RuView Presence

Presence-aware AI agents powered by WiFi sensing. No cameras. No wearables. Just physics.

Your agents detect whether you're physically present using WiFi Channel State Information (CSI). When you leave, they queue non-urgent messages. When you return, they greet you with a digest of everything that happened while you were away.

---

How It Works

State	What Happens
Present	Agents operate normally
Away	Non-urgent messages are queued; urgent ones sent immediately
Returned	Agents deliver a welcome-back digest, then resume normal operation

---

Quick Start

1. Start RuView

docker run -d -p 3000:3000 --name ruview ruvnet/wifi-densepose:latest

Verify it's running:

curl -s http://localhost:3000/health/live
# {"status":"alive","uptime":4}

Works in simulation mode out of the box — no WiFi hardware needed for testing. The source field in API responses will show "simulate" (synthetic data) vs "csi" (real hardware).

2. Install the Plugin

# From local clone
git clone https://github.com/DevvGwardo/openclaw-ruview-presence.git
openclaw plugins install ./openclaw-ruview-presence

# Or link for development
openclaw plugins install -l ./openclaw-ruview-presence

3. Configure

Add to your openclaw.json:

{
  // Plugin configuration (runtime logic)
  "plugins": {
    "entries": {
      "ruview-presence": {
        "enabled": true,
        "config": {
          "ruviewUrl": "http://localhost:3000",
          "confidenceThreshold": 0.3,
          "debounceCount": 2,
          "enableDigest": true
        }
      }
    }
  },

  // Skill configuration (agent instructions)
  "skills": {
    "entries": {
      "ruview-presence": {
        "enabled": true,
        "env": {
          "RUVIEW_API_URL": "http://localhost:3000"
        }
      }
    }
  }
}

4. Install the Bundled Skill

cp -r skills/ruview-presence ~/.openclaw/skills/ruview-presence

That's it. Your agents will start checking presence on their next heartbeat. You'll see this in the logs:

config change detected; evaluating reload (skills)
config change applied (dynamic reads: skills)

---

Configuration Reference

Option	Type	Default	Description
ruviewUrl	string	http://localhost:3000	RuView API base URL
pollIntervalMs	number	10000	How often to poll RuView (ms)
confidenceThreshold	number	0.3	Minimum detection confidence to count as "present"
debounceCount	number	2	Consecutive empty readings before marking as "away"
enableDigest	boolean	true	Show a summary digest when the user returns
enableZoneAwareness	boolean	false	Track which room/zone the user is in

All options can also be set via environment variables:

Environment Variable	Maps To
RUVIEW_API_URL	ruviewUrl
RUVIEW_API_KEY	Auth token (if RuView auth is enabled)

---

RuView API Data

These are the actual responses from RuView that the plugin works with.

Pose Detection (GET /api/v1/pose/current)

The primary endpoint used for presence detection.

{
  "timestamp": 1773088911.824,
  "source": "simulate",
  "total_persons": 1,
  "persons": [
    {
      "id": 1,
      "confidence": 0.78,
      "zone": "zone_1",
      "bbox": { "x": 270.4, "y": 133.2, "width": 136.6, "height": 235.1 },
      "keypoints": [
        { "name": "nose", "confidence": 0.59, "x": 336.5, "y": 151.8, "z": -0.22 },
        { "name": "left_eye", "confidence": 0.61, "x": 326.1, "y": 146.0, "z": -0.22 },
        { "name": "right_eye", "confidence": 0.66, "x": 346.1, "y": 143.2, "z": -0.22 },
        { "name": "left_shoulder", "confidence": 0.77, "x": 311.3, "y": 185.6, "z": -0.22 },
        { "name": "right_shoulder", "confidence": 0.74, "x": 360.6, "y": 186.9, "z": -0.22 }
      ]
    }
  ]
}

Each person includes 17 DensePose-compatible keypoints: nose, left/right eye, left/right ear, left/right shoulder, left/right elbow, left/right wrist, left/right hip, left/right knee, left/right ankle.

