Bubble apps can do a lot, but previously one area Bubble couldn’t help out with was machine learning. But not anymore! We’re excited to announce a new Bubble integration built by our friends at Peltarion - the no-code artificial intelligence platform that helps you build your own deep learning models!
Here’s our Forum announcement to accompany the blog post!
We’re excited to announce a new partnership with Peltarion, a company providing a no-code platform for quickly building and using artificial intelligence / deep learning models! Just as Bubble aims to open up web app development to more users, Peltarion aims to make AI / DL more accessible to more users.
Peltarion lets you upload a data set (yours or one you have access to) and train a DL model with it - they even have pre-trained neural networks to give you a jump start. What’s extra cool is that after you have a model, you can call it via an API to generate a prediction about a new item. This also means that it integrates very well with your Bubble app!
Classic use cases for such an AI model would be image classification, natural language processing and predicting values for new items based on observed, past items. There is a free tier of the Peltarion Platform which you can use to build the model, and they have plenty of tutorials to illustrate its power!
This might be an ignorant question, but how helpful do you think Peltarion might be in sorting activity feed items (a la Facebook news feed) based on individual users’ past behavior? Is this a viable use case?
Nice addition!
If you need a simple recommendations system for your app using AI to recommend products or posts in a social network, you can try my new plug-in RAPI now with early bird pricing !
Also, with my plug-in you don’t have to pay to external providers
Hi, @allenyang
I am excited to use deeplearing in the bubble app.
By the way, I tried to build a test app with the tutorial, but I am facing a trouble.
6th step of 6/ Finishing the workflow, according to the documents we can select “Result Key”, there are only “key”, “value” and “errorMessage”
In case of selecting “key”, the errror says the key is not list.
So, I think the custom state should not be list type.
Finally i could build app without error, but the workflow doesn’t work properly with an error
Given you can run ML in your Browser, and link into AI platforms like Algorithmia - I am not sure sure that this is entirely true.
What does this give me that Algorithmia doesn’t ? Particularly as they have transparent pricing. I refuse to use anything that won’t tell you how much it costs. We all know that means it is a LOT.
It would be a huge shame if Bubble is heading down this particular path. Particularly given the lack of progress in areas that deeply need it. Editor performance is woeful at times and blamed on “the browser”. Server Side Plugins are still a mess. The Editor itself, compared to peers, is way behind. Still haven’t ensured consistency in workflows…etc
Come on Bubble, don’t chase “enterprise” referral fees when you have a lot of work to do elsewhere.
So let me get this straight: You can now do ML stuff with Bubble (via connection to an external service)… but you can’t iterate over an array in (vanilla) Bubble on-page workflows.
@allenyang@vivienne This is all extremely exciting in my opinion. I have some ideas on what I think I would like to do with this in with Bubble.
But if bubblers are like me and not quite sure about AI and ML and would need support along the way to understand and build into bubble, Where would you suggest getting support from along the way.
Hi @ts11 , that’s a cool use case but not supported on the Peltarion platform atm. Please feel free to visit our tutorials for inspiration for other use cases
Hi @asquare747 , would like like to elaborate a bit on what you’re building? I’d say it depends on your application and what you are working on. You could for instance train a model to recognize the face of you and your friends, feeding the model with a collection of images during training.
@KeitaroNakata thanks a lot for pointing out the error in our tutorial. It’s now updated, referring to “key” instead of “Result key”.
We’ve also published an updated version of the plugin where you no longer get the “$jscomp” error. Keep me updated on your progress building your deep learning-powered app! And let me know if you have any further questions.
@timgarrett111 Thanks for sharing your excitement! I’d be happy to help if you have any questions on training and deploying your AI model. The Peltarion tutorials is a great way to get started with training a model.
Hi! Sorry our pricing tiers and what you get with the freemium version were so hard to locate.
The Free version gives you: 10 training hours per month, 1 user, with 1 concurrent experiment.
Our Basic version is priced at €49 per month and gives you: 15 training hours per month, 3 users, and 3 concurrent experiments. This also gives you commercial use rights.
Our Pro version is priced at €499 per month and gives you: 50 training hours per month, 5 users, and 10 concurrent experiments, with 12 hours of support from our deep learning experts. This also gives you commercial use rights, as above.
@johnwilliams@ZubairLK thanks for the kind feedback, and I’m very excited to see what you are building! Don’t hesitate to reach out with any questions you might have