Show HN: I've made LLM generate code for my code generator

1 point

2 years ago

So, recently I've heard that the new Claude model has better code generation capabilities and is faster than the OpenAI model. I've tried to generate a Tetris-like game in the browser and it worked!

Initially, I used the Claude web interface and it has become quite cumbersome for me to copy-paste code from the browser to the local environment, so I've started looking for something that can bridge LLM to my computer. I've found some tools, like claude-engineer, but they seemed kinda unreliable or only CLI-based, so I decided to build my own version.

I didn't want to just build an arbitrary code editor, for quite a while already I've wanted something that can specifically write backend services. I have a tool for that called Mify. It can generate a backend structure with all the necessary glue, but the templates are pretty rigid, and you have to write OpenAPI schema to add handlers. But what if I could merge the AI generator and Mify?

Turns out, I can. Here's the repo: https://github.com/mify-io/mify-llm-editor. It allows you to generate backends based on templates and update code in them! Check it out, if you have any questions ping me in comments or via email [email protected].