Hey so we all probably use AI models in some way or another to build stuff. But they usually struggle with Ash. One particular problem I’m always running into is that they try to create these complex multi-line calls from Phoenix to Ash, instead of simply running the tidy, elegant function provided by the code interface.
So I made a little script exposing a really simple API to gather information about the building blocks of your app.
I find that, once I made an agent aware of the existence of `ash-explorer` and let them run it, it immediately understands the scope of the application, and the right way of consuming the Ash infrastructure.
Here’s the link: GitHub - M-Gonzalo/ash-explorer
No guarantees, PRs welcome, the whole enchilada. Enjoy!
