How makcuai/eventuri-ai-makcu works
Project type and purpose
• This is a Windows desktop AI-powered aimbot and detection GUI.
• Key purpose: capture game frames, run YOLO model inference (v8–v12 in .pt, .onnx, .engine formats) on GPU (CUDA or DirectML), display results in a CustomTkinter GUI, and drive mouse movements or a USB serial device for aim assistance.
File structure
• Top‐level: setup scripts (.bat), README, source folder src/.
• src/ contains:
• main.py, Eventuri-AI.py: application entry points (GUI bootstrap and control loop).
• capture.py: screen/frame capture functionality.
• detection.py: orchestrates model loading and inference.
• models/: directory of multiple YOLO models (Torch, ONNX, TensorRT engines).
• mouse.py, windmouse_smooth.py, patch.py: modules for mouse or hardware-driven aim movement.
• config.py: profile management, saving/loading user settings.
• setup_cuda.py, setup_directml.py: preparation of runtime environment.
• gui_*.py: GUI layout, callbacks, constants, and sections.
README for additional insights
• Confirms dual-mode GPU support (CUDA 12.6 or DirectML).
• Emphasizes a polished modern UI built on CustomTkinter.
• Profiles, device support (USB serial VID:PID), class selection per game, multi-mode aimbot.
Components and interactions
a. Main components
• GUI Layer (CustomTkinter)
• Module
• Inference Engine
• Model Repository (ONNX, .pt, TensorRT .engine)
• Device/Mouse Control Module
• Configuration/Profile Manager
• Environment Setup Subsystem (CUDA or DirectML)
• External Dependencies: GPU drivers, USB serial devices
b. Relationships and interactions
1. GUI Layer triggers Capture → passes frames to Inference Engine.
2. Inference Engine loads selected model from Model Repository (ONNX/PyTorch/TensorRT) using appropriate runtime (DirectML or CUDA).
3. Detection results feed into Mouse Control Module (either OS mouse events via mouse.py/windmouse_smooth.py or serial packets via patch.py to external device).
4. Configuration Manager provides GUI parameters (model path, classes, sensitivity, aimbot mode).
5. Setup scripts (install_setup_cuda.bat, install_setup_directml.bat) prepare environment before runtime.
c. Architectural patterns/design principles
• Modular separation: UI vs. core logic vs. hardware interface.
• Plug-and-play model loader supporting multiple backends.
• Profile/config abstraction for persistency.
• Batch or live inference loop pattern (capture → preprocess → infer → postprocess → act).
d. Key technologies/frameworks
• Python 3.x, CustomTkinter for UI
• PyTorch, ONNX Runtime, TensorRT
• DirectML/CUDA 12.6
• PySerial or equivalent for USB serial communication