.append` swallows each write error, so a full disk means log spam and silent gaps in the recording. > > The startup order with cleanup on failure, the 90 s lease that recovers on its own from a lost STOP, and keeping the last good speaker cache all hold up well. > > I worked from the code only; nothing was tested on hardware. #6 relies on the protocol comment saying the angle is 0 when direction detection is off. > > I'd fix #1, #2, #5 and #6 first, since those are the ones you'd hit in normal use. Want me to start on them? <hr> User: Thanks. Let's fix all of the high and medium issues you found now. Opus 5.5: > All nine high and medium issues are fixed. TS typecheck, lint, 813 TS tests, the Kotlin host tests (4 new), the iOS Kotlin compile and `build.sh` all pass, and I confirmed the new code is in the APK. Nothing has been tested on the glasses yet, and nothing is committed. > > **High** > 1. **Mic taken by other apps:** > - The shared audio tap now tracks who's using it, so an EvenHub app backgrounding no longer cuts off Microphones. > - Transcribe now marks the mic as in use, like the assistant does, so Microphones and EvenHub apps step aside while it runs. > - A check every 3 s now covers both modes. It restarts a dead stock tap and switches back to stock mode if the developer setting is turned off mid-session. > 2. **One bad utterance ending captions:** a decode error now skips that utterance instead of stopping the caption engine (`CaptionEngineCore.kt`). If the engine dies anyway, it's restarted on the same conversation up to 3 times. After that, the Captions toggle shows off and the view says why; turning it on retries. > 3. **Recordings orphaned by a crash:** the recording is linked to its conversation as soon as it starts. At boot, the app closes conversations a crash left open, repairs their WAV headers (`FaceclawWavRecorder.repairHeader`), transcodes them, and deletes files no conversation points to. That cleanup matches by file name, so moving the app to other storage can't delete live recordings. > 4. **Transcoder deleting both copies:** a failure while recording the new file no longer deletes both files. The app now deletes the WAV itself, only after the conversation points at the M4A. The transcoder, re-diarize and model-download callbacks are now held by the app until they fire. > > **Medium** > 5. **Last utterance lost:** stopping captions no longer blocks the UI. The engine finishes the audio still queued, then sends a new `onStopped` callback, and the conversation stays open until that arrives (20 s fallback). A late line from a previous session won't show up in a newly reopened one. > 6. **Beam filter muting captions:** the direction of arrival only filters audio while it's under 1 s old, and it resets when "Direction detection" changes. > 7. **4-mic health check:** a temple that sends no audio for 5 s gets its config re-sent, to that temple only. Mic-control sends now report when they weren't actually queued. The status line says when no audio is arriving, but not after routine restarts like the voice dialog closing. > 8. **Start failures being permanent:** the session now waits ("Waiting for the glasses microphone...") and the 3 s check retries. Captions start only once the mic is up, so waiting doesn't create empty conversations. > 9. **Left/right flapping:** the caption feed only switches temples when the other one is clearly louder for 500 ms. > > Behaviour changes you'll notice: > - The watch shows "Listening…" during Transcribe. > - EvenHub mic apps get no audio while Transcribe runs. > - In 4-mic mode, an EvenHub mic app still takes over the single native audio listener. Fixing that needs a change on the Kotlin side, so I left it. > > On hardware, I'd test: > - The assistant, Transcribe and an EvenHub mic app taking the mic while Microphones is open, then giving it back. > - Reconnecting in both stock and 4-mic modes. > - Killing the app mid-recording, then checking the conversation is recovered at the next start. > > I saved these notes to memory as `microphones-reliability.md`. </details>
This commit is contained in:
1 parent
42c2d5cd55
commit
60c87521b0
23 files changed
+1101
-307
No files matched your search
@@ -13,7 +13,8 @@ import java.io.IOException
|
||||
/**
|
||||
* Transcodes a 16-bit PCM WAV conversation recording into AAC-LC in an M4A
|
||||
* container, cutting speech recordings to roughly a tenth of their WAV size.
|
||||
* Runs on its own thread; the source WAV is deleted on success when requested.
|
||||
* Runs on its own thread; when requested, the source WAV is deleted once the
|
||||
* listener's onDone has returned normally.
|
||||
*/
|
||||
class FaceclawAudioTranscoder(
|
||||
private val wavPath: String?,
|
||||
@@ -66,19 +67,28 @@ class FaceclawAudioTranscoder(
|
||||
private fun run() {
|
||||
try {
|
||||
transcode()
|
||||
val size = File(outputPath).length()
|
||||
if (deleteSourceOnSuccess) {
|
||||
File(wavPath).delete()
|
||||
}
|
||||
if (listener != null) {
|
||||
listener.onDone(outputPath, size)
|
||||
}
|
||||
} catch (t: Throwable) {
|
||||
Log.w(TAG, "transcode failed for $wavPath", t)
|
||||
File(outputPath).delete()
|
||||
if (listener != null) {
|
||||
listener.onError(if (t.message == null) t.toString() else t.message)
|
||||
try {
|
||||
listener?.onError(if (t.message == null) t.toString() else t.message)
|
||||
} catch (listenerError: Throwable) {
|
||||
Log.w(TAG, "transcode error listener failed", listenerError)
|
||||
}
|
||||
return
|
||||
}
|
||||
val size = File(outputPath).length()
|
||||
try {
|
||||
listener?.onDone(outputPath, size)
|
||||
} catch (t: Throwable) {
|
||||
// The caller never took up the new file, so whatever it recorded still names the
|
||||
// source: keep that and drop the output, rather than lose both.
