Please I’ll like to know if there’s anyone that’s been able to use machine learning in any way in their Softr app or if anyone has any idea how to use machine learning. The reason I’m asking this is, I’m building a resource app on Softr, and I’ll like to know if it’s possible for me to display content based on personalised recommendations similar to how Netflix displays content.
One direction you could go is to explore the Open AI integration that Softr offers. Depending on your specific use case, you might be able to generate personalized recommendations by creating a prompt for GPT-3 and then using the results. I haven’t dug into this yet, so I’m not sure what’s possible.
Another direction would be to use an Airtable extension or script to generate recommendations in a table that you could access from Softr pages. If you hunt around on this forum you’ll find posts that describe how to integrate with OpenAI on the Airtable side, and of course you might also be able to integrate with other AI tools or even just use some clever content searching approaches.
And a third way to go would be to explore using Zapier or Make to generate recommendations.
Hope this is helpful, feel free to follow up with more questions.
Addendum: I set up the OpenAI integration and gave it a spin. I didn’t realize until now that by default the only way to create a prompt is to insert an OpenAI text generation block, and then let the user type it into a text field.
So I think if I were going to try to generate personalized recommendations using this block, what I would do is put it on a page that had a list or list details block on it as well, and then write some custom code that iterated through the data in window.records to build up a prompt, filled in the input field with that prompt, and clicked the “Try it” button automatically. For example, if the list block contained a list of favorite album names and artists, I would generate a prompt like “My favorite albums are “ followed by the names and artists of the albums, followed by “What other albums should I listen to?” The results from the API query to OpenAI would then be displayed as a block of text with suggestions for other albums.
If I wanted to reformat those suggestions into links to Airtable records, what I would do is add “Please format the results both as plain text and as JSON with fields artist and album” to the end of the prompt, then when the prompt arrived, I would snip out the JSON, send it off to Make via a webhook, and use it to construct Airtable queries, send the results back as links, and append them to the block on the page.
If this sounds overly complex, well… you’re not wrong. I think I have convinced myself that, while this could probably work, one of the other two options might be simpler.