UNetLab v3 is built on Django along with various other frameworks. I am not a professional developer, nor do I consider myself one, yet I’m often told that my prototypes are extremely useful, especially because I can build them in very short timeframes.

However, since software development is not my main job, my coding activities are irregular, and long breaks between projects often cause me to lose familiarity with the syntax. In the past, Google was my primary tool: I constantly searched for code snippets and constructs just to recall the correct syntax. Using official documentation, Googling for suggestions, and searching for examples to solve specific problems was a very time-consuming process.

With the advent of “artificial intelligence”, I experimented with different approaches until I found a more efficient workflow that I am now using successfully.

ChatGPT for Code Generation

A couple of years ago, I started using ChatGPT to directly generate the code I needed. My prompts were along the lines of:

Create a Python function using Netmiko to connect to a Cisco device, elevate privileges, retrieve the MAC address table, parse it with TextFSM, and return a dictionary where the keys are interface names and the values are lists of MAC addresses.

At the time, the results were almost always subpar: the generated code often included fictional functions and generally didn’t work.

Adapting the generated code took longer than writing it from scratch.

Moreover, for my needs, describing what I want in detail takes significant time, converting my ideas into descriptive text always leads to incomplete or oversimplified instructions.

Copilot in Visual Studio Code

About a year ago, I tried a second approach: following a colleague’s suggestion, I enabled GitHub Copilot in Visual Studio Code and started using it as a coding assistant. The experience was insightful because it made me reflect on my own workflow.

Continue reading the post on Patreon .