Add Pixel10 AI Server — expose on-device AI chip as REST API

Android app that turns a Pixel 10 into a free AI API server by leveraging the Tensor G5's on-device AI capabilities through an embedded HTTP server. Key components: - Dual AI backend: Gemini Nano (ML Kit Prompt API) + MediaPipe LLM fallback - OpenAI-compatible REST API (chat/completions, completions, models, health) - NanoHTTPD embedded server with CORS support and SSE streaming - Foreground service with wake lock for persistent background operation - Material Design 3 dashboard with live request logging All inference runs entirely on-device — zero cloud costs, full offline capability. https://claude.ai/code/session_01GvqMLSMmfMR8uz66BFVXX2
2026-02-28 14:09:41 +00:00
commit 12de200450
24 changed files with 1683 additions and 0 deletions
--- a/.gitignore
+++ b/.gitignore
@@ -0,0 +1,17 @@
+*.iml
+.gradle
+/local.properties
+/.idea
+.DS_Store
+/build
+/captures
+.externalNativeBuild
+.cxx
+local.properties
+/app/build
+/app/release
+*.apk
+*.aab
+*.bin
+*.task
+*.tflite
--- a/README.md
+++ b/README.md
@@ -0,0 +1,142 @@
+# Pixel10 AI Server
+
+Turn your Pixel 10 into a free AI API server. This Android app exposes the Tensor G5's on-device AI chip via a REST API, letting any device on your network make AI inference requests — no cloud, no API keys, no costs.
+
+## How It Works
+
+The app runs an HTTP server directly on your phone that accepts OpenAI-compatible API requests. Under the hood, it uses Google's on-device AI stack:
+
+1. **Gemini Nano** (preferred) — The system-provided model via ML Kit Prompt API, hardware-accelerated on the Tensor G5 TPU with a 32K token context window
+2. **MediaPipe LLM** (fallback) — For custom open-weight models like Gemma 3n or Gemma 2B that you supply yourself
+
+All inference runs entirely on-device. Your data never leaves the phone.
+
+## API Endpoints
+
+| Method | Endpoint | Description |
+|--------|----------|-------------|
+| `POST` | `/v1/chat/completions` | Chat completion (OpenAI-compatible) |
+| `POST` | `/v1/completions` | Text completion |
+| `GET` | `/v1/models` | List available models |
+| `GET` | `/health` | Server status and device info |
+
+## Quick Start
+
+### 1. Install and Launch
+
+Build the APK in Android Studio and install on your Pixel 10 (or Pixel 9/8 series).
+
+### 2. Start the Server
+
+Open the app and tap **Start Server**. The app will:
+- Load the AI model (Gemini Nano or your custom model)
+- Start the HTTP server on the configured port (default: 8080)
+- Display the local IP address to connect to
+
+### 3. Make Requests
+
+From any device on the same WiFi network:
+
+```bash
+# Chat completion
+curl http://<phone-ip>:8080/v1/chat/completions \
+  -H "Content-Type: application/json" \
+  -d '{
+    "messages": [
+      {"role": "system", "content": "You are a helpful assistant."},
+      {"role": "user", "content": "What is the Tensor G5 chip?"}
+    ]
+  }'
+
+# Simple completion
+curl http://<phone-ip>:8080/v1/completions \
+  -H "Content-Type: application/json" \
+  -d '{"prompt": "Explain quantum computing in simple terms"}'
+
+# Health check
+curl http://<phone-ip>:8080/health
+
+# List models
+curl http://<phone-ip>:8080/v1/models
+```
+
+### Use with Python OpenAI Library
+
+```python
+from openai import OpenAI
+
+client = OpenAI(
+    base_url="http://<phone-ip>:8080/v1",
+    api_key="not-needed"  # no auth required
+)
+
+response = client.chat.completions.create(
+    model="pixel10-on-device",
