When I started creating content for the AWS Community Builder program, like many people I started with WordPress. However, it was too complex, had a lot of functionality I didn’t need, and more importantly would have needed a server running somewhere to host it.
All of that led me to looking at static web site generators fairly quickly - a great advantage of this is that you can host them in GitHub or something similar; or given I’m an AWS Community Builder, build it on AWS. And that’s what I described in my article Hosting a Static Website on AWS; just S3, CloudFront and not a server to be seen. I could even tie in a really simple pipeline that would regenerate the site as I commited my changes to a git repo (more info here).
But how do you know that this approach works - I could access the site and see the pages, but how could I see what others were doing and that it worked for them? I did setup Google Analytics, thanks to a Hugo integration, but it felt over the top, and it’s been niggling away for well over a year, that I wanted a different, more aligned solution.
Luckily, my original infrastructure as code (IaC) had configured CloudFront to generate logs, and these were being generated and stored in an S3 bucket, ready for me to come along and review. But manually trawling through log files was an intensive tasks, I’d need to download the files, read through them and work out what was going on. So I decided to create a script to try and parse the data.
What I didn’t expect when I decided to find a solution was that this setup would lead me to test a different kind of tool though; one that would both transform my workflow, let me see how my blog was being viewed, and also, surprisingly, reveal unexpected truths about what happens to a website on the public internet.
Kicking Kiro CLI’s tyres
The other driver for looking at this was that I’ve been using the Kiro IDE a lot recently, but hadn’t had a chance to try out the CLI as I’ve been curious about AI coding assistants that work directly in the terminal for a while; could they help me build something genuinely useful, or would they just generate boilerplate and abandon me when things got complex? To find out, I decided to test Kiro CLI with a practical problem: making sense of my CloudFront access logs.
The goal was simple: create a tool that could fetch, parse, and analyze the raw, gzipped log files from CloudFront that are sitting in S3. What happened next included a few surprises.
Pair programming with an AI
For someone used to working with an AI driven IDE, I found the CLI process surprisingly straightforward. Because I could authenticate myself with AWS credentials, I could access my account via the AWS CLI, try a few prompts and generate code that queried the live results that I could see.