ChatGPT for packaging Cloudron Apps

[robi]

I just thought about asking it to generate a manifest file for Rust Desk and it said it wasn't an existing app.

Then I gave it the URL of the github repo and this is what it generated:

I apologize for the confusion, given that now I have access to the source code of the RustDesk-server, I can provide you with an example of a Cloudron manifest file for this application. Keep in mind that this is just an example and you will need to adjust the values to match the specific requirements of your setup.

{
    "id": "com.example.rustdesk",
    "version": "1.0.0",
    "name": "RustDesk",
    "title": "RustDesk",
    "description": "RustDesk is a open-source, self-hosted, desktop-as-a-service",
    "icon": "rustdesk-icon.png",
    "homepage": "https://rustdesk.com",
    "location": "https://rustdesk.com",
    "type": "webapp",
    "accessRestriction": "public",
    "accessRestrictionReason": "",
    "manifestVersion": 1,
    "dependencies": {
        "domains": [
            {
                "domain": "rustdesk.com"
            }
        ],
        "services": [
            {
                "type": "http"
            }
        ],
        "ports": [
            {
                "name": "http",
                "protocol": "tcp",
                "public": true,
                "port": 3000
            }
        ]
    },
    "config": {
        "env": {
            "DATABASE_URL": "postgres://username:password@host:port/database",
            "APP_SECRET": "secret_key"
        },
        "command": "cargo run --release",
        "appUser": "rustdesk",
        "appGroup": "rustdesk",
        "workingDir": "/app/rustdesk-server"
    }
}


This manifest file assumes that the application is listening on port 3000, that is using a postgres database, and that you have set the DATABASE_URL and APP_SECRET environment variable.
It also assumes that the application is built with rust and cargo and that the source code is located in a folder named "rustdesk-server".

Please keep in mind that you will need to adjust the values for the id, version, name, title, description, icon, homepage and location accordingly.
Also, you will need to adjust the dependencies, config and other fields to reflect the actual requirements of your application.

You can find more information about how to package an application for Cloudron in the official documentation: https://cloudron.io/documentation/apps/packaging/

[BrutalBirdie]

@robi ChatGPT is really awesome.

I really hope tools like this will disrupt the education system, since many of these systems are very broken.

Here! Learn these facts/data and then write an essay.
With ChatGPT these homework tasks or assignments are meaningless. As they where before!

Since it was only about learning facts, spitting them out and forget them again.

The school systems need to focus on learning how to learn again.
How to do proper research, how to validate information, how to use the tools of our time to generate valuable research.

I remember many of my school days simple reading the same page 20 times just to memorize information for an exam and then never using that information ever again.

I could rant to much more about the school system

Looking forward to the Anti-ChatGPT-AI to them be counter by an AntiAnti-ChatGPT-AI and so on.
Like its been for many years in the video game cheating industry.
An arms race but with AI, the AI on AI warfare

[LoudLemur]

@robi You brilliant man!

https://you.com also has an AI code generator. Would you like to try and reproduce the experiment over there and see how the results compare?

[subven]

Another usefull approach would be to fully automate testing new releases for existing Cloudron apps since this is monkey work most of the time. There are already pipelines that look for new releases, build them and run some tests but with AI you could also analyze changes and watch out for migrations + integrate and test them for your specific needs (or at least make some suggestion on what needs to be adjusted).

With a growing app catalogue, this could speed up 80% of the work for updates/upgrades I guess ^^

[timconsidine]

@robi Ha! That’s brilliant.
Does produced code actually work ?
(On small screen so can’t assess)

[robi]

@LoudLemur so far using their interface has been crappy for code as they have a limit of how many lines they will let it print before cutting it off so you cannot get a full output.

If you figure out how to get more than 30 lines of output or so, let me know.

[robi]

@subven yeah, if you first package a testing suite, then generate test scripts that will test each apps functionality and collect and display all that.

You have to have each step pipelined and you can get help with each step from chatGPT. Maybe even the overall pipeline if prompted well enough.

[robi]

@timconsidine from what I can tell the generated file is correct for a rust app with cargo calls, etc. We don't have many rust apps, so not sure if it found an example or extrapolated the json correctly.

[LoudLemur]

@robi It is not easy to find, and you may need to ask your sys admin to have it enabled. (In your case, probably reach out to You.) In the YouCode application, there is a settings button at the top right. Click on that to reveal parameter settings like MaxLines. If it isn't visible, try contacting You.

Finding that box can be tricky too. You might need to try a search and then scroll down to some of the results to find it.

https://you.com/code

In passing, there is an early access too

https://about.you.com/youcode-premium/

[robi]

@LoudLemur lol I am the sysadmin.

I didn't make an account there, but will request the premium access and see what happens.

[LoudLemur]

@robi Haha