TasteKit

Recommendations as a service

4 followers

Recommendations as a service

4 followers

Visit website

Launch tags:API•Developer Tools•Tech

Launch Team

AWS Activate CreditsCredits that fuel your startup’s next milestone

Promoted

PlanetScale Boost

@jsngr Epic idea. Love this and looks well executed, great docs. Are you planning to charge $? Also would love to know the stack behind it. How'd you build it? How long did it take from idea -> launch.

Report

10yr ago

Automator for Figma

Maker

@mscccc Thanks! Haven't thought about $ yet. Wanting to add more features and make it smarter before that happens. It's built on Ruby on Rails, uses Postgres and Redis. I became interested in the Jaccard index when I was playing around with it after learning about it in school. The base of the code can mostly be found here: https://github.com/jordansinger/... It was just a matter of making it so that apps could sign up and get their own API key. Like I mentioned in a comment earlier, the idea came to me to turn it into a service on the plane to Hawaii. Totaled about a days worth of work.

Report

10yr ago

PlanetScale Boost

@jsngr Awesome! We use the jaccard index for calculating similarity within Product Hunt as well :)

Report

10yr ago

Automator for Figma

Maker

@mscccc Very cool!

Report

10yr ago

Hey @jor_el3 : would this work inside a bot experience? Does it have send and receive API capabilities?

Report

10yr ago

Airtable To Figma

@rushabh_doshi @jor_el3 Any more info on this? ;)

Report

9yr ago

intriguing... does the jaccard similarity coefficent allow for weightings on likes / dislikes to influence recommendations results?

Report

10yr ago

Automator for Figma

Maker

@roamlocally It does indeed, that's pretty much how it works.

Report

10yr ago

Slightly confused about how to utilize the API for this. I understand the simplicity of like, dislike, and recommendation. However, there doesn't seem to be an API to populate the content which will be liked or disliked. Feels like it will be hard to do recommendations without that, no?

Report

10yr ago

Automator for Figma

Maker

@daniel_schwartz Items are added as things are liked and disliked, no need to dump your entire set of items in the first place. There's a disclaimer in the API docs stating why there is no create for users or items, it makes things simpler.

Report

10yr ago

@jsngr ah that makes sense (sorry didn't see the disclaimer at the bottom of the page). Definitely an interesting service, will likely give it a shot sometime this week to see what kind of recommendations we get out of our app/platform.

Report

10yr ago

Get After It

Great, how many days you took to launch TasteKit, for Botlytics you said it was done in 2 days. Kindly share some insights on how to launch products 'fast'. Thanks

Report

10yr ago

Automator for Figma

Maker

@tahaqadri Great question, Taha! I recently went on a trip to Hawaii, and the idea came to me on the plane. The flight was 6 hours long, and I finished the majority of the core work then. Just this past weekend I spent more time to focus on the front end and API design, totaling around 8 more hours. The reason I'm able to build things quickly is because I keep things simple and always have an MVP mentality. I determine what the absolute minimum is for the idea to launch, and iterate from there.

Report

10yr ago

Get After It

@jsngr Wow, that's cool, it reminds me of 'Limitless'. Thanks for the reply and explanation.

Report

10yr ago

@jsngr @tahaqadri That is exceptionally amazing! Would love to learn more about how you even gained those skills. Seem to have to be a veritable Full Stack dev genius or at least fairly seasoned before attempting that. Im an ideator practicing with Altucher's 'How to Become an Idea Machine' and would love to 'build my ideas' in a similar style. Feel free to check out some of my ideas on my Twitter and see what you can do with any of that. I will be checking out more of your stuff to figure out exactly how you do this haha.

Report

10yr ago

Automator for Figma

Maker

TasteKit is a recommendation engine API. I built it because there were some occasions where I needed to implement Tinder-style liking and disliking into apps, and the complexity behind intelligently recommending things can get hard. With TasteKit, you simply like and dislike things via the API, and are able to fetch recommendations. There's no need to dump all of your existing set of items. This is because TasteKit uses the Jaccard similarity coefficient. Recommendations are based entirely on what users like and dislike, we couldn't simply recommend you items immediately if you were to dump them all in the database. This is why items get added as users like and dislike things, and users with similar tastes are recommended things they haven't seen that have been liked prior. For example, say you're using a movie app that recommends you movies: you like Toy Story and Finding Nemo. Another user likes Toy Story, Finding Nemo, and Finding Dory. Since you have similar tastes, Finding Dory will be recommended to you.

Report

10yr ago

@jsngr would the recommendations work if I only used the likes API and not the dislikes API? I'm thinking very simple ecommerce recommendations - if a user clicks on an item = like, show recommendations based on what other users clicked on.

Report

10yr ago

Automator for Figma

Maker

@morrjobs Yes indeed. Dislikes act as negative weight, so if you have only likes, then the recommendations will still work!

Report

10yr ago

I did a very small npm module implementing this API if you guys are interested in testing this product : https://www.npmjs.com/package/ta... Have fun !

Report

10yr ago

1 2 3

Reviews