Custom iOS/MacOS App for AI server dashboard- Prompt included

I have a Mac Studio M3 Ultra 512GB and 256GB connected with TB5. I have been monitoring my computers using Exos interface to see RAM usage and would have to go into oMLX in each computer via screen share to see what model is loaded. I decided to use Hermes and GLM-5.3-flash to make my own app for iOS and MacOS to combine the two views into one app. Not only can I see what each computer is using, I can also load and unload models through this. It also made a widget so I can keep on my phone or computer and I can also ask Siri questions about what models are loaded or usage. It’s amazing to wake up with an idea and have something custom built in a few hours. Prompt: Build a Local AI Cluster Monitor Use the prompt below with a capable coding agent. It is intentionally self-contained and does not assume that the user runs oMLX, uses two machines, or has the same network and hardware as anyone else. --- ## Prompt You are a senior Apple-platform engineer and systems-integration engineer. Build a polished, production-quality Local AI Cluster Monitor that lets me monitor one or more machines running local AI inference servers. The finished app should provide the same class of experience as a native cluster dashboard: - one card per compute node - live OS RAM usage and headroom - accelerator/GPU utilization, temperature, and system or accelerator power when the telemetry source exposes them - disk capacity and free space - installed, loaded, and loading AI models - load/unload controls when the selected AI server supports them - clear offline, unavailable, stale, and unsupported states -mixed-server support, so one node may run oMLX while another runs LM Studio, Ollama, exo, or another API -an adaptive dark dashboard for macOS, iPhone, and iPad if I choose those targets - optional widgets and Siri/App Shortcuts Do not hardcode somebody else's IP addresses, node names, API keys, bundle identifiers, Apple developer team, model names, machine sizes, or number of nodes. This must be configurable for my environment. ### Important working rule Do not start implementing immediately. First ask me the setup questionnaire below in one concise message. Explain unfamiliar choices briefly and recommend sensible defaults. Wait for my answers. If I do not know an endpoint or API version, offer to auto-detect it with safe read-only requests after I provide a base URL. Never test load, unload, delete, pull, or inference operations without explicit permission. ## Phase 0 — Setup questionnaire Ask me all applicable questions: 1. Product and platforms - What should the app be called? - Which targets do I want: macOS, iPhone/iPad, or both? -What Xcode version and minimum OS versions must it support? - Do I want a native SwiftUI app, or another client stack? Recommend SwiftUI for an Apple-only app. 2. Node inventory For each machine, ask for: - a stable ID and friendly display name - hardware/OS type: Apple Silicon macOS, NVIDIA Linux/Windows, AMD Linux/Windows, CPU-only, or other - installed RAM or unified memory - how the device is reached: LAN hostname/IP, Tailscale, VPN, or other - whether it has internal and/or external model storage that should be shown separately 3. AI inference server on each node Ask which server each node uses. Offer at least: -oMLX - LM Studio - Ollama - exo - generic OpenAI-compatible server - another/custom server Ask for each server's base URL, API version if known, and authentication mode: - none - Bearer token - custom header -cookie/session login - other Do not ask me to paste production secrets into chat if the app can collect them later through a secure settings screen. Use placeholders during development and have me enter real credentials in the built app. 4. Hardware telemetry Ask which telemetry source I want: - exo /state — recommended when I already run exo on an Apple Silicon cluster - a lightweight companion telemetry agent on every node — recommended when I use LM Studio/Ollama without exo or have mixed operating systems -Prometheus/node_exporter plus a GPU exporter - read-only OS-local metrics for the machine running the app- no hardware telemetry If I am unsure, use this recommendation: - Existing Apple Silicon exo cluster: use exo as the cluster-wide hardware telemetry source. - Apple Silicon machines not otherwise using exo: do not require the full exo inference stack only for monitoring; offer a small agent based on a maintained telemetry tool such as macmon, with a stable JSON wrapper owned by this project.