We have a weird use case. We have an inbound Architect Email flow where field technicians email .wav voice memos to a specific address. The Architect flow routes these emails to a queue. I want the Genesys Cloud Speech Analytics engine to transcribe these .wav attachments so my topic detection rules can process them, just like it does for inbound phone calls. However, the system completely ignores the audio attachments on the emails. Is there an Architect action or a specific setting to force the Speech and Text Analytics (STA) engine to process media files attached to digital interactions?
Hello! I do not think this is natively supported. The STA engine is designed to ingest the media streams directly from the Edge telephony servers or the WebRTC media microservices. It is not designed to parse the attachments array of a digital interaction and extract binary files for transcription. The digital analytics pipeline expects plain text payloads.
The point above is correct. You cannot do this natively within the Architect Email flow. You will have to build a custom solution. In your Email flow, you could use a Data Action to send the attachmentId and a secure download URL to an external AWS Lambda function. That Lambda function would download the .wav file, use Amazon Transcribe (or a similar third-party service) to generate the text, and then use another Genesys Cloud API call to append that text as an Internal Note on the conversation. It is a heavy workaround.
Hi. I just want to add a warning to 's suggestion! If you do the Lambda workaround and paste the transcript as an Internal Note, the Genesys Cloud topic detection still will not analyze it! The STA engine only analyzes the actual customer/agent messages in a digital transcript, it completely ignores Internal Notes. If you want the native topic detection to work, your external Lambda function must inject the transcript back into the conversation using the Guest Chat API to make it look like a message from the customer!