|
||||
Log.w(TAG, "transcode listener failed for $wavPath", t)
|
||||
File(outputPath).delete()
|
||||
return
|
||||
}
|
||||
if (deleteSourceOnSuccess) {
|
||||
File(wavPath).delete()
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@@ -207,7 +207,7 @@ class FaceclawBleCommunicator(context: Context, rightAddress: String?, leftAddre
|
||||
|
||||
fun removeMicStatusListener(listener: FaceclawMicStatusListener?) = core.removeMicStatusListener(listener)
|
||||
|
||||
fun sendFaceclawMicControl(record: ByteArray?, label: String, rightTemple: Boolean, leftTemple: Boolean) =
|
||||
fun sendFaceclawMicControl(record: ByteArray?, label: String, rightTemple: Boolean, leftTemple: Boolean): Boolean =
|
||||
core.sendFaceclawMicControl(record, label, rightTemple, leftTemple)
|
||||
|
||||
fun startG2AudioForwarding(listener: FaceclawAudioPacketListener?): Boolean = core.startG2AudioForwarding(listener)
|
||||
|
||||
@@ -68,6 +68,11 @@ class FaceclawCaptionEngine {
|
||||
session.stop()
|
||||
}
|
||||
|
||||
/** Stop without blocking; the listener's onStopped follows the trailing utterance. */
|
||||
fun requestStop() {
|
||||
session.requestStop()
|
||||
}
|
||||
|
||||
/** Push decoded 16 kHz mono S16LE PCM (any chunking). */
|
||||
fun acceptPcm(pcm16le: ByteArray?) {
|
||||
session.acceptPcm(pcm16le)
|
||||
|
||||
@@ -718,6 +718,72 @@ class FaceclawConversationStore private constructor(context: Context) : SQLiteOp
|
||||
}
|
||||
}
|
||||
|
||||
// ---- crash recovery ----
|
||||
|
||||
/**
|
||||
* Close out sessions a dead process left open (ended_at NULL): the end becomes the last
|
||||
* stored segment's end (or the start), and the tone average comes from the stored segments.
|
||||
* excludeIdsJson is a JSON array of session ids that are legitimately still open. Returns
|
||||
* the sessions it closed as a JSON array of {id, audioPath, audioCodec}.
|
||||
*/
|
||||
fun closeInterruptedSessions(excludeIdsJson: String?): String {
|
||||
val exclude = HashSet<Long>()
|
||||
if (excludeIdsJson != null && !excludeIdsJson.isEmpty()) {
|
||||
val ids = JSONArray(excludeIdsJson)
|
||||
for (i in 0 until ids.length()) {
|
||||
exclude.add(ids.getLong(i))
|
||||
}
|
||||
}
|
||||
val db = writableDatabase
|
||||
val closed = JSONArray()
|
||||
val closedIds = ArrayList<Long>()
|
||||
db.rawQuery("SELECT id, audio_path, audio_codec FROM sessions WHERE ended_at IS NULL", null).use { c ->
|
||||
while (c.moveToNext()) {
|
||||
val id = c.getLong(0)
|
||||
if (id in exclude) continue
|
||||
closedIds.add(id)
|
||||
val row = JSONObject()
|
||||
row.put("id", id)
|
||||
row.put("audioPath", if (c.isNull(1)) JSONObject.NULL else c.getString(1))
|
||||
row.put("audioCodec", if (c.isNull(2)) JSONObject.NULL else c.getString(2))
|
||||
closed.put(row)
|
||||
}
|
||||
}
|
||||
for (id in closedIds) {
|
||||
db.execSQL(
|
||||
"UPDATE sessions SET"
|
||||
+ " ended_at = COALESCE((SELECT MAX(ended_at) FROM segments WHERE session_id = ?), started_at),"
|
||||
+ " avg_sentiment = COALESCE((SELECT AVG(sentiment) FROM segments WHERE session_id = ?), 0)"
|
||||
+ " WHERE id = ?",
|
||||
arrayOf<Any?>(id, id, id))
|
||||
}
|
||||
return closed.toString()
|
||||
}
|
||||
|
||||
/**
|
||||
* Delete recordings in dirPath that no session's audio_path names: files a session row never
|
||||
* learned about, or partial transcoder output. Matched by file name, so a storage move that
|
||||
* changed the directory's absolute path can't make live recordings look orphaned. Only call
|
||||
* while no recording or transcode is in flight. Returns how many files were deleted.
|
||||
*/
|
||||
fun deleteUnreferencedRecordings(dirPath: String?): Int {
|
||||
if (dirPath == null) return 0
|
||||
val files = File(dirPath).listFiles() ?: return 0
|
||||
val referenced = HashSet<String>()
|
||||
readableDatabase.rawQuery("SELECT audio_path FROM sessions WHERE audio_path IS NOT NULL", null).use { c ->
|
||||
while (c.moveToNext()) {
|
||||
referenced.add(File(c.getString(0)).name)
|
||||
}
|
||||
}
|
||||
var deleted = 0
|
||||
for (file in files) {
|
||||
if (file.isFile && file.name !in referenced && file.delete()) {
|
||||
deleted++
|
||||
}
|
||||
}
|
||||
return deleted
|
||||
}
|
||||
|
||||
fun deleteSession(sessionId: Long) {
|
||||
val db = writableDatabase
|
||||
val path = getSessionAudioPath(sessionId)
|
||||
|
||||
@@ -39,7 +39,7 @@ export class VoiceDraft {
|
||||
this.phase = "listening";
|
||||
this.status = "Listening... release to send";
|
||||
this.ownsVoiceActivity = true;
|
||||
voiceActivity.setActive(true);
|
||||
voiceActivity.setActive(true, "voice-draft");
|
||||
this.unsubscribe.push(voiceControlBridge.onTranscript((event) => {
|
||||
if (!this.active || generation !== this.generation) return;
|
||||
if (event.isFinal) {
|
||||
@@ -129,7 +129,7 @@ export class VoiceDraft {
|
||||
}
|
||||
if (this.ownsVoiceActivity) {
|
||||
this.ownsVoiceActivity = false;
|
||||
voiceActivity.setActive(false);
|
||||
voiceActivity.setActive(false, "voice-draft");
|
||||
}
|
||||
this.changed();
|
||||
}
|
||||
|
||||
@@ -17,6 +17,9 @@ import { voiceActivity } from "../../ui/shell/voice-activity";
|
||||
|
||||
declare const com: any;
|
||||
|
||||
// The router's claim on the shared raw-PCM tap (the Microphones app holds its own).