+    messages=[
+        {"role": "user", "content": "Hello from my laptop!"}
+    ]
+)
+print(response.choices[0].message.content)
+```
+
+## Using Custom Models (MediaPipe)
+
+If Gemini Nano isn't available on your device, you can use custom models:
+
+1. Download a compatible model (e.g., [Gemma 3n E2B](https://ai.google.dev/edge/mediapipe/solutions/genai/llm_inference/android))
+2. Push to the device:
+   ```bash
+   adb push gemma-3n-E2B.task /data/data/com.pixel10.ai/files/
+   ```
+3. Restart the app — it will auto-detect the model file
+
+Supported model formats: `.task`, `.bin`, `.tflite`
+
+## Supported Devices
+
+- **Pixel 10 / 10 Pro / 10 Pro XL** — Full Tensor G5 TPU acceleration
+- **Pixel 9 series** — Tensor G4 TPU
+- **Pixel 8 series** — Tensor G3 TPU
+- Other Android 12+ devices — MediaPipe backend with custom models
+
+## Requirements
+
+- Android 12 (API 31) or higher
+- WiFi connection (for network access to the API)
+- For Gemini Nano: Pixel device with AICore support
+- For custom models: Compatible model file placed in app directory
+
+## Building
+
+```bash
+# Clone the repo
+git clone <repo-url>
+cd Pixel10-ai
+
+# Open in Android Studio and build, or:
+./gradlew assembleDebug
+
+# Install on connected device
+adb install app/build/outputs/apk/debug/app-debug.apk
+```
+
+## Architecture
+
+```
+com.pixel10.ai/
+├── inference/
+│   ├── OnDeviceModel.kt      # Unified model interface
+│   ├── GeminiNanoModel.kt     # Gemini Nano via ML Kit Prompt API
+│   └── MediaPipeModel.kt     # Custom models via MediaPipe LLM
+├── server/
+│   ├── AIApiServer.kt        # NanoHTTPD-based REST API server
+│   ├── ApiModels.kt          # Request/response data classes
+│   └── ApiServerService.kt   # Foreground service for background operation
+├── ui/
+│   └── MainActivity.kt       # Server controls and status dashboard
+└── Pixel10AIApp.kt           # Application class
+```
+
+## License
+
+MIT
--- a/app/build.gradle.kts
+++ b/app/build.gradle.kts
@@ -0,0 +1,65 @@
+plugins {
+    id("com.android.application")
+    id("org.jetbrains.kotlin.android")
+}
+
+android {
+    namespace = "com.pixel10.ai"
+    compileSdk = 35
+
+    defaultConfig {
+        applicationId = "com.pixel10.ai"
+        minSdk = 31
+        targetSdk = 35
+        versionCode = 1
+        versionName = "1.0.0"
+    }
+
+    buildTypes {
+        release {
+            isMinifyEnabled = false
+            proguardFiles(
+                getDefaultProguardFile("proguard-android-optimize.txt"),
+                "proguard-rules.pro"
+            )
+        }
+    }
+
+    compileOptions {
+        sourceCompatibility = JavaVersion.VERSION_17
+        targetCompatibility = JavaVersion.VERSION_17
+    }
+
+    kotlinOptions {
+        jvmTarget = "17"
+    }
+
+    buildFeatures {
+        viewBinding = true
+    }
+}
+
+dependencies {
+    // AndroidX
+    implementation("androidx.core:core-ktx:1.15.0")
+    implementation("androidx.appcompat:appcompat:1.7.0")
+    implementation("androidx.activity:activity-ktx:1.9.3")
+    implementation("androidx.lifecycle:lifecycle-runtime-ktx:2.8.7")
+    implementation("com.google.android.material:material:1.12.0")
+    implementation("androidx.constraintlayout:constraintlayout:2.2.0")
+
+    // ML Kit GenAI — Gemini Nano via AICore (recommended for Pixel 10)
+    implementation("com.google.mlkit:genai-prompt:1.0.0-beta1")
+
+    // MediaPipe LLM Inference — for custom models (Gemma, etc.)
