Core Companion 2.2

This release makes Core Companion work on a fresh machine with nothing pre-installed, and takes the guesswork out of choosing a local AI model.

✨ Highlights

First-run setup that just works. Install the mod, join a world, and it downloads and prepares the AI for you — no terminal, no manual model commands.
Automatic, hardware-aware model selection. It checks the memory your machine can actually give an AI model — dedicated GPU VRAM, Apple-Silicon unified memory, or system RAM — reserves a few GB for Minecraft itself so nothing stutters, and picks the best-fitting local model (lightweight, balanced, or largest). It tells you which one it chose for your specs.
A setup that looks intentional. No filesystem paths dumped into chat — just a clean progress bar above your hotbar, and the look/start-mode questions only appear after the model finishes downloading.
She never spawns unable to talk. Your companion only arrives once her model is downloaded and warmed up, so her first reply is instant instead of a long cold-load stall.
Helpful when the AI backend is missing. If Ollama isn't installed or running, you get a clear in-game "install Ollama" guide instead of a cryptic error — and following, mining, and building all keep working without it.

Fixed first-join chat failures (ClosedChannelException) caused by the AI client being rebuilt mid-setup.
The download progress bar now stays on screen and updates smoothly across the whole multi-GB download, instead of flashing once and vanishing.
Dropped connections during a model download now auto-resume and retry instead of failing the whole setup.
Default companion name is now a neutral "Companion" until you name her — and renaming her no longer leaves a stale name baked into her personality.
Better server-side diagnostics when a model download or warm-up fails.