Building a Mini CMS for vCard – Personal Portfolio with GitHub Copilot

This is a submission for the GitHub Copilot Challenge: New Beginnings

What I Built

I created a very basic Content Management System (almost-a-cms) for generating websites on GitHub using the vCard - Personal Portfolio template developed by codewithsadee. The main idea was to simplify the editing of static fields within the template.

To achieve this, I separated the site content into a series of JSON files—one for each section of the site—and paired them with a single HTML template. A Python script reads these JSON files and generates a new index.html file with a single click. This updated file can then be pushed to GitHub, allowing the site to be published seamlessly.

Two Ways to Modify the Content

1. Directly edit the JSON files: open each JSON file, modify them, and then run the Python script to generate or update index.html.
2. Use a small Flask backend: run a local server and edit the JSON content through a basic graphical interface. Once you save your changes, the new index.html file is automatically generated.

Demo

Directly edit the JSON files: Here I show how to modify the name directly from the json file, save the file, run index_html_generator.py, and show the update.

Use a small Flask backend: Here I show how to modify the job title directly from the very minimal backend, save the form, update the webpage, and show the update.

Repo

My GitHub Repo: Almost A CMS_

Copilot Experience

This was my first experience with GitHub Copilot, and it was quite positive. I’m used to having ChatGPT open in a separate tab for work, so I’m accustomed to LLM support. However, having AI integrated directly into VS Code saved me several steps and made my coding flow more streamlined—avoiding constant copy-pasting between a browser and the IDE.

I also appreciated the ability to switch between two different LLMs (OpenAI and Anthropic). However, I regret that only older model GPT-4o and GPT-3.5 Sonnet were available, as it would be nice to have more up-to-date or varied models. Still, for this time-constrained competition, I made maximum use of the in-editor Chat mode, code selection interactions, and the dedicated sidebar to speed up development.

The autocomplete suggestions were particularly precise—much better than the older suggestion systems—and felt more natural to use. Overall, I think GitHub Copilot is a solid product. Even as a free user, the experience was good. If in the future more model choices become available and OpenAI’s offerings align with the latest releases (mini or full versions), I would definitely consider subscribing to GitHub Copilot.

GitHub Models

I didn’t have the chance to fully explore GitHub Models as much as I wanted. I did a quick test and found some older OpenAI models. Seeing older versions dissuaded me from using GitHub Models in this competition context. However, I noticed that many other models (such as Microsoft’s Phi, as well as Llama and Mistral) were available. I plan to return later to test them out and see which ones are most useful. I hope these models will also be updated regularly with newer versions. It feels somewhat surprising that Phi-4 isn’t available yet, given GitHub is owned by Microsoft.

Conclusion

All in all, I’m pleased to have participated in this competition, if only to try out Copilot on a project I actually needed—updating my personal website. Who knows, I might further develop this repo even after the competition ends.