+    implementation("com.google.mediapipe:tasks-genai:0.10.24")
+
+    // Embedded HTTP server
+    implementation("org.nanohttpd:nanohttpd:2.3.1")
+
+    // JSON
+    implementation("com.google.code.gson:gson:2.11.0")
+
+    // Coroutines
+    implementation("org.jetbrains.kotlinx:kotlinx-coroutines-android:1.9.0")
+}
--- a/app/proguard-rules.pro
+++ b/app/proguard-rules.pro
@@ -0,0 +1,10 @@
+# Pixel10 AI API Server - ProGuard Rules
+
+# Keep NanoHTTPD server
+-keep class fi.iki.elonen.** { *; }
+
+# Keep MediaPipe classes
+-keep class com.google.mediapipe.** { *; }
+
+# Keep Gson serialization models
+-keep class com.pixel10.ai.server.** { *; }
--- a/app/src/main/AndroidManifest.xml
+++ b/app/src/main/AndroidManifest.xml
@@ -0,0 +1,54 @@
+<?xml version="1.0" encoding="utf-8"?>
+<manifest xmlns:android="http://schemas.android.com/apk/res/android">
+
+    <!-- Network permissions for serving API -->
+    <uses-permission android:name="android.permission.INTERNET" />
+    <uses-permission android:name="android.permission.ACCESS_NETWORK_STATE" />
+    <uses-permission android:name="android.permission.ACCESS_WIFI_STATE" />
+
+    <!-- Foreground service to keep server alive -->
+    <uses-permission android:name="android.permission.FOREGROUND_SERVICE" />
+    <uses-permission android:name="android.permission.FOREGROUND_SERVICE_SPECIAL_USE" />
+    <uses-permission android:name="android.permission.POST_NOTIFICATIONS" />
+
+    <!-- Wake lock to prevent CPU sleep while serving -->
+    <uses-permission android:name="android.permission.WAKE_LOCK" />
+
+    <!-- Require device AI features -->
+    <uses-feature android:name="android.hardware.npu" android:required="false" />
+
+    <application
+        android:name=".Pixel10AIApp"
+        android:allowBackup="true"
+        android:icon="@mipmap/ic_launcher"
+        android:label="@string/app_name"
+        android:supportsRtl="true"
+        android:networkSecurityConfig="@xml/network_security_config"
+        android:theme="@style/Theme.Pixel10AI">
+
+        <activity
+            android:name=".ui.MainActivity"
+            android:exported="true"
+            android:launchMode="singleTop">
+            <intent-filter>
+                <action android:name="android.intent.action.MAIN" />
+                <category android:name="android.intent.category.LAUNCHER" />
+            </intent-filter>
+        </activity>
+
+        <service
+            android:name=".server.ApiServerService"
+            android:foregroundServiceType="specialUse"
+            android:exported="false">
+            <property
+                android:name="android.app.PROPERTY_SPECIAL_USE_FGS_SUBTYPE"
+                android:value="on_device_ai_inference_server" />
+        </service>
+
+        <!-- Opt-in to on-device Gemini Nano model download -->
+        <meta-data
+            android:name="com.google.android.aicore.BUNDLED_MODEL"
+            android:value="gemini_nano" />
+
+    </application>
+</manifest>
--- a/app/src/main/java/com/pixel10/ai/Pixel10AIApp.kt
+++ b/app/src/main/java/com/pixel10/ai/Pixel10AIApp.kt
@@ -0,0 +1,29 @@
+package com.pixel10.ai
+
+import android.app.Application
+import android.app.NotificationChannel
+import android.app.NotificationManager
+
+class Pixel10AIApp : Application() {
+
+    override fun onCreate() {
+        super.onCreate()
+        createNotificationChannel()
+    }
+
+    private fun createNotificationChannel() {
+        val channel = NotificationChannel(
+            CHANNEL_ID,
+            getString(R.string.notification_channel_name),
+            NotificationManager.IMPORTANCE_LOW
+        ).apply {
+            description = getString(R.string.notification_channel_desc)
+        }
+        val manager = getSystemService(NotificationManager::class.java)
+        manager.createNotificationChannel(channel)
+    }
+
+    companion object {
+        const val CHANNEL_ID = "pixel10_ai_server"
+    }
+}
--- a/app/src/main/java/com/pixel10/ai/inference/GeminiNanoModel.kt
+++ b/app/src/main/java/com/pixel10/ai/inference/GeminiNanoModel.kt
@@ -0,0 +1,125 @@
+package com.pixel10.ai.inference
+
+import android.content.Context
+import android.util.Log
+import com.google.mlkit.genai.prompt.GenerativeModel
+import com.google.mlkit.genai.prompt.type.Content
+import com.google.mlkit.genai.prompt.type.TextPart
+import com.google.mlkit.genai.prompt.type.content
+import com.google.mlkit.genai.prompt.type.generationConfig
+import kotlinx.coroutines.Dispatchers
+import kotlinx.coroutines.flow.fold
+import kotlinx.coroutines.suspendCancellableCoroutine
+import kotlinx.coroutines.withContext
+import kotlin.coroutines.resume
+import kotlin.coroutines.resumeWithException
+
+/**
+ * Gemini Nano backend via ML Kit Prompt API.
