Skip to main content
Solved

How to Maintain Conversation Context When Integrating ChatGPT with WhatsApp in ManyChat?

  • March 27, 2025
  • 1 reply
  • 82 views

John Jairo Parra

Hi, community! I’m integrating ChatGPT with WhatsApp using ManyChat, and I want to ensure that the bot maintains the conversation thread with each user.

My goal is for ChatGPT to remember previous messages within the same chat, so responses stay coherent and natural.

🔹 How can I achieve this within ManyChat?
🔹 Do I need an external storage system for chat memory, or is there a native way to do it?
🔹 What strategies do you recommend to make the user experience smoother and more natural?

I’d appreciate any suggestions or experiences you can share.

Best answer by Gustavo Boregio

@John Jairo Parra 

There are different ways to do it – some simpler than others.

In OpenAI’s ‘Completion’ endpoint, the only way to include memory is by sending all previous interactions (all questions and answers). To do this using the native integration, you need to manage this history yourself, and keep concatenating all questions and answers in the Prompt you send so the AI has context. It's quite a bit of work!

If you do it using the API (connecting Manychat directly to OpenAI), you have more control and options. You’ll still need to manage the memory, but you have flexibility in how you do it. We have a course on how to do this and also a ready-to-install template (you can literally get it working in 5 minutes). That template already handles all the memory—you just need to set your Prompt.

If you want to level up in complexity, you can use an OpenAI Assistant. Assistants have built-in memory (managed by OpenAI), but the integration is much more complex. We also have a ready-to-install template for that.

Take a look at the content we have at https://superarmeonline.com (it’s in Spanish though!)– you’ll find both the courses and the templates there.

And if you have any questions, I’m here to help!

1 reply

Gustavo Boregio
Forum|alt.badge.img+6
  • Manychat Community Moderator
  • Answer
  • March 27, 2025

@John Jairo Parra 

There are different ways to do it – some simpler than others.

In OpenAI’s ‘Completion’ endpoint, the only way to include memory is by sending all previous interactions (all questions and answers). To do this using the native integration, you need to manage this history yourself, and keep concatenating all questions and answers in the Prompt you send so the AI has context. It's quite a bit of work!

If you do it using the API (connecting Manychat directly to OpenAI), you have more control and options. You’ll still need to manage the memory, but you have flexibility in how you do it. We have a course on how to do this and also a ready-to-install template (you can literally get it working in 5 minutes). That template already handles all the memory—you just need to set your Prompt.

If you want to level up in complexity, you can use an OpenAI Assistant. Assistants have built-in memory (managed by OpenAI), but the integration is much more complex. We also have a ready-to-install template for that.

Take a look at the content we have at https://superarmeonline.com (it’s in Spanish though!)– you’ll find both the courses and the templates there.

And if you have any questions, I’m here to help!