Hello everyone. I am currently building a script to export and parse our web messaging transcripts for an external audit. I am using the /api/v2/conversations/messages/{conversationId}/messages endpoint, but I am finding the JSON structure of the transcript to be very complex, especially when it comes to ‘Rich Media’ (like images or buttons) and ‘Bot-to-Agent’ handoffs. Is there a simplified schema or a library that can flatten the transcript into a more readable format for our legal team?
I have built several ‘Transcript Flatteners’ using AWS Lambda. The transcript JSON is indeed very verbose because it includes every single event (like ‘Typing’ or ‘Read Receipts’) alongside the actual messages. You should filter the type field to only include text and structured messages. For the bot-to-agent handoff, look for the originatingEntity field to identify which part of the conversation was handled by the AI vs a live human. It is the only way to make the output legible for non-technical auditors!
I deal with these messy transcripts during our quality reviews. To follow up on Nav90, if you are presenting this to a legal team, you should also include the ‘Timestamp’ for every message in a standardized format. The Genesys API returns UTC, but your auditors will probably want to see the customer’s local time. We had to write a small utility to convert the timestamps based on the customer’s area code to satisfy our compliance department!
I am also a Python SDK contributor and I have been working on a pull request for the Genesys GitHub to improve the transcript parsing. Gre82, you should look at the ‘Conversation History’ API instead of the raw messages endpoint. It provides a more consolidated view of the interaction and handles the threading of multi-session conversations much better. It is still JSON, but it is a lot easier to map to a flat CSV or PDF for your audit reports!