1.3.6400
True Alpha-version
This version of EyeAuras contains dozens of different changes and tweaks in the very core of the system - image capture and processing pipeline.
This was all done to push as much performance as possible by removing things that are not needed at this point or proven to be not good in the long term. Tests are showing, that in most scenarios machine learning-based search is now about ~15% faster overall, with some extreme cases being twice as fast as they were before. But the main parameter which was improved is latency. So now the program reacts much-much faster on anything that happens on the screen - at each stage in the system (capturing the image, applying effects/resizing, preparing it for ML/image analysis and delivering analysis results (predictions and found bounding boxes) are now delivered to triggers much-much faster. In some cases latency has dropped from millisecond-range (i.e. 1+ millisecond) to less than 100 microseconds. This will be very noticeable in high-frequency scenarios, like trying to react on some object on the screen (e.g. item positions on the ground in Path of Exile or aimbot-like minigames where you have to track the object. This is not the latest iteration of improvements, I am planning at least two more large improvements (mainly in machine-learning areas), but they will be delivered later, probably even after the next large release to “Stable” channel at the end of Feb.
Scripts v3.0
Huge update to the scripting subsystem. This was all done as a part of work around Behaviour Trees - since the very beginning having fast and convenient way to prepare smaller bits of code embedded right into the tree itself was a major goal and the current implementation is way too heavy for that - it is powerful, but too verbose and heavyweight.
Right now it is early testing phase and in the coming weeks there will be a bunch of scripting related updates, but if you want to start coding - do not hesitate to try out the newer system and, if you have any questions, contact me in Discord and ask anything. Me and linqse will do our best on writing documentation about the new API with examples and everything.
Few notes:
- Nuget packages - they are already supported, but are disabled for now. I want to cleanup all potential internal issues first and after that allow people to use external dependencies
- C# script Action and Trigger are already using v3, but WebUI is still on v2. Reason is that WebUI has several features that still require porting them from v2 to v3 to match the expected functionality, e.g. exporting/importing code as C# project. This will be implemented in the nearest weeks and released as a part of separate update
Behaviour Trees
Still in development, at this point the core is mostly done and is already functional, but not stable enough. This will be main focus for the next two weeks. Before releasing them the plan is to:
a) Release all things which are now in “alpha” stage and are disabled by default:
- scripts v3
- file-based storage: storing EyeAuras config in an explorer-like way, 1 aura = 1 file, with built-in version control support
- file references: this allows to store and distribute even large binary files as part of auras, e.g. neural networks
b) do exactly 1 release to “stable” branch which will include all of the above, but will not include trees. Lifecycle of stable versions is approximately 2-2.5 months, so this will give just enough time to test out trees in alpha
Here is an example of how behaviour tree looks like in action for Lineage 2 bot - this is a combination of WebUI (window in right part of the screen) and Behaviour Tree for the logic. This bot picks target, attacks using combination of skills, heals/resummons pet if needed and does a few other things.
P.S. This was recorded not for public demo purposes, so there is some glitching going on with bot UI progress bars, the main thing there is the tree