|
||||
const RAW_CAPTURE_OWNER = "evenhub-mic-router";
|
||||
|
||||
export type EvenHubMicClient = {
|
||||
readonly windowId: string;
|
||||
isForeground(): boolean;
|
||||
@@ -83,9 +86,9 @@ class EvenHubMicRouter {
|
||||
// eligibility change so it self-heals after an STT preemption. Returns
|
||||
// false if STT currently owns the mic; a later trigger (assistant modal
|
||||
// close, foreground/screen change) re-tries.
|
||||
voiceControlBridge.startRawCapture({ communicator: activeCommunicator() });
|
||||
voiceControlBridge.startRawCapture({ communicator: activeCommunicator(), owner: RAW_CAPTURE_OWNER });
|
||||
} else {
|
||||
voiceControlBridge.stopRawCapture();
|
||||
voiceControlBridge.stopRawCapture(RAW_CAPTURE_OWNER);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
@@ -39,7 +39,7 @@ export class CaptionsLayer implements Layer {
|
||||
|
||||
image.drawText(font, 12, 4, "Captions", 220);
|
||||
if (!state.captionsActive) {
|
||||
image.drawText(font, 12, 26, "Captions are off (enable in the Microphones menu).", 130);
|
||||
image.drawText(font, 12, 26, state.captionsNote, 130);
|
||||
}
|
||||
|
||||
const rows = this.buildRows(font, width - 36, state.captionLines);
|
||||
|
||||
@@ -17,7 +17,9 @@ const microphonesApp: AppDefinition = {
|
||||
title: "Microphones",
|
||||
icon: "mic",
|
||||
boot: () => {
|
||||
// Prune expired captions/recordings per the retention settings.
|
||||
// Close out conversations and recordings a dead process left open, then
|
||||
// prune expired captions/recordings per the retention settings.
|
||||
micSession.recoverInterruptedConversations();
|
||||
void micSession.applyRetentionSweep();
|
||||
},
|
||||
launch: async (ctx) => {
|
||||
|
||||
@@ -125,6 +125,17 @@ class MicArrayController {
|
||||
return true;
|
||||
}
|
||||
|
||||
/**
|
||||
* Re-send CONFIGURE to one temple whose stream went quiet (a lost write, or
|
||||
* a temple that restarted on its own), leaving the other temple's stream
|
||||
* untouched.
|
||||
*/
|
||||
reconfigureTemple(config: MicConfig, side: MicSide): boolean {
|
||||
if (!templeActive(config, side)) return false;
|
||||
if (config.armHardware && !micControlSupported()) return false;
|
||||
return this.send("configure", config, `configure-${side}`, side === "right", side === "left");
|
||||
}
|
||||
|
||||
query(): boolean {
|
||||
return this.send("query", undefined, "query");
|
||||
}
|
||||
@@ -152,8 +163,8 @@ class MicArrayController {
|
||||
try {
|
||||
this.ensureStatusProxy(communicator);
|
||||
const record = encodeMicControl(op, config);
|
||||
communicator.sendFaceclawMicControl(toJavaBytes(record), label, rightTemple, leftTemple);
|
||||
return true;
|
||||
// False when the session wasn't ready and nothing was queued.
|
||||
return communicator.sendFaceclawMicControl(toJavaBytes(record), label, rightTemple, leftTemple) !== false;
|
||||
} catch (error) {
|
||||
console.warn(`mic control ${label} failed: ${error}`);
|
||||
return false;
|
||||
|
||||
@@ -75,6 +75,9 @@ export type MicModelState = {
|
||||
};
|
||||
|
||||
const downloaders = new Map<MicModelId, any>();
|
||||
// Each download's listener is referenced only from Java while it runs; hold
|
||||
// it here so it isn't GC'd before its callbacks.
|
||||
const downloadListeners = new Map<MicModelId, any>();
|
||||
const progress = new Map<MicModelId, number>();
|
||||
const stateListeners = new Set<(id: MicModelId, state: MicModelState) => void>();
|
||||
|
||||
@@ -152,6 +155,7 @@ function downloadNextFile(id: MicModelId): void {
|
||||
const nextFile = definition.files.find((file) => !isFilePresent(id, file));
|
||||
if (!nextFile) {
|
||||
downloaders.delete(id);
|
||||
downloadListeners.delete(id);
|
||||
notifyStateChanged(id);
|
||||
return;
|
||||
}
|
||||
@@ -169,9 +173,11 @@ function downloadNextFile(id: MicModelId): void {
|
||||
onError: (message: string) => {
|
||||
console.error(`Mic model download failed (${id}/${nextFile.name}): ${message}`);
|
||||
downloaders.delete(id);
|
||||
downloadListeners.delete(id);
|
||||
notifyStateChanged(id);
|
||||
},
|
||||
});
|
||||
downloadListeners.set(id, listener);
|
||||
try {
|
||||
const downloader = new com.faceclaw.app.FaceclawModelDownloader(
|
||||
`${definition.baseUrl}${nextFile.remoteName}`,
|
||||
@@ -186,6 +192,7 @@ function downloadNextFile(id: MicModelId): void {
|
||||
// Without this, a dead entry would pin the model at "downloading".
|
||||
console.error(`Mic model download failed to start (${id}/${nextFile.name}): ${String(error)}`);
|
||||
downloaders.delete(id);
|
||||
downloadListeners.delete(id);
|
||||
notifyStateChanged(id);
|
||||
}
|
||||
}
|
||||
@@ -195,6 +202,7 @@ export function cancelMicModelDownload(id: MicModelId): void {
|
||||
if (!downloader) return;
|
||||
downloader.cancel();
|
||||
downloaders.delete(id);
|
||||
downloadListeners.delete(id);
|
||||
notifyStateChanged(id);
|
||||
}
|
||||
|
||||
|
||||
+667
-242
File diff suppressed because it is too large.
