AI App Development with Networked Storage Best Practices?!?!?!

Question for @Chris_Ching but also would love to hear from any others who have thoughts on the matter…

Since I’m a glutton for punishment and a challenge, the first app I’m planning to build is going to have a heavy networking component as it will be interactive and thus need to store and manage user data in the cloud. I’m planning to develop it using Cursor, following the guidance of the AI App Development YT videos Chris released in the last few months but as I work through editing and updating the build instructions that Claude generated, I’m realizing that many of the Build steps (understandably) include steps that have to be taken in Firebase (cloud storage solution I’m planning to use) which Cursor can’t control (unless it can??).

So my question essentially is, what is the recommended workflow when building an app with AI that requires steps to continuously be taken in an outside system?

One approach I can see is to make a copy of the build instructions with and without those steps and do them manually before running the corresponding build step that includes the Swift/SwiftUI steps that need to be taken so that specific table/column names can be inserted into the generated code based on what was configured in Firebase. It feels like this would best mimic what I understand is best practice when not using AI to code.

Another approach would be just running through all of the Swift/Swift UI steps and ending up with a bunch of placeholders and then going through and doing all of the Firebase buildout afterward and then filling in all of the placeholders. The part of me that hates context switching likes this approach but I imagine this could be messy and make it harder to find and remedy errors.

I imagine this sort of issue also similarly applies to any other 3rd party system thats a key piece of the app architecture so I’m hoping theres some guidance I can tap into as I approach this app development feat in this more modern way.

Thanks!

1 Like

Hey @brandonIsTall ,

Thanks for watching my videos!

This is a really interesting topic because in my AI coding exploration so far, I haven’t considered what it would be like to work with Firebase.

So there IS a way for Cursor to see and work with your Firebase database. MCP is the answer. It gives your AI the ability to go outside of the dev environment and use/access various tools. (My video on it → https://youtu.be/bC3mIQWHZMQ)

It looks like Firebase has an MCP server (Firebase MCP server  |  Develop with AI assistance). I haven’t explored it so I’m not sure what extent of control it gives Cursor.

I’ll have to look into this in the near future as I think using Firebase (or any remote database) opens up so much possibilities for what you can build.

In regards to your two approaches, I think this is an easy answer. You definitely want to take Approach 1 simply because it allows you to test that something it working before moving on.

With Approach 2, you won’t know if anything works until you complete the Firebase side and by that time, you’ve generated so much code that makes it harder to fix things.

It’s best to switch between generating code and configuring Firebase and then confirming that something is working before moving onto to the next step.

As you mentioned, this is how you would do it when not using AI to code.

Thanks again for watching and keep us posted!