speech error handling & tweaks
parent
f0b22eff64
commit
ce63613bdf
|
@ -81,6 +81,7 @@ export default function MessageInput(props: MessageInputProps) {
|
||||||
const temperature = useAppSelector(selectTemperature);
|
const temperature = useAppSelector(selectTemperature);
|
||||||
const message = useAppSelector(selectMessage);
|
const message = useAppSelector(selectMessage);
|
||||||
const [recording, setRecording] = useState(false);
|
const [recording, setRecording] = useState(false);
|
||||||
|
const [transcribing, setTranscribing] = useState(false);
|
||||||
const hasVerticalSpace = useMediaQuery('(min-height: 1000px)');
|
const hasVerticalSpace = useMediaQuery('(min-height: 1000px)');
|
||||||
const recorder = useMemo(() => new MicRecorder({ bitRate: 128 }), []);
|
const recorder = useMemo(() => new MicRecorder({ bitRate: 128 }), []);
|
||||||
const useOpenAIWhisper = useAppSelector(selectUseOpenAIWhisper);
|
const useOpenAIWhisper = useAppSelector(selectUseOpenAIWhisper);
|
||||||
|
@ -104,21 +105,42 @@ export default function MessageInput(props: MessageInputProps) {
|
||||||
}
|
}
|
||||||
}, [context, message, dispatch]);
|
}, [context, message, dispatch]);
|
||||||
|
|
||||||
const onSpeechStart = useCallback(() => {
|
const onSpeechError = useCallback((e: any) => {
|
||||||
|
console.error('speech recognition error', e);
|
||||||
|
|
||||||
|
try {
|
||||||
|
speechRecognition.stop();
|
||||||
|
} catch (e) {
|
||||||
|
}
|
||||||
|
|
||||||
|
try {
|
||||||
|
recorder.stop();
|
||||||
|
} catch (e) { }
|
||||||
|
|
||||||
|
setRecording(false);
|
||||||
|
setTranscribing(false);
|
||||||
|
}, [recorder]);
|
||||||
|
|
||||||
|
const onSpeechStart = useCallback(() => {
|
||||||
|
try {
|
||||||
if (!recording) {
|
if (!recording) {
|
||||||
setRecording(true);
|
setRecording(true);
|
||||||
|
|
||||||
// if we are using whisper, the we will just record with the browser and send the api when done
|
// if we are using whisper, the we will just record with the browser and send the api when done
|
||||||
if (useOpenAIWhisper) {
|
if (useOpenAIWhisper) {
|
||||||
recorder.start().catch((e: any) => console.error(e));
|
recorder.start().catch(onSpeechError);
|
||||||
} else {
|
} else {
|
||||||
|
const initialMessage = message;
|
||||||
|
|
||||||
speechRecognition.continuous = true;
|
speechRecognition.continuous = true;
|
||||||
speechRecognition.interimResults = true;
|
speechRecognition.interimResults = true;
|
||||||
|
|
||||||
speechRecognition.onresult = (event) => {
|
speechRecognition.onresult = (event) => {
|
||||||
const transcript = event.results[event.results.length - 1][0].transcript;
|
let transcript = '';
|
||||||
dispatch(setMessage(transcript));
|
for (let i = 0; i < event.results.length; ++i) {
|
||||||
|
transcript += event.results[i][0].transcript;
|
||||||
|
}
|
||||||
|
dispatch(setMessage(initialMessage + ' ' + transcript));
|
||||||
};
|
};
|
||||||
|
|
||||||
speechRecognition.start();
|
speechRecognition.start();
|
||||||
|
@ -126,10 +148,10 @@ export default function MessageInput(props: MessageInputProps) {
|
||||||
} else {
|
} else {
|
||||||
setRecording(false);
|
setRecording(false);
|
||||||
if (useOpenAIWhisper) {
|
if (useOpenAIWhisper) {
|
||||||
|
setTranscribing(true);
|
||||||
const mp3 = recorder.stop().getMp3();
|
const mp3 = recorder.stop().getMp3();
|
||||||
|
|
||||||
mp3.then(async ([buffer, blob]) => {
|
mp3.then(async ([buffer, blob]) => {
|
||||||
|
|
||||||
const file = new File(buffer, 'chat.mp3', {
|
const file = new File(buffer, 'chat.mp3', {
|
||||||
type: blob.type,
|
type: blob.type,
|
||||||
lastModified: Date.now()
|
lastModified: Date.now()
|
||||||
|
@ -153,17 +175,21 @@ export default function MessageInput(props: MessageInputProps) {
|
||||||
const json = await response.json()
|
const json = await response.json()
|
||||||
|
|
||||||
if (json.text) {
|
if (json.text) {
|
||||||
dispatch(setMessage(json.text));
|
dispatch(setMessage(message + ' ' + json.text));
|
||||||
|
setTranscribing(false);
|
||||||
}
|
}
|
||||||
} catch (e) {
|
} catch (e) {
|
||||||
console.log(e)
|
onSpeechError(e);
|
||||||
}
|
}
|
||||||
|
|
||||||
}).catch((e: any) => console.error(e));
|
}).catch(onSpeechError);
|
||||||
} else {
|
} else {
|
||||||
speechRecognition.stop();
|
speechRecognition.stop();
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
} catch (e) {
|
||||||
|
onSpeechError(e);
|
||||||
|
}
|
||||||
}, [recording, message, dispatch]);
|
}, [recording, message, dispatch]);
|
||||||
|
|
||||||
|
|
||||||
|
@ -197,7 +223,8 @@ export default function MessageInput(props: MessageInputProps) {
|
||||||
<>
|
<>
|
||||||
<ActionIcon size="xl"
|
<ActionIcon size="xl"
|
||||||
onClick={onSpeechStart}>
|
onClick={onSpeechStart}>
|
||||||
<i className="fa fa-microphone" style={{ fontSize: '90%', color: recording ? 'red' : 'inherit' }} />
|
{transcribing && <Loader size="xs" />}
|
||||||
|
{!transcribing && <i className="fa fa-microphone" style={{ fontSize: '90%', color: recording ? 'red' : 'inherit' }} />}
|
||||||
</ActionIcon>
|
</ActionIcon>
|
||||||
<ActionIcon size="xl"
|
<ActionIcon size="xl"
|
||||||
onClick={onSubmit}>
|
onClick={onSubmit}>
|
||||||
|
@ -207,7 +234,7 @@ export default function MessageInput(props: MessageInputProps) {
|
||||||
)}
|
)}
|
||||||
</div>
|
</div>
|
||||||
);
|
);
|
||||||
}, [recording, onSubmit, props.disabled, context.generating]);
|
}, [recording, transcribing, onSubmit, props.disabled, context.generating]);
|
||||||
|
|
||||||
const disabled = context.generating;
|
const disabled = context.generating;
|
||||||
|
|
||||||
|
|
Loading…
Reference in New Issue