Load diff
@@ -453,8 +453,9 @@ export function createMicrophonesAppWindow(options: InProcessAppOptions): InProc
|
||||
}),
|
||||
configToggleItem(
|
||||
"Captions",
|
||||
() => captionsEnabledSetting.get(),
|
||||
() => micSession.setCaptionsEnabled(!captionsEnabledSetting.get()),
|
||||
// Shows off after repeated engine failures, so turning it on retries.
|
||||
() => micSession.captionsEnabled(),
|
||||
() => micSession.setCaptionsEnabled(!micSession.captionsEnabled()),
|
||||
undefined,
|
||||
captionsEnabledSetting.description,
|
||||
),
|
||||
|
||||
@@ -35,6 +35,10 @@ function store(): any {
|
||||
return com.faceclaw.app.FaceclawConversationStore.getInstance(context);
|
||||
}
|
||||
|
||||
// Each run's listener is referenced only from Java while the diarizer works;
|
||||
// hold it here so it isn't GC'd before its callbacks.
|
||||
const activeListeners = new Set<any>();
|
||||
|
||||
export function rediarizeSession(
|
||||
sessionId: number,
|
||||
options: { numSpeakers?: number; threshold?: number },
|
||||
@@ -54,6 +58,7 @@ export function rediarizeSession(
|
||||
const listener = new com.faceclaw.app.FaceclawDiarizer.Listener({
|
||||
onProgress: (progress: number) => callbacks.onProgress(Number(progress)),
|
||||
onDone: (turnsJson: string) => {
|
||||
activeListeners.delete(listener);
|
||||
try {
|
||||
const turns = JSON.parse(String(turnsJson)) as Turn[];
|
||||
callbacks.onDone(applyTurns(sessionId, turns));
|
||||
@@ -61,17 +66,26 @@ export function rediarizeSession(
|
||||
callbacks.onError(`Applying diarization failed: ${String(error)}`);
|
||||
}
|
||||
},
|
||||
onError: (message: string) => callbacks.onError(String(message)),
|
||||
onError: (message: string) => {
|
||||
activeListeners.delete(listener);
|
||||
callbacks.onError(String(message));
|
||||
},
|
||||
});
|
||||
const diarizer = new com.faceclaw.app.FaceclawDiarizer(
|
||||
segmentationModel,
|
||||
embeddingModel,
|
||||
audioPath,
|
||||
options.numSpeakers ?? -1,
|
||||
options.threshold ?? 0.5,
|
||||
listener,
|
||||
);
|
||||
diarizer.start();
|
||||
activeListeners.add(listener);
|
||||
try {
|
||||
const diarizer = new com.faceclaw.app.FaceclawDiarizer(
|
||||
segmentationModel,
|
||||
embeddingModel,
|
||||
audioPath,
|
||||
options.numSpeakers ?? -1,
|
||||
options.threshold ?? 0.5,
|
||||
listener,
|
||||
);
|
||||
diarizer.start();
|
||||
} catch (error) {
|
||||
activeListeners.delete(listener);
|
||||
callbacks.onError(`Re-diarization failed to start: ${String(error)}`);
|
||||
}
|
||||
}
|
||||
|
||||
type StoredSegment = {
|
||||
|
||||
@@ -52,6 +52,7 @@ import { isPreviewOnlyMode } from "../phone-ui/onboarding-state";
|
||||
import { claimGlassesOnboardingAutoLaunch, ONBOARDING_APP_ID } from "../apps/onboarding/onboarding-progress";
|
||||
import { beginRenderPass, endRenderPass } from "../util/render-freshness";
|
||||
import { voiceControlBridge } from "../native/voice-control";
|
||||
import { voiceActivity } from "../ui/shell/voice-activity";
|
||||
import { G2_LENS_HEIGHT, G2_LENS_WIDTH, GrayImage } from "../graphics/image";
|
||||
import { flattenPlanesWithDraws, planesFingerprint, type Plane } from "../graphics/plane";
|
||||
import { prepareFrameDraws } from "../graphics/glyph-wire";
|
||||
@@ -163,6 +164,8 @@ const SHELL_REFRESH_INTERVAL_MS = 60_000;
|
||||
// schedules a refresh), so this just covers anything that slips through.
|
||||
const PREVIEW_INTERVAL_MS = 1_000;
|
||||
const SCREEN_TIMEOUT_CHECK_MS = 1_000;
|
||||
// voiceActivity holder for Transcribe's continuous capture.
|
||||
const CONTINUOUS_CAPTURE_HOLDER = "continuous-capture";
|
||||
// The lock screen is a short notice, not something to read at leisure; it
|
||||
// ignores the screen-timeout setting (including "never").
|
||||
const LOCK_SCREEN_TIMEOUT_MS = 15_000;
|
||||
@@ -2058,11 +2061,16 @@ class DashboardController {
|
||||
}
|
||||
|
||||
private startContinuousVoiceCapture(): void {
|
||||
// Continuous capture (Transcribe) owns the glasses mic for as long as it
|
||||
// runs. Claim it before the capture starts so the Microphones app and
|
||||
// EvenHub mic apps stand down first, as they do for the voice dialog.
|
||||
voiceActivity.setActive(true, CONTINUOUS_CAPTURE_HOLDER);
|
||||
void this.beginVoiceCapture("continuous");
|
||||
}
|
||||
|
||||
private stopContinuousVoiceCapture(): void {
|
||||
voiceControlBridge.stopContinuousCapture();
|
||||
voiceActivity.setActive(false, CONTINUOUS_CAPTURE_HOLDER);
|
||||
}
|
||||
|
||||
/**
|
||||
|
||||
+39
-13
@@ -122,13 +122,17 @@ export class FaceclawVoiceControlBridge {
|
||||
private captureEndpointing = false;
|
||||
// Non-null while a cloud provider owns the transcript; Java only decodes PCM.
