ATLAS System Stats Bridge

The HWiNFO Bridge's sibling: streams whole-PC health - CPU, RAM, GPU, disk, network - into ATLAS, using the same pipeline, for the readings HWiNFO doesn't cover on its own.

The problem it solves

HWiNFO is deep on hardware sensors but doesn't see the OS layer - per-process load, disk queue depth, network throughput as the OS reports it, swap usage. This bridge reads that layer directly via psutil (CPU/RAM/disk/network) and pynvml (NVIDIA GPU), and streams it through the exact same ATLAS Stream API and Kafka broker the HWiNFO Bridge uses. The two are deliberately kept as siblings rather than merged: same ports, same Docker stack, same web-UI shape, so switching between "what's the hardware doing" and "what's the OS doing" costs nothing extra to learn.

ATLAS System Stats Bridge web UI showing connection status, essential preset selected, an analysis note, and grouped live telemetry for CPU, memory, GPU, disk and network

Same UI shape as the HWiNFO Bridge, streaming psutil/pynvml readings instead of HWiNFO's shared memory.

What's involved

Channel presets by category

Essential, CPU, Memory, Disk, Network, GPU, or Full - each preset is a named subset of a fixed channel catalogue (unlike the HWiNFO Bridge, whose channels are discovered live), since psutil/pynvml readings are the same shape on every machine.

Per-core and per-GPU channels

CPU core count and GPU count are detected at startup, so the channel catalogue - and the "full" preset - scales automatically to however many cores and NVIDIA GPUs the host actually has.

Delta-computed throughput

Disk and network throughput and IOPS are derived from the difference between consecutive samples, so the very first sample after starting a session reports no data for those channels - a deliberate, documented tradeoff rather than a bug.

Auto-reconnect

If the Stream API connection drops mid-session, the bridge reconnects on its own rather than requiring the driver to notice and restart it mid-race.

Standalone infrastructure, shared design

Same Docker stack as the HWiNFO Bridge - Kafka, the ATLAS Stream API, and a Key Generator service - with the bridge itself running natively on Windows. Only one of the sibling bridges can run at a time, since they share the same default ports; which one's active depends on whether the session needs hardware-sensor depth or OS-level system health.