Files
ruvnet--RuView/python
Dragan Spiridonov bc0e8fd031 test(python): close the three parity-review findings (native anchor in CI, NaN, cosine)
A cross-vendor review of the parity rework (2febbb81) found three issues; all
verified and fixed here.

1. HIGH — the native≈golden half never ran in CI, so a binding marshalling bug
   could pass. The golden vectors were regenerated through the Python binding,
   and only pytest (binding vs that golden) runs in CI — the native reference
   tests in `python/tests/aether_parity.rs` link against the PyO3 crate and no
   workflow runs them. A stable PyO3 conversion defect present at regeneration
   would therefore be baked into the golden and go undetected.

   Fix: a real native parity test IN the `wifi-densepose-aether` crate
   (`tests/golden_parity.rs`), which is std-only and a member of the v2
   workspace, so it runs under the existing `cargo test --workspace`. It
   recomputes the embedding with no PyO3/marshalling and asserts it matches the
   SAME committed golden within tolerance. Now native≈golden AND binding≈golden
   both run in CI ⇒ binding≈native, and a binding-specific artifact in the
   golden surfaces here as a native mismatch. serde_json added as a
   dev-dependency only (test-only; never linked into the lib or the wheel, so
   the crate stays runtime-dependency-free).

   Independently proven at this commit: native output equals the committed
   golden BIT-FOR-BIT (128/128 exact bits, Δ=0) for both base and loaded — the
   golden is native-faithful today; this test keeps it that way.

2. MEDIUM — the Python parity helper passed non-finite output. `abs(nan - b) >
   tol` is False, so an all-NaN embedding slipped through. Now every element
   must be `math.isfinite` first. Proven: an all-NaN vector is rejected
   ("element 0 is not finite (nan)"); the Rust helpers already caught it via
   `<=`.

3. LOW — a coherent shift inside the per-element tolerance could move the whole
   vector undetected. Added a whole-vector cosine-similarity bound (≥ 1 - 1e-6)
   alongside the per-element check.

Verified on aarch64/macOS: `test_aether` 13, full `python/tests/` suite 227
passed against a `--features sota` build; the new native test passes
(2/2) run standalone (the in-worktree `cargo test -p` only failed on an
un-checked-out submodule; ci.yml checks out submodules recursively).

Co-Authored-By: Ruflo & AQE
2026-07-24 12:29:23 +02:00
..

wifi-densepose

PyPI version Python License: MIT

Detect human presence, count people, read breathing and heart rate, and estimate skeletal pose — using only the WiFi signal already in your home.

No cameras. No wearables. Works through walls and in the dark.

wifi-densepose is the Python binding for the RuView sensing stack: a Rust core that turns the Channel State Information (CSI) emitted by ordinary WiFi chips into ambient-intelligence signals. The wheel ships compiled DSP for fast offline analysis, plus an opt-in Python client for talking to a live RuView sensing-server over WebSocket or MQTT.

Features

  • 17-keypoint pose — full-body skeletal estimate from WiFi CSI, no camera
  • Vital signs — respiratory rate (630 BPM) and heart rate (40120 BPM) with a confidence score and clinical-grade / degraded / unreliable status
  • Presence, person count, fall detection, motion — fused outputs from the same CSI stream
  • 10 semantic primitives (HA-MIND) — someone-sleeping, possible-distress, room-active, bathroom-occupied, fall-risk-elevated, bed-exit, … — ready to wire into Home Assistant or Apple Home automations
  • Beamforming Feedback (BFLD) support — 802.11ac/ax/be compressed feedback matrices on top of the receiver-side CSI path
  • GIL-releasing DSP — extract loops run with the GIL released, so a tokio-backed web server can call into the pipeline without stalling its event loop
  • Tiny wheel — ~240 KB compiled (one binary per OS/arch covers Python 3.10+ via the stable ABI)

Install

pip install wifi-densepose                 # core DSP only
pip install "wifi-densepose[client]"       # + WebSocket/MQTT clients

Wheels are published for Linux (x86_64, aarch64), macOS (x86_64, arm64), and Windows (amd64).