|
||||
private cloudClient: CloudSttClient | null = null;
|
||||
// Raw-PCM tap (EvenHub mic apps): when active, the controller runs in the
|
||||
// decode-only "cloud" mode and every decoded PCM frame is broadcast to these
|
||||
// listeners. STT capture preempts it — a live assistant/transcribe session
|
||||
// owns the single G2 mic and the raw tap is stopped for its duration.
|
||||
// Raw-PCM tap (EvenHub mic apps, the Microphones app): when active, the
|
||||
// controller runs in the decode-only "cloud" mode and every decoded PCM
|
||||
// frame is broadcast to these listeners. STT capture preempts it — a live
|
||||
// assistant/transcribe session owns the single G2 mic and the raw tap is
|
||||
// stopped for its duration.
|
||||
private readonly rawPcmListeners = new Set<RawPcmListener>();
|
||||
private readonly frameMetaListeners = new Set<FrameMetaListener>();
|
||||
private rawActive = false;
|
||||
// Who wants the raw tap. It runs while any owner does, so one owner letting
|
||||
// go (an EvenHub app backgrounding) doesn't silence another.
|
||||
private readonly rawOwners = new Set<string>();
|
||||
// Set when the current capture session runs "my voice only" verification;
|
||||
// a non-wearer verdict arrives just before the final transcript and makes
|
||||
// the bridge swallow it.
|
||||
@@ -220,11 +224,33 @@ export class FaceclawVoiceControlBridge {
|
||||
}
|
||||
|
||||
/**
|
||||
* Start the decode-only raw-PCM tap for an EvenHub mic app. Fails (returns
|
||||
* false) if an STT capture already owns the mic; the caller re-tries when its
|
||||
* eligibility is re-evaluated. Idempotent while already running.
|
||||
* Start the decode-only raw-PCM tap on behalf of `owner`. Fails (returns
|
||||
* false) if an STT capture already owns the mic; the owner stays registered
|
||||
* and re-tries later (the EvenHub router when eligibility is re-evaluated,
|
||||
* the Microphones app from its capture check). Idempotent while already
|
||||
* running.
|
||||
*/
|
||||
startRawCapture(options: { communicator: any }): boolean {
|
||||
startRawCapture(options: { communicator: any; owner: string }): boolean {
|
||||
if (!global.isAndroid) return false;
|
||||
this.rawOwners.add(options.owner);
|
||||
return this.startRawTap(options.communicator);
|
||||
}
|
||||
|
||||
/** `owner` no longer wants the raw tap; it stops once no owner does. */
|
||||
stopRawCapture(owner: string): void {
|
||||
this.rawOwners.delete(owner);
|
||||
if (this.rawOwners.size === 0) this.endRawTap();
|
||||
}
|
||||
|
||||
/**
|
||||
* Whether the raw tap is running and its glasses mic enable still holds
|
||||
* (or is still coming up). False after an STT preemption or a session drop.
|
||||
*/
|
||||
isRawCaptureLive(): boolean {
|
||||
return this.rawActive && this.micIsLive();
|
||||
}
|
||||
|
||||
private startRawTap(communicator: any): boolean {
|
||||
if (!global.isAndroid) return false;
|
||||
if (this.rawActive && this.micIsLive()) return true;
|
||||
// An STT holder owns the mic even while its stream waits on a new glasses
|
||||
@@ -234,7 +260,7 @@ export class FaceclawVoiceControlBridge {
|
||||
if (this.started || this.rawActive) this.teardownCapture();
|
||||
this.suspendedRaw = false;
|
||||
this.ensureController();
|
||||
this.controller?.setCommunicator(options.communicator);
|
||||
this.controller?.setCommunicator(communicator);
|
||||
// The raw tap is strictly the G2 stream; clear any phone-mic flag a
|
||||
// preview-mode capture may have left on the shared controller.
|
||||
this.controller?.setUsePhoneMic(false);
|
||||
@@ -254,7 +280,7 @@ export class FaceclawVoiceControlBridge {
|
||||
return true;
|
||||
}
|
||||
|
||||
stopRawCapture(): void {
|
||||
private endRawTap(): void {
|
||||
this.suspendedRaw = false;
|
||||
if (!this.rawActive) return;
|
||||
this.rawActive = false;
|
||||
@@ -266,7 +292,7 @@ export class FaceclawVoiceControlBridge {
|
||||
if (!global.isAndroid) return;
|
||||
// STT preempts the raw-PCM tap: a live transcription session owns the mic,
|
||||
// and an EvenHub app gets no audio for its duration.
|
||||
if (this.rawActive) this.stopRawCapture();
|
||||
if (this.rawActive) this.endRawTap();
|
||||
// Whether the mic is already running is a question about the stream, not
|
||||
// about this set: a holder whose capture died with the glasses session
|
||||
// (or whose mic enable was refused) is still in it. Trusting the set here
|
||||
@@ -428,8 +454,8 @@ export class FaceclawVoiceControlBridge {
|
||||
for (const holder of holders) {
|
||||
this.acquireCapture(holder, { ...options, endpointing: this.captureEndpointing });
|
||||
}
|
||||
if (resumeRaw && !holders.length) {
|
||||
this.startRawCapture({ communicator: options.communicator });
|
||||
if (resumeRaw && !holders.length && this.rawOwners.size > 0) {
|
||||
this.startRawTap(options.communicator);
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@@ -1,8 +1,13 @@
|
||||
/**
|
||||
* A tiny global signal for whether a voice-input modal (the assistant dialog)
|
||||
* is currently open. The shell sets it; the EvenHub mic router reads it to
|
||||
* decide whether an app may hold the glasses microphone — an app receives no
|
||||
* audio while the assistant is active, since both share the one G2 mic.
