- Added custom SquidEntity support for lookcontrols (fixing squid issues with following and looking at player)
- Refactored look controls to be simpler and easier to modify

Refactor controls classes into new package, and add custom speed settings for Illagers and WanderingTraders

Watch for entities loading and unloading, and adjust chatdata (i.e. remove dead chat data). Also, when loading a world, init friendship targeting for any > 0 friendeship scores (so friends wont attack players)

Adding new mixin to dynamically affect how the canTarget method works on LivingEntities. This prevents them from targeting any Player users if the friendship score > 0. Also changed behavior of Follow to pause when they are near the player, allowing other goals to run.

Adding custom entity speed adjustments for a variety of Enties, that follow too slow or too fast. Most work fine with a speed of 1.0 though.

Adding accesswidener for SlimeMoveControl access. New helper code to make Slime's look at the player (in survival mode) and correctly follow the player.

Improved Attack AI to run towards the player, slow down, leap and attack. The player can now sprint to outrun the attacking entity.

Fixed entity heights for EnderDragon (lowered chat bubble) and fixed tilt calculation to use the padding above the entity height.

 when sending chunked lite JSON data to new player on server (i.e. someone reconnects or logins in).

Update player login message from server to be sent in chunks for 32000 characters, to avoid the max String length errors. The client will reassemble all chunks and then load the JSON.

system-chat prompt engineering, to prevent output of ...response..., and give a more clear output syntax example.

Added a new FleePlayerGoal, if an entity feel scared or threatened by the player. They run away at 1.5x speed. Also updated the system-chat examples to provide a more sequential list of messages.

- Added moveConflictingGoals function (to prevent duplicate goal priorities)
- Fixed FollowPlayerGoal to not clear the this.targetPlayer variable when stop() is called. This can be temporarily stopped due to a higher priority goal.
- Updated TALK priority to 2, and FOLLOW priority to 3, for maximum compatability across entities, without overriding their most important survival/attack goals.

Fixing issue with message order in Chat GPT request, to add them in reverse order (limiting them by token usages), and then reversing the list at the end.

- Added a long list of random "no response" phrases. Sometimes the LLM just outputs a behavior and no text.
- Added the actual output with behaviors to chat history (for better context)
- Fixed a bug which was broadcasting the new generated message at the wrong time.