🌙 OpenSleep

  

---

OpenSleep is a next-generation, premium sleep tracking application that redefines personal sleep science through a uncompromising 100% on-device, private-first architecture.

Unlike standard trackers that transmit your personal bio-data to remote servers, OpenSleep does all of its telemetry processing and advanced AI analysis directly on your smartphone. Featuring a fully integrated, localized Gemma 4 AI Sleep Coach, OpenSleep delivers personalized wellness guidance with complete cloud isolation.

---

🚀 Key Pillars

1. 🛡️ 100% On-Device & Zero-Server Privacy

• Local Active Telemetry: Ultrasonic sonar echo processing and motion actigraphy are analyzed completely on-device. Raw audio is processed in memory and never stored or transmitted.
• No External Servers: Zero analytics endpoints, zero tracking SDKs, and zero database synchronization in the cloud. Your data belongs to you alone.
• Health Integration: Syncs directly and securely with native system aggregates (Apple Health on iOS, Google Health Connect on Android) through highly secure local APIs.

2. 🧠 Local Gemma 4 AI Coach

• Fully Offline Intelligence: OpenSleep runs a highly optimized, localized Gemma 4 LLM directly on-device. No API requests, no cloud latency, and no chance of conversation leakage.
• Smart Sleep Analytics: The AI coach reads your sleep patterns locally to offer customized tips for improving sleep hygiene, tracking circadian rhythm, and managing daytime sleepiness.
• Dynamic Chat Interface: A gorgeous, reactive conversation dashboard supporting markdown rendering, tabular sleep summaries, and granular context window controls.

3. 📊 High-Precision Dual-Mode Sleep Tracking

• Contactless Sonar (Active Ultrasonic): Uses the device speaker to emit inaudible high-frequency sound chirps (~18 kHz - 22 kHz) and records the reflected echo via the microphone. An on-device DSP pipeline (high-pass Butterworth filtering, FFT power spectrum analysis, and correlation mapping) detects respiration rates and body movement contactless from a nightstand.
• Mattress Actigraphy (3-Axis Motion): Uses low-latency accelerometer and gyroscope updates to record physical body movement when the phone is placed flat on the mattress.
• Scientific Sleep Staging: Combines actigraphy and sonar activity to compute precise Deep, Light, REM, and Awake sleep stages through aligned heuristic models.

---

🏗️ Architecture & Data Flow

Below is the design of the on-device environment showing the active tracking loops and cloud isolation:

graph TD
    subgraph Device ["User Device (iOS & Android)"]
        subgraph Sensors ["Sensor Layer"]
            Speaker["Ultrasound Chirps<br>(18 - 22 kHz)"]
            Mic["Microphone Echo Input"]
            Motion["3-Axis Motion Sensors<br>(Accel & Gyro)"]
        end

        subgraph DSP ["On-Device DSP Pipeline"]
            IIR["Butterworth IIR Filter"]
            FFT["Real FFT Power Spectrum<br>(vDSP / JTransforms)"]
            Aggregator["Low-Level Activity Aggregator"]
        end

        subgraph Staging ["Sleep Staging & Sync"]
            Analyzer["Sleep Stage Analyzer"]
            Storage["Local DB & Sync<br>(Apple Health / Health Connect)"]
        end

        Speaker -.->|Acoustic Echo| Mic
        Mic -->|Raw PCM Buffer| IIR
        IIR -->|Filtered Signal| FFT
        FFT -->|Correlation & Spectrum| Aggregator
        Motion -->|Direct Actigraphy| Aggregator
        Aggregator -->|Normalized Activity| Analyzer
        Analyzer -->|Sleep Stages| Storage
        
        Storage -->|Contextual Sleep Data| AI["Local Gemma 4 AI Coach<br>(On-Device LiteRT-LM)"]
        User["User Interaction"] <-->|Chat Interface| AI
    end
    
    subgraph Cloud ["External Network"]
        direction LR
        Server["External Server"] -.->|PROHIBITED / BLOCKED| Device
    end
    
    style Device fill:#0f172a,stroke:#3b82f6,stroke-width:2px,color:#fff
    style Cloud fill:#1e293b,stroke:#f43f5e,stroke-width:1px,stroke-dasharray: 5 5,color:#cbd5e1
    style Sensors fill:#1e293b,stroke:#10b981,stroke-width:1px,color:#fff
    style DSP fill:#1e293b,stroke:#8b5cf6,stroke-width:1px,color:#fff
    style Staging fill:#1e293b,stroke:#3b82f6,stroke-width:1px,color:#fff
    style AI fill:#1e293b,stroke:#ec4899,stroke-width:2px,color:#fff
    style User fill:#334155,stroke:#94a3b8,stroke-width:1px,color:#fff
    style Server fill:#1e293b,stroke:#f43f5e,stroke-width:1px,color:#94a3b8

Loading

---

🛠️ Technology Stack

Platform	Frontend	Local Storage	AI Engine
iOS	Swift & SwiftUI (Premium UI Design)	SwiftData & Apple HealthKit	LiteRT-LM (Swift Package)
Android	Kotlin & Jetpack Compose (Material 3)	Room & Google Health Connect	LiteRT-LM (Kotlin Library)

Important

OpenSleep does not connect to the internet. It never collects, stores, or sells your personal information, sleep logs, audio metadata, or conversation histories. All AI calculations are run directly on the physical processor of your device.

---

📬 Support & Community

OpenSleep is an open-source project created and maintained for the benefit of developers and health enthusiasts alike.

• GitHub Repository: https://github.com/timmyy123/opensleep
• Support Email: If you have any inquiries, feedback, or need assistance, feel free to reach out directly to the creator at timmy@opensleep.tech.

---

OpenSleep — Your Sleep, Your Data, 100% On-Device.