Using AI to build a package

[eddowding]

As an experiment, I just used Claude to see if I could package an app. However, I have no idea if it's correct or not. Could someone please let me know how close this is?

https://github.com/getfider/fider/tree/v0.22.0

Claude:

I'll help you package Fider for Cloudron. Let me create the necessary configuration files for packaging Fider v0.22.0.Cloudron Package Configuration for Fider.

To package this for Cloudron:

1. Create a new directory and place these files in it:

mkdir fider-cloudron
cd fider-cloudron
# Place the files from the artifact here // code below & https://claude.site/artifacts/c9929cf9-5697-47e6-8fc4-b5ac74777e1b

2. Download the Fider logo:

wget https://raw.githubusercontent.com/getfider/fider/v0.22.0/public/images/logo-100x100.png -O logo.png

3. Build and package:

cloudron build
cloudron install

This configuration:

• Uses PostgreSQL for the database
• Configures SMTP for email sending
• Maintains persistent storage for uploads
• Handles automatic backups
• Provides health checking
• Sets up proper environment variables

https://claude.site/artifacts/c9929cf9-5697-47e6-8fc4-b5ac74777e1b

// CloudronManifest.json
{
  "id": "org.fider.cloudron",
  "title": "Fider",
  "author": "Fider Team",
  "description": "Open source feedback and feature request collector",
  "tagline": "Get feedback from your customers",
  "version": "0.22.0",
  "healthCheckPath": "/health",
  "httpPort": 3000,
  "addons": {
    "postgresql": {},
    "sendmail": {}
  },
  "manifestVersion": 2,
  "website": "https://fider.io",
  "contactEmail": "support@cloudron.io",
  "icon": "logo.png",
  "tags": ["feedback", "uservoice", "feature-requests"],
  "mediaLinks": [
    "https://raw.githubusercontent.com/getfider/fider/v0.22.0/docs/screenshots/1.png",
    "https://raw.githubusercontent.com/getfider/fider/v0.22.0/docs/screenshots/2.png"
  ],
  "minBoxVersion": "7.0.0",
  "forumUrl": "https://forum.cloudron.io/",
  "documentationUrl": "https://github.com/getfider/fider/tree/v0.22.0/docs",
  "postInstallMessage": "Fider has been successfully installed! Visit the app to complete initial setup."
}

// Dockerfile
FROM getfider/fider:0.22.0

# Add Cloudron-specific configurations
ENV NODE_ENV=production
ENV HOST_MODE=false
ENV SSL_CERT=false
ENV PORT=3000

# Cloudron will provide these through environment variables
ENV POSTGRES_URL=postgresql://${POSTGRESQL_USERNAME}:${POSTGRESQL_PASSWORD}@${POSTGRESQL_HOST}:${POSTGRESQL_PORT}/${POSTGRESQL_DATABASE}
ENV EMAIL_SMTP_HOST=${SMTP_SERVER}
ENV EMAIL_SMTP_PORT=${SMTP_PORT}
ENV EMAIL_SMTP_USERNAME=${SMTP_USERNAME}
ENV EMAIL_SMTP_PASSWORD=${SMTP_PASSWORD}

# Create data directory for uploads
RUN mkdir -p /app/data
VOLUME ["/app/data"]

# Setup healthcheck
HEALTHCHECK --interval=30s --timeout=30s --start-period=30s --retries=3 \
    CMD curl -f http://localhost:3000/health || exit 1

EXPOSE 3000

# start.sh
#!/bin/bash
set -e

# Wait for PostgreSQL to be ready
until nc -z -w5 ${POSTGRESQL_HOST} ${POSTGRESQL_PORT}; do
    echo "Waiting for PostgreSQL to be ready..."
    sleep 2
done

# Update database configuration
export JWT_SECRET=${CLOUDRON_APP_TOKEN}
export BASE_URL=${CLOUDRON_APP_ORIGIN}

# Start Fider
exec /app/fider

// README.md
# Fider for Cloudron

This package allows you to run Fider on Cloudron. Fider is an open source feedback and feature request collector.

## Installation

1. Install via the Cloudron App Store
2. Access your Fider instance at the configured domain
3. Complete the initial setup process

## Configuration

The app automatically configures:
- PostgreSQL database connection
- SMTP email settings
- Base URL and JWT settings

## Backup and Restore

Cloudron automatically handles backup and restore of:
- PostgreSQL database
- Uploaded files in /app/data

## Updates

Updates can be installed through the Cloudron App Store when new versions are available.

## Support

For issues related to:
- Fider: Visit https://github.com/getfider/fider/issues
- This Cloudron package: Post on Cloudron forum

[eddowding]

This is one-shot in Claude. I imagine a trained AI might do better.

[nebulon]

Nice try, seems like a very long way to go though. Especially the Dockerfile. It has valid syntax mostly, that is a good start, but otherwise nearly no single line makes any sense in the Cloudron context to be honest. Pretty much the same can be said about the start.sh. Last time I tried though chatgpt did perform much better, it just really struggled with the CloudronManifest.json

[eddowding]

Have you tried training them on example files?

[robi]

Here is an example session of how to better "talk to" the AI so it understands what you're doing.
https://skylarbpayne.com/posts/cursed-cursor

Having it analyze a similar app and start with the manifest, then have a template process to follow will get it closer to what you want. It's all about clearer communication on your part.