Tools for AI powered coding?

As I never managed to learn coding, AI is a godsend to me. I could already create some useful tools for every day’s use; one of these I published at Tryton Community. There is more to come. Till now, I could only do scripts with console outputs, not sure if I’ll one day be able to proceed to actually do modules.

As AI tends to speculate and takes a lot of assumptions, seemingly some come from Odoo structure, I found it necessary to exactly describe in which database tables and columns data are located. This description is the time-consuming part, so I thought that a general map of relations would save a lot of time. And maybe there is a better way of representation this info than graphics.

Does somebody have a good idea how such a piece of information could be achieved more easy than drawing a map by hand?

Greetings,
EG

AI needs exact context, but this does not always mean that the user must manually describe the whole domain or draw all relations by hand.

A good workflow could be:

  1. First let the AI scan the local Tryton folder and generate a technical map of models, fields, relations, views, actions and dependencies.
  2. Then ask it to work on a very specific task.
  3. Require it to explain which files/classes it used before changing code.

So the most important thing is not only “AI coding”, but “AI coding with repository context”.

This description is stored in ir.model and ir.model.field models of the database. You can export them using an script to and md file so the AI can read it directly or even just allow the AI to connect to a local instance of the database.