Hi guys, we need help with apparent different behaviour between published web bot and internsl BO test chat. We have a chat history output which we are using as input to do heavy data extraction for input into a product recommendation platform. There are four different AI tasks in botpress studio simulators which take around 10 seconds to complete. These tasks are performing well in the internal chat test, but in the published Web Bot interface those tasks apparently are not getting executed. The bot seems to go back to the start of a new conversation. Also KB based bots seem to display same problem for us. They work from simulator but not from Web. Any pointers/ help will be appreciated. The overall aim is an e-commerce bot that uses a nlp chat to get product recommendations, and extract the attributes of the products recommended in NL and feed back to a product recommendation platform which presents the matching products.

Hey @mammoth-barista-45892, since the AI tasks take a while to finish, your bot might be getting caught up in the same bug that is causing duplicate messages for other people. We're hoping to get a fix out today!

I’m having a similar problem, AI Task is just not generating the same message as it is in the emulator. It fails to answer sometimes and other times it’ll answer in incomplete messages like in the picture attached, it only answers half the question and fails to actually recommend anything.

Cheers Gordy