SOTA extras (ADR-185)

Three optional subsystems bind the Rust SOTA modules as compiled-feature wheels. Each raises a clear ImportError if you import it without the extra:

Extra Module What it adds
[aether] wifi_densepose.aether Contrastive CSI embeddings / re-identification (ADR-024) — EmbeddingExtractor, cosine_similarity, info_nce_loss
[meridian] wifi_densepose.meridian Cross-environment domain generalization (ADR-027) — HardwareNormalizer, GeometryEncoder, RapidAdaptation, CrossDomainEvaluator
[mat] wifi_densepose.mat Mass-Casualty Assessment disaster-survivor detection + START triage — DisasterResponse, Survivor, TriageStatus
[sota] all three Convenience superset
pip install "wifi-densepose[aether]"       # re-identification embeddings
pip install "wifi-densepose[meridian]"     # cross-room calibration
pip install "wifi-densepose[mat]"          # disaster triage
pip install "wifi-densepose[sota]"         # all three

Runnable examples: examples/reid_from_csi.py, examples/cross_room_calibrate.py, examples/mat_triage.py.

Usage

Extract breathing rate from a CSI stream

from wifi_densepose import BreathingExtractor

br = BreathingExtractor.esp32_default()     # 56 subcarriers @ 100 Hz, 30s window

for residuals, weights in your_csi_source:  # one frame at a time
    est = br.extract(residuals=residuals, weights=weights)
    if est is not None:
        print(f"{est.value_bpm:.1f} BPM  (confidence={est.confidence:.2f})")

Heart rate is the same shape — HeartRateExtractor.esp32_default() with a 0.82.0 Hz band-pass and a 15-second window.

Subscribe to a live sensing-server

import asyncio
from wifi_densepose.client import SensingClient, EdgeVitalsMessage

async def main():
    async with SensingClient("ws://your-ruview-node:8765/ws/sensing") as c:
        async for msg in c.stream():
            if isinstance(msg, EdgeVitalsMessage):
                print(msg.presence, msg.breathing_rate_bpm, msg.heartrate_bpm)

asyncio.run(main())

React to Home Assistant semantic primitives

from wifi_densepose.client import (
    RuViewMqttClient, SemanticPrimitive, SemanticPrimitiveListener,
)

listener = SemanticPrimitiveListener()
listener.on(SemanticPrimitive.BedExit, lambda e: print("bed exit:", e.node_id))
listener.on(SemanticPrimitive.PossibleDistress, lambda e: alert(e))

client = RuViewMqttClient(broker_host="homeassistant.local")
client.on_message(
    "homeassistant/+/wifi_densepose_+/+/state",
    listener.handle_mqtt_message,
)
client.start()
client.wait_connected()

Decode 802.11ax beamforming feedback

import numpy as np
from wifi_densepose import BfldFrame, BfldKind

# Parse compressed BFR from a Wireshark capture into a Complex64 ndarray ...
fb = np.zeros((2, 1, 996), dtype=np.complex64)  # Nr=2 Nc=1 Nsc=996 for HE80

frame = BfldFrame.from_compressed_feedback(
    timestamp_ms=ts,
    sounding_index=seq,
    sta_mac="aa:bb:cc:dd:ee:ff",
    kind=BfldKind.CompressedHE80,
    feedback_matrix=fb,
)
print(frame.n_subcarriers, frame.mean_amplitude)

Hardware

Works with any WiFi chip that exposes CSI. Reference setups (ESP-IDF firmware, build scripts, witness-verified test bundles) are in the RuView repo:

Device Cost Role
ESP32-S3 (8MB flash) ~$9 WiFi CSI sensing node
ESP32-S3 SuperMini (4MB) ~$6 WiFi CSI (compact)
ESP32-C6 + Seeed MR60BHA2 ~$15 mmWave HR/BR/presence add-on

The legacy v1 line (Wi-Pose-style FastAPI server) is end-of-life; wifi-densepose==1.99.0 is a tombstone that raises ImportError pointing to v2 with a migration URL.

License

MIT.