WIP trash

100% of code by Claude free and GPT-Codex.

Somewhat useable but not really what it is fully intended to be yet.

Deterministic duplicate finding is... fine, I guess. Only completely and entirely useful for 100% similarity. Otherwise YMMV.

Haven't tested or made any prompts for LLM connection - either local or API - to allow for non-deterministic duplicate finding.

Auto-generates config .json on start, index .json on run.

Auto-generates duplicates and trash folders in main directory on run.

Everything added is at least functional from testing. It's not enough for the scope of the intent in my opinion.

To do:

Move away from tkinter.

Test and improve initial LLM support.

Fix graph preview for large workflows.

Consider further refactoring.

Idk, some other shit.