Hi. I’ve been very happy with Make.com so far but ran into a seemingly unnecessary limitation today when trying to create a customGPT (OpenAI) with actions using Make.com webhooks.
In short the OpenAI action schema does not support parameters to make separate, unique, actions. I.e. I cannot create separate actions using [webhookURL]?action=action1, [webhookURL]?action=action2, and so on.
Instead the OpenAI schema supports separate actions via paths in the webhook URL. I,e [webhookURL]/action1, [webhookURL]/action2, and so on. Turns out Make.com webhooks does not support paths for the URL this surprised me and seems like a very simple implementation to do for the webhook?
Now I feel like Make.com is not the tool to use when developing customGPTs. I guess I have the option of creating separate scenarios with a unique webhook for each action I want to make but this does not feel very clean and will quickly get messy. Another option is to use POST requests I guess and pass the action parameter in the body but I’m not sure I can make this in to separate actions in ChatGPT.
Is this issue on the Make.com’s development team’s radar? Will we get path support in webhooks soon? Or does anyone in the community have another suggestion on how to solve this without creating multiple scenarios to cover each action I want to implement?
Welcome to the Make community!
That’s a great idea! I’m sure others would agree that this is a nice feature to have.
You can submit this suggestion to the Idea exchange, under App improvement ideas
Don’t forget to search for it first, just in case someone already suggested it, so that you don’t end up creating a duplicate.
Hope this helps! Let me know if there are any further questions or issues.
— @samliew
P.S.: Investing some effort into the Make Academy will save you lots of time and frustration using Make.
Just wanted to follow up on this one and correct my misconception that make.com is not a good tool for custom GPTs. While it is true that the only way to define actions in the openAI schema is to have different endpoints (paths in the URL) it is of course fully possible to use parameters for different capabilities of whatever automation you are making. The trick is you have to explain to the LLM how to use the parameters and hope it follows the instructions. Seems to work well for me so far anyway.
Tbh the real issues here is that custom GPT actions are still quite limited so I find myself having to do a lot of processing on the Make.com side of things. I started leveraging DumplingAI for some of the tasks and working as expected so far.
Wanted to leave a comment about this here in case someone else trying to implement actions for custom GPTs. There’s not that much great documentation about it.