|
||||
* A tiny global signal for whether speech capture owns the glasses
|
||||
* microphone: the assistant's voice dialog, a dictation draft, or
|
||||
* Transcribe's continuous capture. The EvenHub mic router and the
|
||||
* Microphones app read it to yield the one G2 mic while it is active — an
|
||||
* app receives no audio meanwhile, since they all share it.
|
||||
*
|
||||
* Each user holds it under its own name, so one finishing doesn't clear
|
||||
* another's hold. Set it before starting the capture and clear it after
|
||||
* releasing, so other mic users stand down first and resume after.
|
||||
*
|
||||
* Kept as a dependency-free leaf so both the shell and the EvenHub layer can
|
||||
* import it without a cycle.
|
||||
@@ -10,16 +15,19 @@
|
||||
|
||||
type Listener = (active: boolean) => void;
|
||||
|
||||
let active = false;
|
||||
const holders = new Set<string>();
|
||||
const listeners = new Set<Listener>();
|
||||
|
||||
export const voiceActivity = {
|
||||
isActive(): boolean {
|
||||
return active;
|
||||
return holders.size > 0;
|
||||
},
|
||||
setActive(next: boolean): void {
|
||||
if (active === next) return;
|
||||
active = next;
|
||||
setActive(next: boolean, holder = "voice-input"): void {
|
||||
const was = holders.size > 0;
|
||||
if (next) holders.add(holder);
|
||||
else holders.delete(holder);
|
||||
const active = holders.size > 0;
|
||||
if (active === was) return;
|
||||
listeners.forEach((listener) => listener(active));
|
||||
},
|
||||
subscribe(listener: Listener): () => void {
|
||||
|
||||
@@ -13,6 +13,30 @@ constructor(
|
||||
private val sampleRate: Int,
|
||||
private val channels: Int,
|
||||
) {
|
||||
companion object {
|
||||
/**
|
||||
* Rewrite the header of a recording whose writer never finished (the process died
|
||||
* mid-recording, leaving the empty header written at start) to cover the PCM actually on
|
||||
* disk, trimming any partial sample frame. Returns the PCM byte count, or -1 when the
|
||||
* file doesn't exist.
|
||||
*/
|
||||
@JvmStatic
|
||||
@Throws(IOException::class)
|
||||
fun repairHeader(path: String, sampleRate: Int, channels: Int): Long {
|
||||
val target = File(path)
|
||||
if (!target.exists()) return -1
|
||||
val headerBytes = BinaryEncoding.wavHeader(0, sampleRate, channels).size.toLong()
|
||||
val frameBytes = channels * 2L
|
||||
val pcmBytes = maxOf(0L, target.length() - headerBytes) / frameBytes * frameBytes
|
||||
RandomAccessFile(target, "rw").use {
|
||||
it.setLength(headerBytes + pcmBytes)
|
||||
it.seek(0)
|
||||
it.write(BinaryEncoding.wavHeader(minOf(pcmBytes, Int.MAX_VALUE.toLong()).toInt(), sampleRate, channels))
|
||||
}
|
||||
return pcmBytes
|
||||
}
|
||||
}
|
||||
|
||||
private var file: RandomAccessFile?
|
||||
private var dataBytes = 0L
|
||||
|
||||
|
||||
+57
-7
@@ -312,6 +312,21 @@ class CaptionSession(
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Stops without waiting. The worker still flushes the trailing utterance and releases the
|
||||
* models, then reports onStopped, so a caller can close out after the last line arrives
|
||||
* instead of blocking its thread on the decode.
|
||||
*/
|
||||
fun requestStop() {
|
||||
lock.withLock {
|
||||
if (!started) {
|
||||
return
|
||||
}
|
||||
started = false
|
||||
}
|
||||
queue.wake()
|
||||
}
|
||||
|
||||
/** Push decoded 16 kHz mono S16LE PCM (any chunking). */
|
||||
fun acceptPcm(pcm16le: ByteArray?) {
|
||||
if (!started || pcm16le == null || pcm16le.size < 2) {
|
||||
@@ -322,6 +337,7 @@ class CaptionSession(
|
||||
|
||||
private fun runLoop(finished: Latch) {
|
||||
workerThreadName = currentThreadName()
|
||||
var failure: String? = null
|
||||
try {
|
||||
loadModels()
|
||||
segmenter.reset()
|
||||
@@ -329,18 +345,35 @@ class CaptionSession(
|
||||
val chunk = queue.take(250) { started } ?: continue
|
||||
segmenter.processChunk(chunk)
|
||||
}
|
||||
// Flush a trailing utterance so its text isn't lost on stop.
|
||||
// Finish the audio already queued when the stop landed, then flush the trailing
|
||||
// utterance so its text isn't lost on stop.
|
||||
while (true) {
|
||||
val chunk = queue.take(0) { false } ?: break
|
||||
segmenter.processChunk(chunk)
|
||||
}
|
||||
segmenter.flush()
|
||||
} catch (t: Throwable) {
|
||||
logError("caption engine failed", t)
|
||||
emitStatus("Captions failed: " + t.message)
|
||||
failure = t.message ?: t.toString()
|
||||
emitStatus("Captions failed: $failure")
|
||||
} finally {
|
||||
decoder?.release()
|
||||
try {
|
||||
decoder?.release()
|
||||
} catch (t: Throwable) {
|
||||
logError("caption model release failed", t)
|
||||
}
|
||||
decoder = null
|
||||
lock.withLock {
|
||||
workerFinished = null
|
||||
workerThreadName = null
|
||||
// A worker that died on its own leaves the session stopped, so the owner can
|
||||
// tell (and acceptPcm stops queueing for nobody).
|
||||
if (workerFinished === finished) {
|
||||
started = false
|
||||
workerFinished = null
|
||||
workerThreadName = null
|
||||
}
|
||||
}
|
||||
// Posted before the latch opens, so a blocking stop() returns after it.
