[googletts] Missing service configuration

[Artur-Fedjukevits]

Expected Behavior

Some CommunicationException should not reset the oAuthService variable.

Current Behavior

After some WARN like

2023-10-13 06:15:16.652 [WARN ] [ce.googletts.internal.GoogleCloudAPI] - Error initializing Google Cloud TTS service: An unexpected IOException occurred: java.util.concurrent.TimeoutException: Total timeout 5000 ms elapsed

or

2023-10-05 09:08:16.316 [WARN ] [ce.googletts.internal.GoogleCloudAPI] - Error initializing Google Cloud TTS service: An unexpected IOException occurred: java.util.concurrent.ExecutionException: java.lang.IllegalArgumentException: Buffering capacity 2097152 exceeded

the oAuthService variable is reset to null and Google TTS returns the Missing service configuration until openHAB is restarted.

Possible Solution

public byte[] synthesizeSpeech(String text, GoogleTTSVoice voice, String codec) {
        String format = getFormatForCodec(codec);
        try {
            return synthesizeSpeechByGoogle(text, voice, format);
        } catch (AuthenticationException e  ̶|̶ ̶C̶o̶m̶m̶u̶n̶i̶c̶a̶t̶i̶o̶n̶E̶x̶c̶e̶p̶t̶i̶o̶n̶ ̶e̶) {
            logger.warn("Error initializing Google Cloud TTS service: {}", e.getMessage());
            if (oAuthService != null) {
                oAuthFactory.ungetOAuthService(GoogleTTSService.SERVICE_PID);
                oAuthService = null;
            }
            voices.clear();
        } catch (CommunicationException e) {
            logger.warn("Error initializing Google Cloud TTS service: {}", e.getMessage());
            voices.clear();
        }
        return null;
    }

Probably like this, but I'm not sure.

Your Environment

Version used: openHAB 4.0.3
Environment name and version: OpenJDK Runtime Environment Zulu17
Operating System: Ubuntu 20.04.3

[openhab-bot]

This issue has been mentioned on openHAB Community. There might be relevant details there:

https://community.openhab.org/t/google-tts-missing-service-configuration/149498/13

[jlaur]

I checked #14937, but actually this seems to be a long-standing bug:

openhab-addons/bundles/org.openhab.voice.googletts/src/main/java/org/openhab/voice/googletts/internal/GoogleCloudAPI.java

Lines 331 to 338 in 2d92fda

	} catch (AuthenticationException | CommunicationException e) {
	logger.warn("Error initializing Google Cloud TTS service: {}", e.getMessage());
	if (oAuthService != null) {
	oAuthFactory.ungetOAuthService(GoogleTTSService.SERVICE_PID);
	oAuthService = null;
	}
	voices.clear();
	}

I would suspect that CommunicationException should be handled differently, i.e. only AuthenticationException should reset OAuth2. @lolodomo, do you agree?

EDIT: And only now do I read your suggestion, @Artur-Fedjukevits, which is exactly the same as my suggestion. So we are probably right. 😉

[jlaur]

@Artur-Fedjukevits - is anything else logged after that? E.g. "Could not synthesize text via Google Cloud TTS Service"?

[jlaur]

@Artur-Fedjukevits - can you test the fix?

• org.openhab.voice.googletts-4.0.4-SNAPSHOT.jar
• org.openhab.voice.googletts-4.1.0-SNAPSHOT.jar

[jlaur]

Some notes for future refactoring improvements:

openhab-addons/bundles/org.openhab.voice.googletts/src/main/java/org/openhab/voice/googletts/internal/GoogleCloudAPI.java

Lines 327 to 340 in 2d92fda

	public byte[] synthesizeSpeech(String text, GoogleTTSVoice voice, String codec) {
	String format = getFormatForCodec(codec);
	try {
	return synthesizeSpeechByGoogle(text, voice, format);
	} catch (AuthenticationException | CommunicationException e) {
	logger.warn("Error initializing Google Cloud TTS service: {}", e.getMessage());
	if (oAuthService != null) {
	oAuthFactory.ungetOAuthService(GoogleTTSService.SERVICE_PID);
	oAuthService = null;
	}
	voices.clear();
	}
	return null;
	}

This method is basically a wrapper for:

openhab-addons/bundles/org.openhab.voice.googletts/src/main/java/org/openhab/voice/googletts/internal/GoogleCloudAPI.java

Lines 353 to 384 in 2d92fda

	private byte[] synthesizeSpeechByGoogle(String text, GoogleTTSVoice voice, String audioFormat)
	throws AuthenticationException, CommunicationException {
	AudioConfig audioConfig = new AudioConfig(AudioEncoding.valueOf(audioFormat), config.pitch, config.speakingRate,
	config.volumeGainDb);
	SynthesisInput synthesisInput = new SynthesisInput(text);
	VoiceSelectionParams voiceSelectionParams = new VoiceSelectionParams(voice.getLocale().getLanguage(),
	voice.getLabel(), SsmlVoiceGender.valueOf(voice.getSsmlGender()));
	SynthesizeSpeechRequest request = new SynthesizeSpeechRequest(audioConfig, synthesisInput,
	voiceSelectionParams);
	HttpRequestBuilder builder = HttpRequestBuilder.postTo(SYTNHESIZE_SPEECH_URL)
	.withHeader(HttpHeader.AUTHORIZATION.name(), getAuthorizationHeader())
	.withContent(gson.toJson(request), MimeTypes.Type.APPLICATION_JSON.name());
	try {
	SynthesizeSpeechResponse synthesizeSpeechResponse = gson.fromJson(builder.getContentAsString(),
	SynthesizeSpeechResponse.class);
	if (synthesizeSpeechResponse == null) {
	return null;
	}
	byte[] encodedBytes = synthesizeSpeechResponse.getAudioContent().getBytes(StandardCharsets.UTF_8);
	return Base64.getDecoder().decode(encodedBytes);
	} catch (JsonSyntaxException e) {
	// do nothing
	} catch (IOException e) {
	throw new CommunicationException(String.format("An unexpected IOException occurred: %s", e.getMessage()));
	}
	return null;
	}

Outcomes:

• Line 365: AuthenticationException
• Line 375: Return null
• Line 381: Return null
• Line 383: CommunicationException

The error handling is inconsistent. The wrapper handles OAuth2 and logging, and converts exceptions into returning null. If already null because of JSON deserialization or missing response, nothing will be logged. In all cases we end throwing here:

openhab-addons/bundles/org.openhab.voice.googletts/src/main/java/org/openhab/voice/googletts/internal/GoogleTTSService.java

Lines 329 to 332 in 2d92fda

	byte[] audio = apiImpl.synthesizeSpeech(trimmedText, (GoogleTTSVoice) voice, requestedFormat.getCodec());
	if (audio == null) {
	throw new TTSException("Could not synthesize text via Google Cloud TTS Service");
	}

Suggestions:

• synthesizeSpeechByGoogle should check response before calling fromJson. If null or empty, don't call. In this case fromJson will never return null.
• synthesizeSpeechByGoogle should never return null. If response is missing or cannot be deserialized, throw CommunicationException (or custom exception). Chain JsonSyntaxException.
• The wrapper synthesizeSpeech should only handle OAuth2 and in this case rethrow AuthenticationException. Don't log.
• In GoogleTTSService catch exceptions and rethrow as TTSException. Log if needed, I don't know if throwing this exception will already be logged by the framework.