Windows Systems Utility AI-Powered Intelligence v1.0.0 Release Windows 10 / 11 (x64)

NetX — AI Network Intelligence

Standalone Windows desktop network utility combining live OS hardware counters, socket-to-process inspection, and an interactive Gemini-powered AI diagnostic troubleshooter for automated connection recovery.

NetX — AI Network Troubleshooter & Assistant
NetX interactive AI network troubleshooter and diagnostics view
NetX — Real-Time Telemetry HUD & Bandwidth Monitor
NetX live running desktop interface showing real-time interface telemetry, bandwidth graphs, and local network diagnostics
Real Windows Kernel & Subsystem Counters ZERO SIMULATED VALUES
Interface Bandwidth .NET GetIPStatistics() & netstat -e
Process Sockets netstat -ano TCP/UDP + tasklist PIDs
Wi-Fi Radio Telemetry netsh wlan show interfaces (RSSI/PHY)
Active Subnet Sweep arp -a + Parallel ICMP Ping

Overview & Problem Statement

Network diagnostic utilities for Windows are typically bifurcated into two extremes: overly complex enterprise packet analyzers (like Wireshark) that overwhelm users with low-level protocol dissects, or superficial consumer widgets that display simulated animations or inaccurate ping meters. Developers and network administrators need an unbloated desktop utility that answers essential questions quickly: Which executable is consuming bandwidth right now? What are the live radio characteristics of the current Wi-Fi link? And how many devices are active on the local subnet?

Key Implemented Features

  • AI Network Troubleshooter & Diagnostics: Integrated conversational network assistant powered by Google Gemini API and custom aiToolRegistry. Performs automated multi-hop latency triage, DNS resolution verifications, rogue process detection, and guided network remediation plans.
  • Process-to-Socket Mapping: Correlates active TCP/UDP sockets from netstat -ano with running Windows processes queried via tasklist /FO CSV. Performs asynchronous reverse-DNS lookups to resolve remote server hostnames, exposing unauthorized outbound connections instantly.
  • Persistent Local Usage Store: Writes rolling byte counters to %APPDATA%\NetX\netx-usage-store.json across hourly, daily, and monthly intervals, providing a verifiable local history of internet data consumption.
  • Wi-Fi Spectrum & BSSID Analyzer: Executes netsh wlan show networks mode=bssid to detect surrounding 2.4 GHz and 5 GHz wireless channels, authentication ciphers (WPA2/WPA3), signal strengths (dBm), and channel congestion.
  • Edge Speed Benchmarking: Measures true throughput against Cloudflare's global edge network (speed.cloudflare.com) using concurrent HTTPS streams, recording real download and upload transfer rates with millisecond jitter calculation.
  • Interactive Diagnostic Suite: Built-in DNS resolver (testing A, AAAA, MX, NS, TXT records), live ICMP Traceroute, and raw TCP port handshakes via Node net.Socket.

My Work & Engineering Contributions

I designed and built NetX as a modern native Windows systems telemetry utility:

  • AI Diagnostics & Tool Calling Architecture: Built the AiAssistantView and tool execution engine enabling deterministic LLM function calls (DNS lookups, ICMP pings, socket inspection) with conversational confirmation dialogs.
  • Windows Telemetry Engine: Developed background services querying native GetIPStatistics counters and child process streams without admin elevation.
  • Correlation Algorithms: Implemented real-time socket-to-process parsing, linking port endpoints with executable binaries and window handles.
  • Desktop HUD & View Suite: Designed the React 18 frontend featuring a live bandwidth HUD, command palette (Ctrl+K), and responsive views for WiFi, LAN, and speed testing.
  • Packaging & Distribution: Configured electron-builder pipelines generating zero-dependency standalone Windows executables.

Technical Implementation & Architecture

NetX decouples heavy operating system execution into child process workers, delivering real-time telemetry over an IPC channel:

  • Tool-Grounded AI Registry: Modeled a strict TypeScript tool registry (aiToolRegistry.ts) defining JSON schema contracts for diagnostics tools. The LLM cannot hallucinate arbitrary commands; each diagnostic action is dispatched through validated system execution handlers.
  • Differential Polling Architecture: Fast interface counters poll every 1,000ms while heavier operations (ARP sweeps, reverse-DNS lookups) are executed on-demand in separate worker queues.
  • Telemetry Local Storage: Rolling bandwidth history writes to an atomic JSON file store in AppData, avoiding large embedded database runtimes.

Engineering Challenges & Solutions

  • Challenge: CPU Throttling During Rapid Subnet Scanning: Probing an entire /24 subnet sequentially can stall desktop processes for several minutes.
    Solution: Implemented parallel asynchronous ICMP ping sweeps with 800ms timeouts, batching ARP table queries to discover active subnet hosts in under 3 seconds.
  • Challenge: Accurate Non-Elevated Telemetry: Many Windows telemetry tools require full Administrator rights, creating security warnings on startup.
    Solution: Engineered the metrics collection layer around standard Win32 user-mode APIs (.NET NetworkInterface.GetIPStatistics() and netsh), capturing 100% genuine hardware telemetry without requiring elevated permissions.