|
||||
emitStopped(failure)
|
||||
finished.countDown()
|
||||
}
|
||||
}
|
||||
@@ -366,8 +399,20 @@ class CaptionSession(
|
||||
|
||||
override fun onUtterance(samples: ShortArray, length: Int, startMs: Long, endMs: Long, peakRms: Double) {
|
||||
val currentDecoder = decoder ?: return
|
||||
val text = currentDecoder.recognize(samples, length)
|
||||
val embedding = currentDecoder.embed(samples, length)
|
||||
// One utterance the models choke on must not end captioning for the session: deliver
|
||||
// what did decode and keep listening.
|
||||
val text = try {
|
||||
currentDecoder.recognize(samples, length)
|
||||
} catch (t: Throwable) {
|
||||
logError("caption recognition failed", t)
|
||||
""
|
||||
}
|
||||
val embedding = try {
|
||||
currentDecoder.embed(samples, length)
|
||||
} catch (t: Throwable) {
|
||||
logError("caption voice-print failed", t)
|
||||
null
|
||||
}
|
||||
val currentListener = listener ?: return
|
||||
val embeddingCopy = embedding?.copyOf()
|
||||
host.dispatcher.post { currentListener.onUtterance(text, embeddingCopy, startMs, endMs, peakRms) }
|
||||
@@ -377,4 +422,9 @@ class CaptionSession(
|
||||
val currentListener = listener ?: return
|
||||
host.dispatcher.post { currentListener.onStatus(status) }
|
||||
}
|
||||
|
||||
private fun emitStopped(error: String?) {
|
||||
val currentListener = listener ?: return
|
||||
host.dispatcher.post { currentListener.onStopped(error) }
|
||||
}
|
||||
}
|
||||
+6
@@ -19,4 +19,10 @@ interface FaceclawCaptionEngineListener {
|
||||
fun onSpeechStart(startMs: Long): Unit
|
||||
|
||||
fun onStatus(status: String?): Unit
|
||||
|
||||
/**
|
||||
* The worker has exited, after delivering any trailing utterance. error is null after
|
||||
* stop()/requestStop(), or the failure that ended captioning on its own.
|
||||
*/
|
||||
fun onStopped(error: String?): Unit
|
||||
}
|
||||
+6
-4
@@ -885,16 +885,17 @@ class GlassesSessionCore(
|
||||
* Queue a CFW mic_control record (['M','C',ver,op,...]) as a settings
|
||||
* field-103 write to both temples, or to a single one. Fire-and-forget:
|
||||
* the firmware answers with a field-104 status notify per temple, which
|
||||
* arrives through addMicStatusListener.
|
||||
* arrives through addMicStatusListener. Returns false when nothing was
|
||||
* queued (no session, or a malformed record).
|
||||
*/
|
||||
fun sendFaceclawMicControl(record: ByteArray?, label: String, rightTemple: Boolean, leftTemple: Boolean) {
|
||||
fun sendFaceclawMicControl(record: ByteArray?, label: String, rightTemple: Boolean, leftTemple: Boolean): Boolean {
|
||||
if (record == null || record.size < 4) {
|
||||
return
|
||||
return false
|
||||
}
|
||||
monitor.withLock {
|
||||
if (!running || !sessionReady) {
|
||||
logLine("skip mic control ($label); session not ready")
|
||||
return
|
||||
return false
|
||||
}
|
||||
if (rightTemple) {
|
||||
pendingMessages.addLast(messageBuilder.faceclawMicControl(record, label, false))
|
||||
@@ -905,6 +906,7 @@ class GlassesSessionCore(
|
||||
logLine("queue mic control $label")
|
||||
}
|
||||
interruptibleSleep.interrupt()
|
||||
return true
|
||||
}
|
||||
|
||||
/**
|
||||
|
||||
@@ -181,7 +181,7 @@ class IosGlassesSession internal constructor(
|
||||
|
||||
fun removeMicStatusListener(listener: FaceclawMicStatusListener?) = core.removeMicStatusListener(listener)
|
||||
|
||||
fun sendFaceclawMicControl(record: NSData?, label: String, rightTemple: Boolean, leftTemple: Boolean) =
|
||||
fun sendFaceclawMicControl(record: NSData?, label: String, rightTemple: Boolean, leftTemple: Boolean): Boolean =
|
||||
core.sendFaceclawMicControl(record?.byteArray(), label, rightTemple, leftTemple)
|
||||
|
||||
fun startG2AudioForwarding(listener: FaceclawAudioPacketListener?): Boolean =
|
||||
|
||||
@@ -0,0 +1,24 @@
|
||||
package com.faceclaw.app
|
||||
|
||||
import java.io.File
|
||||
import kotlin.test.Test
|
||||
import kotlin.test.assertEquals
|
||||
|
||||
class WavRecorderRepairTest {
|
||||
@Test
|
||||
fun repairsTheHeaderOfARecordingThatWasNeverFinished() {
|
||||
val file = File.createTempFile("unfinished", ".wav")
|
||||
try {
|
||||
// What a process death mid-recording leaves: the empty header from the start, the PCM
|
||||
// written since, and a torn half-sample at the end.
