Risk Colorwheel — Perceptual, Personal, Live

Meet your LLM-guided dial. The wheel blends many signals into a single reading: a smooth harm ratio from tranquil green → amber drift → alert crimson. In Guess mode, the server (via psutil) samples CPU/RAM & jitter entropy as context; in Route mode, you add coordinates so the model can reason about the trip. The wheel’s breathing changes with risk—slower & softer when clear, deeper & brisk when elevated.

Open Dashboard Your tone: A3 Strict-JSON LLM PSUTIL signals
--%
INITIALIZING
Calibrating renderer…

Tip: if your OS has “Reduce Motion”, animations automatically calm down.

How it decides

We call the LLM in two ways and always require strict JSON back. A lightweight worker smooths noise, and the wheel renders a perceptual color ramp with a risk-linked breathing halo and ring glow.

  • LLM Guess — server bundles psutil CPU %, RAM %, loadavg, and a tiny entropy sketch; no destination.
  • LLM Route — includes lat/lon → dest_lat/dest_lon so the LLM can reason about the segment you care about.
Source
Why this reading Conf: --%
  • Waiting for LLM response…
Perceptual ramp

Colors blend in OK-ish space so equal changes feel equal. We also tint ~18% toward your accent for subtle identity.

Breathing & calm tech

The halo’s breath rate and amplitude follow risk & confidence. Clear → slower and softer; elevated → quicker and fuller.

Privacy by design

Guess mode never sends a destination. Route mode only sends the numbers you provide. Accent hue is local, not uploaded.