mirror/nanobot
1
0
Fork 0
mirror of https://github.com/HKUDS/nanobot.git synced 2026-08-04 08:28:36 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(webui): avoid false microphone silence errors

Browse Source
This commit is contained in:
chengyongru 2026-07-29 23:59:59 +08:00 committed by chengyongru
parent 0fe3c5aa2c
commit fc73d5ff39
2 changed files with 8 additions and 52 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

50
webui/src/hooks/useVoiceRecorder.ts
View File

@ -8,7 +8,6 @@ import {
const VOICE_RECORDING_MAX_MS = 120_000;
const VOICE_RECORDING_MIN_MS = 650;
const VOICE_NO_INPUT_HINT_MS = 1_100;
const VOICE_HOLD_START_MS = 140;
const VOICE_WAVEFORM_BAR_COUNT = 64;
const VOICE_WAVEFORM_SILENT_HEIGHT = 3;
@ -30,7 +29,6 @@ export type VoiceRecorderState = "idle" | "recording" | "transcribing";
export type VoiceRecorderErrorKey =
| "failed"
| "noDevice"
| "noInput"
| "notConfigured"
| "permission"
| "tooLong"
@ -61,7 +59,6 @@ export function useVoiceRecorder({
const audioRef = useRef<VoiceAudioState | null>(null);
const startedAtRef = useRef(0);
const maxTimerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
const inputHintTimerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
const holdTimerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
const holdActiveRef = useRef(false);
const startPendingRef = useRef(false);
@ -69,15 +66,10 @@ export function useVoiceRecorder({
const suppressClickRef = useRef(false);
const suppressClickTimerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
const shortcutActiveRef = useRef(false);
const levelObservedRef = useRef(false);
const peakLevelRef = useRef(0);
const levelReliableRef = useRef(false);
const noInputHintVisibleRef = useRef(false);
const [state, setState] = useState<VoiceRecorderState>("idle");
const [elapsedMs, setElapsedMs] = useState(0);
const [levels, setLevels] = useState<number[]>(VOICE_WAVEFORM_IDLE_LEVELS);
const clearInputHintTimer = useCallback(() => clearTimer(inputHintTimerRef), []);
const clearSuppressClickTimer = useCallback(() => clearTimer(suppressClickTimerRef), []);
const suppressNextClick = useCallback(() => {
@ -128,16 +120,6 @@ export function useVoiceRecorder({
}
current.analyser.getByteTimeDomainData(current.data);
const level = voiceLevelFromSamples(current.data);
levelReliableRef.current = true;
levelObservedRef.current = true;
peakLevelRef.current = Math.max(peakLevelRef.current, level);
if (level >= VOICE_MIN_LEVEL) {
clearInputHintTimer();
if (noInputHintVisibleRef.current) {
noInputHintVisibleRef.current = false;
onClearError();
}
}
setLevels((currentLevels) => [
...currentLevels.slice(1),
waveformHeightFromLevel(level),
@ -150,11 +132,10 @@ export function useVoiceRecorder({
} catch {
stopWaveform();
}
}, [clearInputHintTimer, onClearError, stopWaveform]);
}, [stopWaveform]);
const cleanupRecording = useCallback(() => {
clearTimer(holdTimerRef);
clearInputHintTimer();
clearTimer(maxTimerRef);
stopWaveform();
streamRef.current?.getTracks().forEach((track) => track.stop());
@ -162,8 +143,7 @@ export function useVoiceRecorder({
mediaRecorderRef.current = null;
startPendingRef.current = false;
shortcutActiveRef.current = false;
noInputHintVisibleRef.current = false;
}, [clearInputHintTimer, stopWaveform]);
}, [stopWaveform]);
const stopRecording = useCallback(() => {
const recorder = mediaRecorderRef.current;
@ -197,10 +177,6 @@ export function useVoiceRecorder({
streamRef.current = stream;
mediaRecorderRef.current = recorder;
startedAtRef.current = Date.now();
levelObservedRef.current = false;
peakLevelRef.current = 0;
levelReliableRef.current = false;
noInputHintVisibleRef.current = false;
setElapsedMs(0);
startWaveform(stream);
recorder.ondataavailable = (event) => {
@ -210,10 +186,6 @@ export function useVoiceRecorder({
const chunks = chunksRef.current.splice(0);
const durationMs = Math.max(0, Date.now() - startedAtRef.current);
const mimeType = recorder.mimeType || "audio/webm";
const hasMeasuredSilence =
levelReliableRef.current
&& levelObservedRef.current
&& peakLevelRef.current < VOICE_MIN_LEVEL;
cleanupRecording();
if (chunks.length === 0) {
setState("idle");
@ -224,11 +196,6 @@ export function useVoiceRecorder({
onError("tooShort");
return;
}
if (hasMeasuredSilence) {
setState("idle");
onError("noInput");
return;
}
setState("transcribing");
const blob = new Blob(chunks, { type: mimeType });
const audioPromise = wantsWav ? convertBlobToWav(blob) : blobToDataUrl(blob);
@ -242,19 +209,6 @@ export function useVoiceRecorder({
setState("recording");
onClearError();
maxTimerRef.current = setTimeout(stopRecording, VOICE_RECORDING_MAX_MS);
inputHintTimerRef.current = setTimeout(() => {
const recording = mediaRecorderRef.current?.state === "recording";
if (
!recording
|| !levelReliableRef.current
|| !levelObservedRef.current
|| peakLevelRef.current >= VOICE_MIN_LEVEL
) {
return;
}
noInputHintVisibleRef.current = true;
onError("noInput");
}, VOICE_NO_INPUT_HINT_MS);
} catch (error) {
cleanupRecording();
setState("idle");

10
webui/src/tests/thread-composer.test.tsx
View File

@ -825,10 +825,10 @@ describe("ThreadComposer", () => {
expect(onTranscribeAudio).not.toHaveBeenCalled();
});
it("warns during recording when microphone input is silent", async () => {
it("transcribes recorded audio even when waveform samples are silent", async () => {
mockVoiceRecorder();
mockVoiceAudioInput();
const onTranscribeAudio = vi.fn(async () => "should not appear");
const onTranscribeAudio = vi.fn(async () => "quiet voice");
render(
<ThreadComposer
onSend={vi.fn()}
@ -843,9 +843,11 @@ describe("ThreadComposer", () => {
await new Promise((resolve) => setTimeout(resolve, 1_150));
});
expect(screen.getByText("No microphone input detected.")).toBeInTheDocument();
expect(screen.queryByText("No microphone input detected.")).not.toBeInTheDocument();
fireEvent.click(await screen.findByRole("button", { name: "Stop recording" }));
expect(onTranscribeAudio).not.toHaveBeenCalled();
await waitFor(() => expect(onTranscribeAudio).toHaveBeenCalledTimes(1));
expect(screen.getByDisplayValue("quiet voice")).toBeInTheDocument();
});
it("does not treat unavailable microphone levels as silence", async () => {