Custom Action, database and AI

Here is one for the gurus and the geniuses that I m not ( it’s a great idea though lol)

I want to create a custom action, or a button, that will fetch a user’s specific collection information, and then use Openai to summarize that information and write it in a text box …

Who can help me do this?
Thanks in advance

Hey @Senuoy,

You can definitely do this with the ChatGPT response component in the marketplace.

I can make a tutorial on how it works later.

1 Like

Yeah this would be quite easy. Have a button with custom action that sends current record data to OpenAI API.

2 Likes

If you still need help, I can easily assist you on this

1 Like

I’ve followed your tutorial and successfully accessed user collection data, which I can easily have ChatGPT summarize. However, I’m encountering difficulties when trying to access the Feedbacks data from the input screen. Despite spending hours attempting to resolve the issue, I’ve been unable to get ChatGPT to work with the Feedbacks data. It appears that the main obstacle is the custom action requiring a reference to an input field, which can accept information from a component but is unable to read lists through magic text.

I’ve tried adding the feedback information using a custom list and included a feedback component for the specific user. However, neither the custom action nor magic text seems to be able to extract data from components within lists, which appears to be a limitation. After two hours of attempting to resolve this issue, I eventually gave up.

Although I will look into components available in the marketplace, I prefer building things myself to better understand their functionality and facilitate troubleshooting when necessary. I’m not a fan of simply plugging in components. Any suggestions on how to resolve this issue would be greatly appreciated. Thank you all for your assistance.

Summary, Input field is using magic text to read user info, hire date, rating, etc, so in tern summarizing it using any prompt I use. however, Feedback collection data is another data, so not showing when try to using magic text, and obviously there is a lot of entries for that collection per user. so I created a list to show all feedbacks, magic text or custom action cant read lists. so I created a custom list with components, again, cant read components inside lists…chatgpt tried to help as well, suggested some temp collections, zappier, etc, complicated…

Not sure if the video I just did is any help: ChatGPT API with Xano and Adalo No-code app development - YouTube

Do you have any screenshots of your setup with the feedback?

1 Like

I m going to go through your video now, sharing the screenshot in the meantime. Thank you,



the feedback components on the right is what I cant reference into the magic texts so AI can analyze it and give me the summary I want from the prompt. each employee has his own multiple feedbacks, so it needs to read all that data. Chatgpt suggested I create temp collection relation to store all that into one text field,but the database will grow like crazy too fast and I m not sure it will work…

I think you need to use something like Xano or Make.com to get the data and call the ChatGPT API, and this should get around the issue of not being able to access the list data.

1 Like

How would you do it using make? i ve setup the database call in make, but do a bolean on adalo to call in its own data? from the custom action?

You could get Make to call Adalo database, and then filter the returned fields, and send the data to ChatGPT. It might be a little cumbersome calling the records from Adalo though if it’s a large dataset as the collections API only gives 100 records at a time. Not decided on what the best way to do it. It’s a potential future video also.

1 Like

What do you suggest in this case? thanks

I thought of an idea that might work, but fried all my brain cells for the day so hope you can help: What if I created a database user collection field, where I concatenate the feedbacks progressively, and then I call upon that field since I can easily reference any user field in the prompt field? so the question is how do I copy all past feedbacks and concatenate them into one text field??

Thanks

so it works, but I need to know how to concatenate the feedbacks into one field progressively without deleting the previous one.

I ve been meaning to ask you this @theadaloguy since I saw your tutorial on youtube. How would you recreate the animation that Chatgpt has, when writing text, instead of text just appearing at once, gives it more cool…thanks

for an animation you can have a loading lottie with visiblity based on a hidden an input field.

On your button for the generation, first action should be to change input value to ‘on’.
The loading lottie is sometimes visible if input field = on.
Then after the custom action on the button, you have an action to change input value of the input field to ‘off’.

This means the lottie should display during the duration it takes for the generation to return a result.

1 Like

I guess you’d concatenate them each time they get generated.

So you have a user > feedback field which is just a bulk of text of concatenated previous feedbacks.

should be like this:

1 Like

This was a success, now I’m curious, say a user goes in to delete that database entry, comment, or whatever, how do you automate the deletion of that specific entry from the concatenated database? Thank you everyone

Within adalo im not sure you can, might need to call an API and use a replace function with the original text and search text (what to delete) then return the full string

1 Like