Zone Summary (GET /api/v1/pose/zones/summary)

{
  "zones": {
    "zone_1": { "person_count": 1, "status": "monitored" },
    "zone_2": { "person_count": 0, "status": "clear" },
    "zone_3": { "person_count": 0, "status": "clear" },
    "zone_4": { "person_count": 0, "status": "clear" }
  }
}

Vital Signs (GET /api/v1/vital-signs)

{
  "vital_signs": {
    "breathing_rate_bpm": 9.4,
    "breathing_confidence": 0.83,
    "heart_rate_bpm": 44.4,
    "heartbeat_confidence": 0.67,
    "signal_quality": 0.52
  },
  "source": "simulate",
  "tick": 19612
}

Full Sensing Data (GET /api/v1/sensing/latest)

Returns everything above plus raw signal features (mean RSSI, spectral power, motion/breathing band power), per-sensor subcarrier amplitudes, RF tomography voxel grid, and classification.

{
  "classification": {
    "presence": true,
    "motion_level": "present_still",
    "confidence": 0.78
  },
  "features": {
    "mean_rssi": -37.0,
    "variance": 15.1,
    "spectral_power": 249.0,
    "dominant_freq_hz": 1.85,
    "breathing_band_power": 16.4,
    "motion_band_power": 13.7,
    "change_points": 8
  }
}

Health Check (GET /health/live)

{ "status": "alive", "uptime": 1961 }

---

Architecture

Plugin Layer (index.ts)

Hooks into OpenClaw's before_prompt_build lifecycle event to poll RuView and manage state transitions. When the user returns after being away, the plugin prepends a digest summary to the agent's context.

Types match the actual RuView API response format — RuViewPoseResponse, RuViewZoneSummary, RuViewVitalSigns, and RuViewSensingLatest are all typed to the real payloads.

Skill Layer (skills/ruview-presence/)

Provides standing orders that agents follow during heartbeat cycles. Gives agents explicit instructions for presence-aware behavior — checking the API, interpreting results, and acting on state changes. Includes the actual JSON response format so agents can parse responses correctly.

Gateway RPC Methods

The plugin exposes two methods on the OpenClaw gateway for programmatic access:

Method	Returns
ruview.presence	{ state, zone, awaySince, queuedEvents, detectedPersons, source }
ruview.queueEvent	{ queued: true, total: <count> }

---

Hardware Options

RuView runs in simulation mode by default (source: "simulate" in responses). For real-world presence detection (source: "csi"):

Option	Hardware	Cost	Capability
No hardware	Any computer	$0	Simulation mode (synthetic data)
Basic	Any WiFi laptop	$0	RSSI-only presence (coarse)
Recommended	3-6x ESP32-S3 + router	~$54	Full CSI: pose, breathing, heartbeat, motion
Research	Intel 5300 / Atheros AR9580	~$50-100	Full CSI with 3x3 MIMO

---

Fault Tolerance

• If RuView is unreachable, the plugin keeps the last known state and retries on the next heartbeat
• The debounce mechanism prevents false "away" triggers from momentary signal drops (requires 2 consecutive empty readings by default)
• Urgent messages are always delivered immediately, regardless of presence state
• The source field in RuView responses distinguishes simulation from real hardware data

---

Troubleshooting

Problem	Solution
Agent says "rate limit reached"	Your model provider is rate-limited. Switch to a different model with openclaw models set <model>
Agent tries to use read() tools	The Chat Completions API doesn't support tool calls. Send the agent the data directly or use the Responses API
Port 3000 is already in use	Map to a different port: docker run -d -p 3002:3000 --name ruview ruvnet/wifi-densepose:latest and update ruviewUrl in config
Skill not detected	Check openclaw logs for config change detected; evaluating reload (skills) — if missing, restart the gateway
Always shows "present"	In simulation mode, RuView always returns a synthetic person. Use real CSI hardware for actual presence/absence detection

---

Project Structure

openclaw-ruview-presence/
  index.ts                 Plugin entry point (state machine, polling, digest)
  openclaw.plugin.json     Plugin manifest and config schema
  package.json             Package definition
  docs/
    hero-banner.png        README hero banner
    architecture.png       System architecture diagram
    statemachine.png       State machine diagram
  skills/
    ruview-presence/
      SKILL.md             Agent standing orders with API response formats
      HEARTBEAT.md         Heartbeat trigger

---

Requirements

• OpenClaw agent runtime (>=2026.3.1)
• RuView sensing server (Docker or native)

License

MIT

---

_{Built with RuView WiFi sensing and OpenClaw agent runtime.}