Skip to main content

While manychat do have the option to integrate ChatGPT but it does not follows up the older customer reply. I mean it treats every reply message as a new message and to resolve it you need to create a very long and messy flow in order to not miss any chat.

 

It is also not great in terms of assigning tags, or handing over other flow. I managed to solve it by introducing a custom instruction of manychat in bot field to reply a keyword if it thinks i should add a particular tag or assign it to new flow.

 

And then I use every reply and put up under set of conditions so that it can filter out some conversations for tags and new flows.

 

Areas of improvement:

  1. First fix this bot field text value, it just gives one small line for a long set of instructions, The size of the input box should have been extended based on the length of text. This can help us to edit the custom instructions right into manychat.
  1. Add an option where the chatbot takes the whole conversation not just a single reply. If it could take a whole conversation it could reply the follow up questions based on the whole conversation instead of replying one question and then handling it to the flow.

As of now, I can suggest of these two features only, If you guys have any other suggestions please do share.

Hi there,

Yes, adding memory would be awesome!

 

We have been able to add memory to ManyChat + OpenAI/ChatGPT using external requests and our own template. It takes a bit more work and variable management, but it works to perfection, and we can decide how much memory the bot has (usually the last 3 messages is enough in most cases, but we have bots that remember the last 12 interactions running)!

We also have the Assistants API integrated with ManyChat (through our custom template), which has memory built-in and lets you train the bot with your own information. This opens up a whole world of possibilities for larger clients with extensive knowledge bases.

Happy to discuss more if there’s interest.