How To Connect ManyChat to ChatGPT (Ultimate Tutorial)

[@chatimize]

Get started with my FREE template: chatimize.com/manychat-chatgpt-template/

[@elkinshikhaliyev4154]

This is the first and the best video about connecting an AI assistant through ManyChat!!! Thank you so much! I did everything in the video and everything works smoothly and without any issues!

[@chatimize]

Thank you! I'm glad it helped you!

[@elkinshikhaliyev4154]

Thanks a lot. But, there is a problem with gtp-4o-mini assistant. Through API it gives worth answers than directly on openAI page. Why? 😢

[@chatimize]

I am not sure... I also addressed this in my video. I think the good news is that it gives good answers in the API (and not the other way around)

[@swarupsarkar937]

Great one ❤❤❤solved all of my issue

[@chatimize]

I'm glad it helped you!

[@mocacoca1]

Great Video, Is there any advantage or differences to Connecting Manychat to Chatgpt directly vs the previous video using Cody?

[@chatimize]

Yes, ChatGPT is cheaper than Cody. But I prefer Cody, because it's more user friendly and ready for live applications.

[@albertomoreno1982]

Amazing video! But I have a problem or rather a question, the chat does not remember the context of the conversation, when it recognizes the id, does it recognize the id of the previous message or the id of the entire conversation? Because when I test it, it repeats the same things again

[@chatimize]

Hey Alberto, thank you! Yes, it should remember the entire conversation, that’s what the Thread ID is for.

[@albertomoreno1982]

@ Thank you very much for your reply! I will make changes to the Open AI assintant, which may be what is failing

[@chatimize]

Awesome!

[@smuhv7534]

Hey Joren, Just a quick question for the file search on openai assistant, does it consume alot of tokens? or is it stored on vector databases?

[@chatimize]

It depends on how big the query is and how many files you have. My recommendation = Just test and you know it instantly.

[@Motorkout]

Can manychat reply to comments on facebook using CGPT?

[@chatimize]

Unfortunately not… it is possible with Uchat (chatimize.com/uchat/). This is another tool with great features, but somewhat less beginner friendly.

[@TheGrilledCheese]

Great tutorial and was definitely interested but wanted to know, I'm feeling like the ai on manychat with chatgpt is already is pretty solid. When I add info into it's Ai and fill out it's task it seems to answer pretty intelligently. I say that to ask why do you think I should do it this way and what are the real benifits?

[@chatimize]

I think you are referring to the AI step in this case. I think the problem with that feature is that it isn't flexible. If someone exactly follows the steps in the script, it works. But when someone wants to do something else (that isn't in the script), it isn't flexible enough. And that's why I prefer to use ChatGPT via this integration.

[@TheGrilledCheese]

@ after using it I can see how you would want that. You can only upload a certain amount of info per ai chat; when I need it to know a more complex list of services and projects. I did connect two ai chats together so when you ask a certain question it will switch to other chat that has more info about a certain subject.

[@chatimize]

Yes, exactly. My recommendation in general: Just test whether it works for you. If it works, it works. If it doesn't, use this integration.

[@rodrigochung3797]

Can this chatbot you created answer multiple user questions? In other words, a customer asks 2 questions in sequence and your automation responds broken down to these answers?

[@chatimize]

To be honest: I am not sure how you would handle that. Because then you need to have some logic that waits on messages until X period. And then send all the messages in one go. I think it's doable with using a Datastore in Make and saving the messages there, but it will be a much more difficult integration. How many times do you encounter this?

[@magiskedaniel]

Do you recommend using this or Cody?

[@chatimize]

I recommend using Cody. OpenAI Assistants is still in beta, so I don't recommend it now. And it can also be a bit slow at times.

[@WTPleo]

Can we make that WhatsApp bot initiate a conversation with a lead instead of what you showed in the video, which is someone needing to initiate the conversation with the WhatsApp bot?

[@chatimize]

Yes, that is possible. Here is a video on sending WhatsApp broadcasts: kzbin.info/www/bejne/mYDdZ2ehj5lkeKssi=KFejNKidgMASSxty You can also it similarly with Rules or Sequences.

[@CmtePereira]

Another excellent tutorial, do you have any tips to avoid the error "[400] Can't add messages to thread_xxxxxxx while a run run_xxxxxx is active." in OpenAI in make when the user sends multiple DM messages in sequence?

[@chatimize]

Thank you! To be honest: I am not sure how you would handle that. Because then you need to have some logic that waits on messages until X period. And then send all the messages in one go. I think it's doable with using a Datastore in Make and saving the messages there, but it will be a much more difficult integration. How many times do you encounter this error?

[@CmtePereira]

@@chatimize When the user sends, for example, 3 messages separately in a short sequence, the manychat trigger apparently sends all of them and the moment the first hits OpenAI and there is an active run it triggers error 400 informing that there is already a run running for another attempt, I tried to work around it with delay but without success, it would be perfect if Manychats could separate the message queue from the trigger before send to make webhook or any tipe of store before set on openAI.

[@chatimize]

Yes, if you want to do this, you should add these messages to a Datastore in Make. And then you need logic in Make to handle when you want to send this to OpenAI to answer.

[@elbenerdrums5203]

Hi there! Nice video, I'm just facing one issue, I'm having 2 answers at the beginign of the conversation, how can I solve this? This justs happens when it's a new "costumer"

[@chatimize]

Thanks for bringing this up. The template had a mistake. I have updated it, so you can download the new one or update your existing one. If you want to update your existing one, do the following: 1. Go to your Make Scenario 2. At the router, click on the second route (where it says "2nd" 3. Set the option "Set the route as a fallback" to "yes". Doing this ensures that the 2nd route is only being used when the first route is not used. Please let me know if it worked!

[@elbenerdrums5203]

@@chatimize Damn! It wordek like a charm, Thank you so much!

[@chatimize]

@@elbenerdrums5203 Awesome, have fun!

[@RodrigoCoelhoRj]

Trigger Make option doesnt appear for me at my Many Chat account. Any idea gow to solv it?

[@chatimize]

Yes! Connect the first node in the Make scenario to your Manychat account. Once you’ve done that, it will appear in Manychat!

[@himarossi]

How to avoid, the operation to be run twice? When testing when the first message and the user doesnt have Thread ID, it runs first the top part. But then instantly it recognises the response that the user got as a question and sends a second message. So always when user has no thread ID he receives 2 messages

[@chatimize]

I can help you with this! Make sure the second route of the Router is set to fallback. So now, just click on where it says “2nd”, and the a pop-up opens where it says “Run as fallback”. You want to check that. Once you do that, it will work!

[@himarossi]

@@chatimize Ahhh i see, i complicated it way more by creating a 2nd condition, saying that if thread ID not equal to 0 not to run it, seems to work now

[@chatimize]

Awesome!