The further the user progresses, the slower the interaction becomes
Hi everyone! I'm experiencing the following issue: the further the user progresses through the responses, the slower the interaction becomes. I'm not sure if this is due to the number of typebots being called during the interaction, the complexity of each typebot, or other factors. Is there a way to make the interaction more efficient as long as it is going on? Thank you!
12 Replies
What do you mean exactly by the slower the interaction becomes? It takes more time to answer? Can you record a video showing this?
Hi Baptiste, thank you for your kind response. I meant that the more the interaction goes on (because many questions are being asked) the longer it takes for the next question to appear. The user might have to wait even more than 20 seconds before the next question shows up.
Can you provide a reproduction? I can't reproduce that with my bots so maybe it is someting on your config? Maybe at least record a video to show what's in your flow? 🙂
So you sent a video recording by email. For now I can only guess since I don't know how your bot flow looks like.
I guess that you have an Agent Q&A loop and you feed the agent the new conversation history. So the more you provide history, the longer it will take the AI to answer. It is expected in that case.
Hi Baptiste, the agent is called at the end of the interaction, not during the interaction itself, and it does not process the history, it does something else (it is a "light" processing). So I do not think that's the problem...
Interested in seeing the flow then 🙂
Difficult to guess without knowing what's underneath
How can I show you the flow?
After each question-response, we update the value of a variable by adding a short string to it. Do you think this could be one of the causes?
You could record a screen recording presenting it, you could share it via a public link or export it into a file
Hi Baptiste, the flow implemented in our typebot is described in the atteched image.
It is rather simple:
1) The user goes through a sequence of questions based on his initial choice among a set of options.
2) Each option is associated to a given set of questions which are implemented in a seperate typebot. The typebot of an option contains both questions and conditions (these last are applied to the responses of the questions).
3) Some questions are common to more options, so we sometimes jump to a further typebot (Common) to manage common questions.
4) At the end of all questions an output is generated and the bot exits.
5) An AI agent is used to process the output in the very last block of the Orchestrator (it processes a very short text).
If we simulate the selection of one or, at most, 2 options, the flow is fine. If we select more than two options, at one point the interatction is slowing down.
The number of blocks in each typebot related to an option can be high (60-70 blocks).
Always when we are testing longer interactions, sometimes we encounter another problem: we receive the message "failed to fetch while sending message", "check your internet connection". If this happens, the bot stops, even if no implementation error is there.
Can you please help us to find out potential causes? 🙂
Please send a bot link (by DM if you'd prefer), I can only trust what's in Typebot, I can't guess anything! Thanks 🙂
Can you please tell me what DM means? 🙂
Direct message on Discord or by email, your call 🙂