🎯 NVIDIA Technology Integration Guide

This document outlines the specific NVIDIA hardware and software technologies utilized by the Mission Control Gaming Assistant.

🚀 NVIDIA NIM (Inference Microservices)

The assistant leverages NVIDIA NIM to provide world-class reasoning without local VRAM penalties.

Hybrid Inference: Intelligent switching detects connectivity. When offline, the system bypasses cloud calls to prevent latency spikes and falls back to local reasoning.
Strategic Reasoning: Uses meta/llama-3.1-8b-instruct for fast long-term strategy.
Tactical Analysis: Uses meta/llama-3.1-8b-instruct for real-time combat advice.
Vision Enrichment: Uses meta/llama-3.2-11b-vision-instruct for robust scene understanding.
Voice Engine: Hybrid system utilizing Google Cloud TTS and ElevenLabs for high-fidelity responses, ensuring reliability beyond deprecated NIM audio endpoints.

The vision pipeline is powered exclusively by TensorRT 10.x.

Pure TRT Mode: Unlike standard AI apps, Mission Control does not load the PyTorch runtime during gameplay. This saves ~1,000MB of VRAM.
Engine Optimization: YOLOv8 models are compiled into FP16 .engine files specifically tuned for the user's GPU architecture (Turing, Ampere, Ada, or Blackwell).

Real-time hardware monitoring is handled via pynvml.

Metrics: GPU Core Util, VRAM usage, Temperature, Power Draw (Watts), and Clock Speeds.
Performance Advisor: Recommends DLSS, Frame Generation, Reflex, and Path Tracing settings based on live bottleneck analysis.
Feature Verification: Automatically detects support for Overdrive Mode and hardware-accelerated Path Tracing for compatible GPUs.

Mission Control is future-proofed for the latest hardware features:

DLSS 3.x/4.x: Monitoring and advice for Frame Generation and Multi-Frame Gen.
Blackwell (RTX 50-series): Specialized support for architectural features and thermal profiles of the latest GPUs.
NVIDIA Reflex: Integrated latency monitoring to ensure sub-ms input response.