We recently pushed a new version of our API's Message model which introduced a couple minor bugs. We're happy to announce we resolved all of them. Here's a list of what was fixed:
After a conversation capture session, we would not train the original question if we received negative feedback on a subsequent question. This sounds complicated because it was, but it's fully tested and working now.
We broke the self training feature in certain circumstances. That's now fully tested to be working again. We developed very comprehensive testing for this system to prevent us breaking it in the future.
Deleting tags were not always removing the orphaned tag ids from answers. That's fixed now.
Original questions from conversations were not appearing at the top of Handoff emails for a few days there. That's fixed.
We fxed a long time issue where the conversation log didn't reflect the actual back and forth between your chatters and the bot during capture sessions.
That's it 🎉
Give your customers 24/7 support
Ada can help you build and train a virtual agent, so your customers get answers wherever—and whenever—they are in the world