Using Ai with Wappler - Admin Panel by DB Scaffolding

Hi Wapplers,

I just downloaded WindSurf by Codeium, looking for possible integration with Wappler and how it can accelerate the development process.

I believe one of the tedious things in building a website is the (Admin Panel), so I decided to give this tool a shot!

I searched for an SQLite sample database, and I found Chinook by Kaggle

Then, I created a new NodeJS project in Wappler, selected chinook.db for storage, and created a test page called "the_db.ejs"

Next, open the project folder in WindSurf IDE (Download - 2 weeks free trial)

To explore the database I asked WindSurf

Analyze the database chinook.db, and in /views/the_db.ejs file, make a hierarchy visualization for all tables and their fields, with modern and professional UI

here is what I got!

"show all relationships between the tables?"

now make a visual diagram for all tables and the relationships between them

After understanding the database, I started building the admin panel

in the admin_panel.ejs, create a side-bar containing links for all of the tables, then generate the related CRUD pages of each table

check routs.json look for missing routes, and add the new pages to be accessible

The Artists Management page "/admin/artists" does not show any data from the database

..and the flow goes on!

I know many things are needed to get a professional Admin Panel, however I believe using AI with Wappler is promising and could make the users' lives easier, and database scaffolding is one example.

On the other hand, I encourage everyone to try WindSurf, I think it raises the bar in coding experience

Look forward to seeing your trails not only with WindSurf but also with other tools that could make our work smoother and faster!

Regards,
Samy

3 Likes

While I think that AI has a future into web development, I also think that it is too early to settle on one platform to be used.

Here is a list of open-source alternatives:

3 Likes

We have tried both ChatGPT and Claude using their Pro Subscriptions and have found that both repeatedly generate errors and ignore prompting until this situation comes to a grinding halt as you run out of credit having to redo the same thing over and over and over again (which both LLMs recognise as their fault, yet you do not receive any rebate for their errors, they just continue to consume). Claude is definitely the superior LLM to ChatGPT though (for code), by well and by far! Still in saying that if your code exceeds 300 or so lines expect them both to throw a wobbly and fail to generate the entire source! They'll both make you consume more usage by repeating the output in full, and then fail to render/generate it all once again, looping again and again. With careful prompting you can lessen the likelihood of it occurring... Well sometimes anyway, other times it will just go off on one inadvertently... Result is once again you'll have to wait a couple of hours to be able to prompt again...

This brings me on to any Wappler based A.I integration... A free (Open Source) LLM would be so much more reliable and avoid this failure. Which one though? Maybe we need to consider getting a few Community members to try out various free LLM's to ascertain the best result? The flip-side is using one of the paid LLMs and just have your usage consumed within half an hour or less if your application has any complexity to it whatsoever...

:thinking:

4 Likes

Quoting myself, excuse me for that, but an example of how to save on response length...

I don't think we have to worry about Skynet anytime soon.....

2 Likes