DEV Community

Amy Liu
Amy Liu

Posted on

What's the best stack for building web based AI applications?

OpenAI, chatGPT, DALL-E have excellent APIs but what are the best tools for building SaaS products based on them and offering services to end users? Products with a free tier are preferred.

Top comments (18)

steve_val profile image

By far the best stack for getting started quickly today is:

All these tools have a free tier and will scale nicely to 1m+ users if you get good traction.

awalias profile image

Example of using this exact stack to search and summarize technical documentation in a web app

steve_val profile image

A nice bonus with Supabase is you can also use the file storage system to keep a library of any AI generated images/large files etc.

supabase_io profile image

We have some docs for the pgvector extension.

pgvector README also contains some great usage guiadence.

saltcod profile image
Terry Sutton

A great, in-depth guide to that exact stack ^^

amyliumaiyi profile image
Amy Liu

Thanks for sharing. I like the fact that supabase is just postgres, and I'm already well acquainted with next.js for frontend, will give this a shot.

sweetpapa profile image
Forrester Terry

I've been using the following for a while:

  • Typescript (usually)
  • Quasar Framework (A Vue.js framework that bundles Electron, Capacitor, and various components all together)
  • Firebase Free Tier (Firestore NoSQL database, Firebase Functions (need to enable billing, but there is a free tier, Realtime Database, Firebase Storage)

Reasons I like this tech stack:

Typescript: |
Simply put, for larger projects or projects I intend to drop and come back to later - Typescript's rules and features helps make scalable projects. Having types and interfaces alone help me both document and plan out my apps. However, Typescript can certainly take some time to get used to, and IMO is not an important place to start if you are just learning programming. Certainly, JavaScript will work well.

Quasar / Vue.js | |

I like Vue.js because it was really easy for me to learn and teach to my team. The framework has been solid, many folks use it in production, and I have enjoyed Vue 3.

Quasar allows you to easily build cross platform mobile, web or desktop apps using any combination of Vue with Electron (Win, macOS, Linux), Capacitor (iOS/Android) or just Vue.js by itself for a SPA or SSR app.

Quasar is nice because you can build an app once, and then use the same code base for different platforms. It also streamlines the various processes involved with these different frameworks into a single workflow.

Firebase by Google |

Firebase leverages GCP, and has been very reliable for me. The Firebase SDK and integrations for JavaScript have been easy enough to work with -- especially if you are newer to backend dev (lots of documentation and examples to work off of).

Handling authentication, database management with Firestore, storage, hosting and even cloud functions are pretty easy to work with via the Firebase SDKs, CLI and backend GUI. Firebase also offers some nice ways of handling authorization to these features.

What using this stack looks like:

I will often build a frontend using VueJS, which is then either deployed as a site, or if I am building something else, it is paired with Electron or Capacitor.

For websites, it is great to pair Firebase functions directly with Firebase hosting, because I can easily and securely call my functions directly from my website without having to pass or expose any credentials. The API routes to Firestore are locked to the web hosting site, so things are nice and secure.

For mobile apps or desktop apps, I can setup access to Cloud Functions via Oauth like workflows if I like, or many other ways.

I have used this at my workplace to build at least 4 or 5 production apps which see between 5000-50,000 daily active users depending on the app.

If you ask me why I did not get into React or other stacks, it is simply preference and based on what I thought would be a good combination of:

  • Easiest to teach my team
  • Easiest to hire for
  • Safe to use for big/important projects

React and others can certainly fit this bill, but I liked Vue the best.

amyliumaiyi profile image
Amy Liu

Thanks for sharing! I'm a huge fan of Vue.js, the community is very large and helpful. I haven't tried Quasar yet, looks interesting.

rorstro profile image


matijasos profile image
Matija Sosic

If you want to build a React + Node.js + GPT API SaaS as quickly as possible, go with! Zero setup and you get free full-stack auth, async jobs and CRUD working out of the box.

Here's a GPT-powered app made with Wasp that recently got a lot of attention:

You can see the full source code and adapt it to your needs here:

vincanger profile image

thanks for the shout, @Matija

If anybody has any questions regarding my stack. let me know!

supa_community profile image
Supabase Community

Any questions on this stack we're always here to help:

amyliumaiyi profile image
Amy Liu

Thanks! That's really useful.

accreditly profile image

Lots of great recommendations here, but in reality it boils down to your exact requirements and experience.

Great SaaS products can be built in most languages. A good developer understands design patterns for building applications and those can be applied to any language or framework.

There are OpenAI libraries available for the following languages/frameworks:

  • Python
  • JS/Node
  • C#/.Net
  • Crystal
  • Go
  • Java
  • Kotlin
  • PHP
  • R
  • Ruby
  • Scala
  • Swift
  • Unity
  • Unreal Engine

Ultimately it comes down to what you're most comfortable developing in.

thorwebdev profile image
Thor 雷神

There's a bunch of templates on the vercel marketplace also:

iamaliahad profile image
Ali Ahad

FastApi and svelte!

lotfijb profile image
Lotfi Jebali

If you are thinking about a project and have an idea, we can collaborate on it and make it bigger ;)

ews3rd profile image

Insights are accurate. Time to delivering a finished product is now. As a mainframe programmer, I challenged ChatGPT to give up its secrets to error free code generation, and found the keys to error free generation of COBOL, JCL, SQL, sample test data and delivery of validated test results and other mainframe artifacts. GPT LLM has an unbelievable ability to generate simple to complex code using limited language prompts coupled with template instruction examples. I'm looking for the proactive client who wants to see how quickly a properly designed project can be coded and tested in a few sit-down sessions with the BA and designers.