+ *
+ * This runs the system-provided Gemini Nano model on the Pixel 10's Tensor G5
+ * TPU through Android's AICore service. The model is managed by the OS —
+ * no manual download or file management needed.
+ *
+ * Key advantages:
+ *  - Hardware-accelerated on Tensor G5 TPU (2.6x faster than G4)
+ *  - 32,000 token context window on Pixel 10
+ *  - ~3 GB model always resident in RAM for instant inference
+ *  - Fully offline, private — data never leaves the device
+ */
+class GeminiNanoModel private constructor(
+    private val generativeModel: GenerativeModel
+) : OnDeviceModel {
+
+    override val backendName = "Gemini Nano (ML Kit)"
+
+    @Volatile
+    override var isReady: Boolean = true
+        private set
+
+    override suspend fun generate(
+        prompt: String,
+        maxTokens: Int,
+        temperature: Float
+    ): String = withContext(Dispatchers.Default) {
+        try {
+            val request = content { text(prompt) }
+            val response = generativeModel.generateContent(request)
+            response.text ?: ""
+        } catch (e: Exception) {
+            Log.e(TAG, "Gemini Nano generation error", e)
+            throw OnDeviceModel.InferenceException("Gemini Nano generation failed: ${e.message}", e)
+        }
+    }
+
+    override suspend fun generateStreaming(
+        prompt: String,
+        onToken: (String) -> Unit
+    ): String = withContext(Dispatchers.Default) {
+        try {
+            val request = content { text(prompt) }
+            generativeModel.generateContentStream(request)
+                .fold("") { acc, response ->
+                    val chunk = response.text ?: ""
+                    if (chunk.isNotEmpty()) onToken(chunk)
+                    acc + chunk
+                }
+        } catch (e: Exception) {
+            Log.e(TAG, "Gemini Nano streaming error", e)
+            throw OnDeviceModel.InferenceException("Streaming failed: ${e.message}", e)
+        }
+    }
+
+    override fun close() {
+        isReady = false
+        generativeModel.close()
+    }
+
+    companion object {
+        private const val TAG = "GeminiNanoModel"
+
+        suspend fun create(context: Context): GeminiNanoModel = withContext(Dispatchers.IO) {
+            // Check if Gemini Nano is available on this device
+            val model = GenerativeModel.newBuilder()
+                .setContext(context)
+                .build()
+
+            // Verify feature is available — will throw if not supported
+            suspendCancellableCoroutine { continuation ->
+                model.isAvailable()
+                    .addOnSuccessListener { available ->
+                        if (available) {
+                            continuation.resume(Unit)
+                        } else {
+                            continuation.resumeWithException(
+                                OnDeviceModel.InferenceException(
+                                    "Gemini Nano is not available on this device"
+                                )
+                            )
+                        }
+                    }
+                    .addOnFailureListener { e ->
+                        continuation.resumeWithException(
+                            OnDeviceModel.InferenceException(
+                                "Failed to check Gemini Nano availability: ${e.message}", e
+                            )
+                        )
+                    }
+            }
+
+            // Trigger model download if needed
+            suspendCancellableCoroutine { continuation ->
+                model.downloadModel()
+                    .addOnSuccessListener { continuation.resume(Unit) }
+                    .addOnFailureListener { e ->
+                        Log.w(TAG, "Model download issue (may already be available): ${e.message}")
+                        // Don't fail — model might already be cached
+                        continuation.resume(Unit)
+                    }
+            }
+
+            Log.i(TAG, "Gemini Nano model ready via ML Kit Prompt API")
+            GeminiNanoModel(model)
+        }
+    }
+}
--- a/app/src/main/java/com/pixel10/ai/inference/MediaPipeModel.kt
+++ b/app/src/main/java/com/pixel10/ai/inference/MediaPipeModel.kt
@@ -0,0 +1,149 @@
+package com.pixel10.ai.inference
+
+import android.content.Context
+import android.util.Log
+import com.google.mediapipe.tasks.genai.llminference.LlmInference
+import kotlinx.coroutines.Dispatchers
+import kotlinx.coroutines.suspendCancellableCoroutine
+import kotlinx.coroutines.withContext
+import java.io.File
+import kotlin.coroutines.resume
+import kotlin.coroutines.resumeWithException
+
+/**
+ * MediaPipe LLM Inference backend for custom open-weight models.
