Using Claude Code To Improve Eagle
In general, I haven’t been very outspoken about LLMs in this blog, or in general. For no specific reason other than not using them that much. I’m now revisiting the topic, as I’ve implemented a few features with the help of Claude Code in Eagle, the little program behind my website.
Some context about how my website works: it’s fully statically generated with Hugo, except for a few dynamic features powered by Eagle. Eagle is just a wrapper around the Hugo website. Most of the time it just serves static files. However, it does quite a few things on the side.
Some time ago, I wrote about my experience using LLMs for code development. At the time, I tried both ChatGPT on their website, as well as GitHub Copilot through VS Code. It did some nice things, but I wasn’t really blown away. So I went on using GitHub Copilot as a glorified autocomplete.
In the meanwhile, and after constantly seeing stuff online about it and how great it is, I got a month subscription for Claude Code because I wanted to try it out. Mind you: I have used Claude before, but only through its web interface. This is specific about Claude Code.
The tasks I gave Claude were things that have been on my ever growing TODO list for quite some time. They’re not very complex tasks, but they’re things that I don’t really get that much pleasure from doing, and they’re also not things that I really want. They’re good to haves. So… perfect for Claude. Here’s what I wanted:
* Refactor the session management in order to not use JWTs, and instead have a database-based sessions. A goal here was to be able to revoke sessions, and have a better overview of which sessions are active.
* Refactor the IndieAuth token management in a similar way to the sessions, by no longer using JWTs, storing the access tokens on the database, and be able to revoke them. In addition, also implement refresh token.
* Replace the usage of BoltDB with SQLite using Gorm. This has also been on my list for quite some time. The good thing here is that I did not need to migrate DB/preserve any data. So it was just replacing the implementation.
As I said, the tasks are not very complex, just mildly boring. After asking Claude to do one by one, and reviewing the code it made, I still had to make a few adjustments, or ask Claude to do them for me, whatever I deemed to be faster. In the end, everything worked quite well, and I’m a bit impressed.
Just like I said last time, I think it’s a great tool, if used well. If you know exactly what you need it to do, and give a direction on how it should be done, it seems to do its thing. I have a few other things I want to add to Eagle, and they’re a bit more complex than the tasks I asked Claude to do. So I’m very curious to see how Claude ends up performing. I’m a bit more positive than on my last article!
In the meanwhile, Anthropic was offering 6 months of Claude Max for open source maintainers. After applying and waiting a few days, I got a link to activate the subscription. I haven’t been able to activate it yet, because of my current subscription. I’m hoping the link doesn’t expire in the next few days!
Using Claude Code To Improve Eagle In general, I haven’t been very outspoken about LLMs in this blog, or in general. For no specific reason other than not using them that much. I’m now revisiti...
#Artificial #Intelligence #Machine #Learning #Tools #Meta
Origin | Interest | Match