BusDK Update

bus-top: a live process monitor that explains your machine with AI

bus-top is an early live process and system monitor with AI built in. It combines native samplers, process classifiers, and focused explanations so an operator can understand what the machine is doing while it is happening.

The current work is experimental and practical in shape: live process rows remain useful on their own, while AI can add targeted context for selected processes, host conditions, and cached detail views when that assistance is available.

June 10, 2026bus-topprocess monitorAI explanations

At a glance

TL;DR

bus-top samples the running system and classifies process roles so the display carries operational context, not only raw counters.
Focused AI explanations can describe selected process families, host signals, and detail views without turning every row into an AI request.
A cache-only mode lets teams use stored explanations and labels when they want tighter control over live AI usage.

The monitor is aimed at day-to-day operations: watch the machine, notice the busy process, then ask for an explanation that is narrow enough to be useful. The current bus-top brings the pieces together: a configured AI backend, focused AI detail views, process-family selection for useful requests, cached labels in text snapshots, cache warm and cache-only operation, process classifier updates, and a native Darwin sampler.

bus-top
Bus Top  host workstation.local  sampler darwin-native  ai cache-only
PID    CPU   MEM   CLASS     NAME              NOTE
1284   7%    410M  browser   browser-helper    cached: render workload
2411   3%     96M  build     go test           local: compiling package
817    1%     64M  system    launchd           macOS service manager

The value is not that the interface hides system data. The value is that the numbers, names, and process families have room for operator-facing interpretation. A native sampler can keep the live view grounded in the local machine, classifiers can turn familiar daemons and workloads into readable roles, and AI can focus on explaining the part of the screen that needs attention.

The AI path is intentionally bounded. Cache-only mode keeps the monitor useful with previously stored explanations, and focused request selection avoids treating every process as equally important. That makes the feature usable in early deployments where teams are still deciding when live AI assistance belongs in an operations loop.

See docs.busdk.com for bus-top usage.