twilio-labs / flex-hackathon-template Goto Github PK

Some issues @Kunalmighty found while testing the IVR:

Within the IVR's gather widgets we were somehow hitting the timeout with no input (maybe speech wasn't recognized?) and the call was dropped.

The variable to read back the order number was not being set for some reason.

Suggestion: we might significantly simplify the IVR to just do Intro -> Transfer to agent. Folks can customize as needed.

I've seen this stack trace pop up sporadically when making an inbound call. It seems that in some cases the sync stream is not being created. Haven't yet determined the cause.

Line throwing the error: https://github.com/Kunalmighty/SKO24Flex/blob/main/apps/transcriptionEngine/src/index.js#L70

Error writing to Sync stream TypeError: Cannot read properties of null (reading 'publishMessage')
at /Users/cweems/workspace/SKO24Flex/apps/transcriptionEngine/src/index.js:70:28
at speechRecognizer.recognizing (/Users/cweems/workspace/SKO24Flex/apps/transcriptionEngine/src/recogniser/microsoft.js:33:7)
at SpeechServiceRecognizer. (/Users/cweems/workspace/SKO24Flex/apps/transcriptionEngine/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/common.speech/SpeechServiceRecognizer.js:90:59)
at step (/Users/cweems/workspace/SKO24Flex/apps/transcriptionEngine/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/common.speech/SpeechServiceRecognizer.js:48:23)
at Object.next (/Users/cweems/workspace/SKO24Flex/apps/transcriptionEngine/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/common.speech/SpeechServiceRecognizer.js:29:53)
at /Users/cweems/workspace/SKO24Flex/apps/transcriptionEngine/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/common.speech/SpeechServiceRecognizer.js:23:71
at new Promise ()
at __awaiter (/Users/cweems/workspace/SKO24Flex/apps/transcriptionEngine/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/common.speech/SpeechServiceRecognizer.js:19:12)
at SpeechServiceRecognizer.processTypeSpecificMessages (/Users/cweems/workspace/SKO24Flex/apps/transcriptionEngine/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/common.speech/SpeechServiceRecognizer.js:66:16)
at SpeechServiceRecognizer. (/Users/cweems/workspace/SKO24Flex/apps/transcriptionEngine/node_modules/microsoft-cognitiveservices-speech-sdk/distrib/lib/src/common.speech/ServiceRecognizerBase.js:646:56)

@cfavila @Kunalmighty Hi guys! I made some progress on getting this set up but am still a bit stuck. Jotting down where I left off for tomorrow:

1. It seems like there's a fetch that needs to be uncommented and pointed to the serverless service (probably on localhost 3000): https://github.com/Kunalmighty/SKO24Flex/blob/main/src/components/ConversationSummary.js#L21-L36. We should document in the README because it fails silently.
2. Having uncommented that fetch, it doesn't seem to be supplying the body parameters (transcript and language) that the ai-suggestion.js function requires?
3. Without a transcript array, the transcript.forEach call fails here: https://github.com/Kunalmighty/SKO24Flex/blob/main/apps/serverless/functions/realtime-voice-suggestions/ai-suggestion.js#L34

Let me know if I am off-base here - perhaps the wrong function is being called?