kennss/SiliconScope

Sudoless Apple Silicon system monitor (native SwiftUI GUI) with ANE / Media Engine / memory-bandwidth tracking

What it solves

SiliconScope is a system monitor for Apple Silicon Macs that surfaces deep hardware metrics that standard tools like Activity Monitor cannot. It specifically targets the needs of developers working with on-device AI and media workloads by providing real-time visibility into the Neural Engine (ANE), Media Engine, and memory bandwidth—the primary bottlenecks for local LLMs and AI models.

How it works

The app is a native SwiftUI dashboard and menu-bar suite. It accesses hardware data using a combination of public and private Apple APIs without requiring root (sudo) privileges:

  • IOReport Framework: Used to extract per-domain power (CPU, GPU, ANE, DRAM), cluster residency, and memory bandwidth.
  • SMC & HID: Uses curated per-generation maps of SMC FourCC keys and HID sensor sets to provide accurate per-unit temperatures for E-cores, P-cores, GPU, and memory.
  • Process Inspector: Utilizes proc_pid_rusage to show per-process power, IPC, and specifically the Neural-Engine memory footprint.
  • Fleet Management: A remote agent can be installed on other Macs or Linux/NVIDIA boxes to stream metrics (including ANE for remote Macs) to a central dashboard via TLS-encrypted connections.

Who it’s for

  • AI Developers: Those running local LLMs or CoreML models who need to know if their workload is compute-bound, bandwidth-bound, or thermally throttled.
  • Power Users: Mac users who want a detailed, iStat-style menu-bar monitor with deep hardware insights.
  • ** uma Headless Mac/Linux Admins**: Users managing multiple machines who want a single pane of glass for their GPU and AI accelerator utilization.

Highlights

  • ANE & Media Engine Tracking: Real-time power and bandwidth monitoring for Apple's AI and media accelerators.
  • Process Inspector: Sudoless per-process metrics including P/E core split, IPC, and Neural-Engine memory usage.
  • Record & Replay: A DVR-like feature that records system metrics to a .ssrec file for later analysis and scrubbing.
  • AI Workload Classifier: Automatically identifies if a system is bandwidth-bound, compute-bound, or thermally throttled.
  • Fleet Support: Remote monitoring for other Macs (including ANE) and Linux/NVIDIA boxes (showing VRAM and Ollama models).
  • Menu-bar Suite: Highly customizable, iStat-style glyphs and dropdowns for every major system metric.

Related

  • Project
    Sentient-OS-Labs/sentient-os

    An on-device AI operating system for macOS that builds a private, local knowledge base of your digital life to power proactive agents and computer-use automation.

  • Project
    Layr-Labs/d-inference

    Darkbloom (d‑inference) is an open‑source, decentralized inference platform that lets owners of Apple‑silicon Macs rent out idle GPU resources as an OpenAI‑compatible backend. A Go coordinator runs in a GCP Confidential VM, handling authentication, billing, and routing, while a hardened Swift CLI on each Mac runs models in‑process via MLX. Every request is end‑to‑end encrypted (NaCl Box) and the provider process is locked down with the Hardened Runtime, SIP, and kernel‑level debugger denial, ensuring the Mac’s owner cannot read prompts or responses. Consumers use the standard OpenAI/Anthropic API by pointing their SDKs to `https://api.darkbloom.dev/v1`. Providers install the CLI with a one‑liner, link their account, and earn per‑token revenue; self‑route and direct modes enable free, private inference on one’s own machine. Pricing is roughly half of major hosted APIs, and the system is currently in public alpha.

  • Project
    screenpipe/screenpipe

    A local-first screen and audio recording tool that creates a searchable AI memory of your computer activity to provide context for AI agents.

  • Project
    FluidInference/FluidAudio

    A Swift SDK for local, low-latency audio AI on Apple devices, providing transcription, text-to-speech, and speaker diarization offloaded to the Apple Neural Engine.

  • Project
    soniqo/speech-swift

    Speech Swift is an open‑source Swift package that provides dozens of on‑device speech AI models (ASR, TTS, translation, voice agents, enhancement, etc.) for Apple Silicon. It runs via Core ML and the MLX framework, supports hundreds of languages, offers multiple quantised sizes, and can be added to a project with a single Swift‑Package‑Manager dependency. The repo includes extensive docs, pre‑trained models, and example code for transcription, streaming dictation, and UI components, all without any cloud calls.