|
||||
file.writeBytes(BinaryEncoding.wavHeader(0, 16000, 1) + ByteArray(3201) { 1 })
|
||||
assertEquals(3200L, FaceclawWavRecorder.repairHeader(file.path, 16000, 1))
|
||||
val bytes = file.readBytes()
|
||||
assertEquals(44 + 3200, bytes.size)
|
||||
assertEquals(BinaryEncoding.wavHeader(3200, 16000, 1).toList(), bytes.copyOf(44).toList())
|
||||
assertEquals(-1L, FaceclawWavRecorder.repairHeader(file.path + ".missing", 16000, 1))
|
||||
} finally {
|
||||
file.delete()
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -90,6 +90,7 @@ class CaptionEngineCoreTest {
|
||||
}
|
||||
override fun onSpeechStart(startMs: Long) { events.add("start:$startMs"); startLatch.countDown() }
|
||||
override fun onStatus(status: String?) { events.add("status:$status") }
|
||||
override fun onStopped(error: String?) { events.add("stopped:$error") }
|
||||
}
|
||||
val host = object : CaptionHost {
|
||||
override val dispatcher = CallbackDispatcher { it() }
|
||||
@@ -123,5 +124,105 @@ class CaptionEngineCoreTest {
|
||||
assertEquals("status:Captions listening...", events[1])
|
||||
assertEquals("start:0", events[2])
|
||||
assertEquals("utterance:hi:2:0", events[3])
|
||||
assertEquals("stopped:null", events.last())
|
||||
}
|
||||
|
||||
/** Records listener events; `stopped` opens once the worker reports onStopped. */
|
||||
private class RecordingListener : FaceclawCaptionEngineListener {
|
||||
val events = mutableListOf<String>()
|
||||
val stopped = Latch(1, testPlatform())
|
||||
var utteranceLatch = Latch(1, testPlatform())
|
||||
override fun onUtterance(text: String?, embedding: FloatArray?, startMs: Long, endMs: Long, peakRms: Double) {
|
||||
events.add("utterance:$text")
|
||||
utteranceLatch.countDown()
|
||||
}
|
||||
override fun onSpeechStart(startMs: Long) {}
|
||||
override fun onStatus(status: String?) {}
|
||||
override fun onStopped(error: String?) {
|
||||
events.add("stopped:$error")
|
||||
stopped.countDown()
|
||||
}
|
||||
// Read only after a latch: callbacks run on the worker (the test dispatcher is inline).
|
||||
fun snapshot(): List<String> = events.toList()
|
||||
}
|
||||
|
||||
private fun captionHost(transcriber: () -> OfflineTranscriber?) = object : CaptionHost {
|
||||
override val dispatcher = CallbackDispatcher { it() }
|
||||
override fun loadTranscriber(modelDir: String): OfflineTranscriber? = transcriber()
|
||||
override fun loadEmbedder(modelPath: String): SpeakerEmbedder? = null
|
||||
}
|
||||
|
||||
private fun speak(session: CaptionSession, loudChunks: Int = 10, quietChunks: Int = 8) {
|
||||
val loud = AudioSegmentation.pcm16ToLittleEndian(chunk(3000), 0, 800)
|
||||
val quiet = AudioSegmentation.pcm16ToLittleEndian(chunk(0), 0, 800)
|
||||
val idle = AudioSegmentation.pcm16ToLittleEndian(chunk(10), 0, 800)
|
||||
repeat(2) { session.acceptPcm(idle) }
|
||||
repeat(loudChunks) { session.acceptPcm(loud) }
|
||||
repeat(quietChunks) { session.acceptPcm(quiet) }
|
||||
}
|
||||
|
||||
@Test
|
||||
fun captionSessionSurvivesAnUtteranceTheModelRejects() {
|
||||
var calls = 0
|
||||
val transcriber = object : OfflineTranscriber {
|
||||
override fun recognize(samples: FloatArray, count: Int): String {
|
||||
calls++
|
||||
if (calls == 1) throw IllegalStateException("bad input")
|
||||
return "second"
|
||||
}
|
||||
override fun release() {}
|
||||
}
|
||||
val listener = RecordingListener()
|
||||
val errors = mutableListOf<String>()
|
||||
val session = CaptionSession(captionHost { transcriber }, testPlatform(), "caption-reject-test") { m, _ -> errors.add(m) }
|
||||
session.setListener(listener)
|
||||
session.setAsrModelDir("models")
|
||||
session.setSilenceMs(300)
|
||||
session.start()
|
||||
speak(session)
|
||||
assertTrue(listener.utteranceLatch.await(5000))
|
||||
listener.utteranceLatch = Latch(1, testPlatform())
|
||||
speak(session)
|
||||
assertTrue(listener.utteranceLatch.await(5000))
|
||||
session.stop()
|
||||
// The rejected utterance still arrives (without text), and the worker kept going.
|
||||
assertEquals(listOf("utterance:", "utterance:second", "stopped:null"), listener.snapshot())
|
||||
assertEquals(listOf("caption recognition failed"), errors.toList())
|
||||
}
|
||||
|
||||
@Test
|
||||
fun requestStopDeliversTheTrailingUtteranceBeforeStopped() {
|
||||
val transcriber = object : OfflineTranscriber {
|
||||
override fun recognize(samples: FloatArray, count: Int): String = "last words"
|
||||
override fun release() {}
|
||||
}
|
||||
val listener = RecordingListener()
|
||||
val session = CaptionSession(captionHost { transcriber }, testPlatform(), "caption-flush-test") { m, e -> error("$m: $e") }
|
||||
session.setListener(listener)
|
||||
session.setAsrModelDir("models")
|
||||
session.start()
|
||||
// Speech with no closing silence: the utterance is still open when the stop lands.
|
||||
speak(session, loudChunks = 10, quietChunks = 0)
|
||||
session.requestStop()
|
||||
assertTrue(listener.stopped.await(5000))
|
||||
assertEquals(listOf("utterance:last words", "stopped:null"), listener.snapshot())
|
||||
}
|
||||
|
||||
@Test
|
||||
fun aModelLoadFailureReportsStoppedWithTheError() {
|
||||
val listener = RecordingListener()
|
||||
val session = CaptionSession(
|
||||
captionHost { throw IllegalStateException("model corrupt") },
|
||||
testPlatform(),
|
||||
"caption-load-test",
|
||||
) { _, _ -> }
|
||||
session.setListener(listener)
|
||||
session.setAsrModelDir("models")
|
||||
session.start()
|
||||
assertTrue(listener.stopped.await(5000))
|
||||
assertEquals(listOf("stopped:model corrupt"), listener.snapshot())
|
||||
// The dead worker left the session stopped, so a later stop is a no-op.
|
||||
session.stop()
|
||||
session.requestStop()
|
||||
}
|
||||
}
|
||||
Reference in new issue
Block a user