+ *
+ * Use this to run models like Gemma 3n E2B, Gemma 2B, or other compatible
+ * LLMs that you download and place on the device yourself.
+ *
+ * On Pixel 10, MediaPipe automatically leverages the Tensor G5 GPU/NPU
+ * for accelerated inference.
+ *
+ * To use:
+ *  1. Download a compatible model (e.g. gemma-3n-E2B.task)
+ *  2. Push to device: adb push model.task /data/local/tmp/llm/
+ *     or copy to app files dir via the app
+ */
+class MediaPipeModel private constructor(
+    private val llmInference: LlmInference,
+    private val modelName: String
+) : OnDeviceModel {
+
+    override val backendName = "MediaPipe ($modelName)"
+
+    @Volatile
+    override var isReady: Boolean = true
+        private set
+
+    override suspend fun generate(
+        prompt: String,
+        maxTokens: Int,
+        temperature: Float
+    ): String = withContext(Dispatchers.Default) {
+        try {
+            llmInference.generateResponse(prompt)
+        } catch (e: Exception) {
+            Log.e(TAG, "MediaPipe inference error", e)
+            throw OnDeviceModel.InferenceException("Generation failed: ${e.message}", e)
+        }
+    }
+
+    override suspend fun generateStreaming(
+        prompt: String,
+        onToken: (String) -> Unit
+    ): String = withContext(Dispatchers.Default) {
+        suspendCancellableCoroutine { continuation ->
+            val fullResponse = StringBuilder()
+            try {
+                llmInference.generateResponseAsync(prompt).addResultListener { partialResult, done ->
+                    val chunk = partialResult ?: ""
+                    fullResponse.append(chunk)
+                    onToken(chunk)
+                    if (done) {
+                        continuation.resume(fullResponse.toString())
+                    }
+                }
+            } catch (e: Exception) {
+                Log.e(TAG, "Streaming error", e)
+                continuation.resumeWithException(
+                    OnDeviceModel.InferenceException("Streaming failed: ${e.message}", e)
+                )
+            }
+        }
+    }
+
+    override fun close() {
+        isReady = false
+        llmInference.close()
+    }
+
+    companion object {
+        private const val TAG = "MediaPipeModel"
+
+        private val MODEL_FILENAMES = listOf(
+            "gemma-3n-E2B.task",
+            "gemma-3n-E4B.task",
+            "gemma-2b-it-gpu-int4.bin",
+            "gemini-nano.bin",
+            "model.bin"
+        )
+
+        suspend fun create(context: Context): MediaPipeModel = withContext(Dispatchers.IO) {
+            val modelPath = findModelPath(context)
+                ?: throw OnDeviceModel.InferenceException(
+                    "No MediaPipe model file found.\n" +
+                    "Place a compatible .bin or .task file in:\n" +
+                    "  ${context.filesDir.absolutePath}/\n" +
+                    "Supported: ${MODEL_FILENAMES.joinToString()}"
+                )
+
+            val modelName = File(modelPath).name
+            Log.i(TAG, "Loading MediaPipe model: $modelPath")
+
+            try {
+                val options = LlmInference.LlmInferenceOptions.builder()
+                    .setModelPath(modelPath)
+                    .setMaxTokens(2048)
+                    .setTopK(40)
+                    .setTemperature(0.7f)
+                    .setRandomSeed(42)
+                    .build()
+
+                val inference = LlmInference.createFromOptions(context, options)
+                Log.i(TAG, "MediaPipe model loaded: $modelName")
+                MediaPipeModel(inference, modelName)
+            } catch (e: Exception) {
+                throw OnDeviceModel.InferenceException(
+                    "Failed to load MediaPipe model from $modelPath: ${e.message}", e
+                )
+            }
+        }
+
+        private fun findModelPath(context: Context): String? {
+            // Search standard locations
+            val searchDirs = listOfNotNull(
+                context.filesDir,
+                File(context.filesDir, "models"),
