This is a submission for the GitHub Copilot Challenge : New Beginnings
What I Built
I have built a minimalist note taking app that focuses on simplicity.
Demo
https://qwertyiscoding.github.io/Chalkboard/
Repo
https://github.com/QwertyIsCoding/Chalkboard
Copilot Experience
Copilot was quite handy throughout the entire development process. The integration with my repositories came in quite handy when referencing certain files when prompting and Copilot having access to the entire codebase helps with code context as well as continuity.
I first played around with the model switcher to see how different models respond to prompts, and I found out that Anthropic's Claude 3.5 Sonnet worked best in terms of accuracy -- for example a feature request. However, OpenAI's GPT 4o works better when it comes to more complex prompts (as those that can be seen in the 'Prompts' folder in the repo) and has better continuity between responses if the LLM can't provide the whole response in one go.
The edits and chat features built into VS Code were also quite useful, since I did not have to go to the GitHub site every time I wanted to make a quick prompt. Furthermore, I like how it implements the prompt into the file and highlights them while allowing me to test and debug further. It feels like prompting again and again on a prompt, and this technique in my experience makes the code more well polished instead of just entering large prompts in the first place. The autocomplete feature also felt very “at home” with the other features of Copilot. It made repetitive tasks easy (such as creating buttons, etc. in my HTML files).
GitHub Models
Yes, I used GitHub models for prototyping my app.
GPT 4o — Used for complex tasks such as large prompts and cross-referencing several files.
Perplexity — This was used to create the foundation for the app. This is where it all started, and its PRO mode was phenomenal in providing crisp and logical code.
Claude 3.5 Sonnet — This was mainly used for feature ideas -- such as to create a font switcher in the settings page which was already built with Perplexity and implemented with GPT 4o
Conclusion
I had fun working on this project and it was exciting to see how fast I could work with Copilot and other LLms. I created this app because I was quite tired of switching between note taking apps that were constantly getting bloated with features a note taking app does not need -- weather, music, chats and so many other "features" that pull the users away from their notes. And I wanted to fix that.
Credits
This project was build by Pranav Somalinga in one day.
Author Of article : Pranav Read full article
