RuView

ruvnet/RuViewReviewed by AI

What it does

A WiFi sensing platform. Small ESP32 boards (about $9 each) read how radio waves bounce around a room, and the software turns that into presence, movement, breathing and heart-rate readings, plus alerts like falls, and publishes them to Home Assistant, Apple Home, Google Home or Alexa. There is a Docker image with simulated data to try it without hardware. For tinkerers: you buy and flash boards, run a server and calibrate. Free and MIT-licensed; the project itself labels several of its models and vital-sign readings as unvalidated.

How easy it is to install

D · terminal, several tools or a specific graphics card · Windows

Windows D
macOS D
Linux D

D on Windows, Mac and Linux alike: the quickest look is a Docker image fed with simulated data (and it refuses to start without a token you generate yourself), while anything real means buying ESP32 boards, flashing firmware with esptool or ESP-IDF, provisioning WiFi, running a server and calibrating the room. A laptop with no CSI-capable board only gets coarse presence from signal strength.

Coming soon: install in 1 click with the RepoPicks app.

Numbers on GitHub

Stars: 101,882on Oct 31, 2026
Downloads: 110,071551 this week
Growth: +2,121stars in 5 days

Stars over time

Facts

Runs on: Windows, macOS, Linux
You need to install: Docker, Python, Other tools
To build from source: Python, NVIDIA graphics card
Disk space: 6 MB–17 MB · from the installer
Accounts or keys: None
Cost: Free · The software is MIT and runs locally, with no cloud account. You pay for hardware: the README quotes about $9 for one ESP32-S3 node, around $54 for a mesh of 3-6 nodes, and about $140 for a full kit with a Cognitum Seed.
License: MIT (Permissive)
Can replace: Origin Wireless
Last update: Oct 29, 2026

Risk signals (6)

AttentionInstall command runs a script from the internet
Install command pipes a remote script into a shell: curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
automatic check
Attentionhealth and safety readings are not validated
The project's own MetaHarness safety list says "Device-reported vitals are labelled as unvalidated", yet the module catalogue offers `sleep-apnea`, `seizure-detect`, `cardiac-arrhythmia` and `respiratory-distress`; the `smoke-fire` module states it is "Not a UL-listed replacement for code-required smoke alarms." Nothing here is a medical or life-safety device.
AI review
Attentionadvertised pose feature is not wired up
README, "Model weights: what's real, what's not": the committed on-device pose model scores "PCK@20 = 3.0%", below the project's own ≥35% target, and "Its runtime path in `cog-pose-estimation/src/inference.rs` is still a centred-skeleton stub returning `confidence=0` — the weights are not yet wired in." The published encoder file also fails to load ("the published file's header is NUL-padded, which the reference `safetensors.torch.load_file` rejects"), and an earlier "100% presence" figure was retracted.
AI review
Attentionwatching people who have not agreed to it
README: "See through walls with WiFi ... Detect people, measure breathing and heart rate ... with no cameras or wearables", "no app on the user's phone", and multi-channel scanning that "uses your neighbors' routers as free radar illuminators". Monitoring people, or leaning on other people's networks, is restricted by privacy law in many countries; check the rules where you live before pointing it at a room.
AI review
Attentionexample configuration ships wide open
`example.env` sets `HOST=0.0.0.0`, `ALLOWED_HOSTS=*`, `CORS_ORIGINS=*`, `DEBUG=true`, `ENABLE_AUTHENTICATION=false`, `ENABLE_TEST_ENDPOINTS=true` and `SECRET_KEY=your-secret-key-here-change-for-production`. The file lists ten things to change for production, but a copied-as-is server on your network has no login at all.
AI review
Infovery large claim surface and release churn
367 GitHub releases (three of them tagged on 2026-10-26), 504 open issues, and a README offering a "105-cog catalog" from vital signs to `ghost-hunter` and `weapon-detect`. The project keeps its own honesty tables for what is measured and what is not, which is also a sign of how much is claimed.
AI review

Evidence is shown in its original language.

What Radar could not check

Whether presence, breathing and heart rate work in a normal home, outside the author's measurements.
Whether through-wall sensing reaches the "up to ~5 m" the README mentions in your own walls.
Whether firmware flashing and WiFi provisioning on an ESP32-S3 or C6 go as described.
Real disk and memory use once models and the Python stack are installed.
Whether the signed edge modules do what their one-line descriptions say, and what they send to `seed.cognitum.one`.
Whether using it on a given room, or on neighbouring networks, is legal where you live.