v0.0.1

For the first time, we uploaded the mod to modrinth and curseforge. The xos branch PR for these updates that just got merged is found exactly here: xlateai/xos#127 and that should be merged to main soon. The commit hash that was uploaded as a gradle-in-build was this exactly this one.

Modrinth Download: https://modrinth.com/mod/code-mc
CurseForge Download: https://www.curseforge.com/minecraft/mc-mods/code-mc
Release Video: https://youtu.be/GAXOnJ4zLIU

embedded `xos` panel in chat! (paired with xos v0.3)

In xos v3 we added the ability to compile towards jni/Java such that we can embed an xos game engine window inside of our minecraft client inside the chat screen (pressing T ingame).

In the next iteration of this feature, I'll be adding dev apis for minecraft from xos so we can use import mc just like we can use import xos to control the game itself and read features about it from python programmed within!

Training Improvements & Groups + Colors

A pretty big set of updates across Python and Java. The idea is to improve training both algorithmically and tooling for training as well.

Python Changes

1. GPU forward and backward pass for rl operator agents.
2. wandb setup (with in-game printing)
3. change from fmc updates every 10 ticks (twice per second) to once per round (should slow down training but increase policy divergence/exploration)
4. stochastic policies instead of deterministic for better action diversity

Java Changes

1. agent names display heart count
2. spectral effect applied to top k agents
3. training mode free for all has groups so we can have multiple isolated groups of agents training within small clusters instead of one massive one
4. colored agents per group

Stable Free For All! (Super basic policies)

Super basic policies are being learned, but the infrastructure should be ready for the FFA gamemode. Just need to improve training for FMC and the networks!