+                context.getExternalFilesDir(null),
+                File("/data/local/tmp/llm")
+            )
+
+            for (dir in searchDirs) {
+                if (!dir.exists()) continue
+                for (name in MODEL_FILENAMES) {
+                    val file = File(dir, name)
+                    if (file.exists()) {
+                        Log.i(TAG, "Found model: ${file.absolutePath}")
+                        return file.absolutePath
+                    }
+                }
+                // Also check for any .task or .bin file
+                dir.listFiles()?.firstOrNull {
+                    it.extension in listOf("task", "bin", "tflite")
+                }?.let { return it.absolutePath }
+            }
+
+            return null
+        }
+    }
+}
--- a/app/src/main/java/com/pixel10/ai/inference/OnDeviceModel.kt
+++ b/app/src/main/java/com/pixel10/ai/inference/OnDeviceModel.kt
@@ -0,0 +1,82 @@
+package com.pixel10.ai.inference
+
+import android.content.Context
+import android.util.Log
+import kotlinx.coroutines.Dispatchers
+import kotlinx.coroutines.withContext
+
+/**
+ * Unified interface for on-device AI inference on the Pixel 10.
+ *
+ * Supports two backends:
+ *  1. **Gemini Nano** via ML Kit Prompt API — uses the system-managed model
+ *     through AICore, accelerated by the Tensor G5 TPU. Zero setup needed
+ *     on supported Pixel devices.
+ *  2. **MediaPipe LLM** — for custom open-weight models (Gemma 2B/3n, etc.)
+ *     that you supply yourself. Place the .bin/.task file in the app's
+ *     files directory.
+ *
+ * The factory method tries Gemini Nano first (preferred), then falls back
+ * to MediaPipe if a local model file is found.
+ */
+interface OnDeviceModel {
+    val backendName: String
+    val isReady: Boolean
+
+    suspend fun generate(
+        prompt: String,
+        maxTokens: Int = 1024,
+        temperature: Float = 0.7f
+    ): String
+
+    suspend fun generateStreaming(
+        prompt: String,
+        onToken: (String) -> Unit
+    ): String
+
+    fun close()
+
+    companion object {
+        private const val TAG = "OnDeviceModel"
+
+        /**
+         * Create the best available on-device model.
+         * Tries Gemini Nano (AICore) first, falls back to MediaPipe.
+         */
+        suspend fun create(context: Context): OnDeviceModel = withContext(Dispatchers.IO) {
+            // Try Gemini Nano via ML Kit Prompt API first
+            try {
+                Log.i(TAG, "Attempting Gemini Nano via ML Kit Prompt API...")
+                val nano = GeminiNanoModel.create(context)
+                Log.i(TAG, "Gemini Nano ready!")
+                return@withContext nano
+            } catch (e: Exception) {
+                Log.w(TAG, "Gemini Nano not available: ${e.message}")
+            }
+
+            // Fall back to MediaPipe with a local model file
+            try {
+                Log.i(TAG, "Attempting MediaPipe LLM with local model...")
+                val mediapipe = MediaPipeModel.create(context)
+                Log.i(TAG, "MediaPipe model ready!")
+                return@withContext mediapipe
+            } catch (e: Exception) {
+                Log.w(TAG, "MediaPipe model not available: ${e.message}")
+            }
+
+            throw InferenceException(
+                "No AI model available.\n\n" +
+                "Option 1: Use a Pixel device with Gemini Nano support " +
+                "(Pixel 10/9/8 series)\n\n" +
+                "Option 2: Place a MediaPipe-compatible model (.bin or .task) in:\n" +
+                "  ${context.filesDir.absolutePath}/\n" +
+                "  Supported: gemma-3n-E2B.task, gemma-2b-it-gpu-int4.bin, etc.\n\n" +
+                "Download models from:\n" +
+                "  https://ai.google.dev/edge/mediapipe/solutions/genai/llm_inference/android"
+            )
+        }
+    }
+
+    class InferenceException(message: String, cause: Throwable? = null) :
+        Exception(message, cause)
+}
--- a/app/src/main/java/com/pixel10/ai/server/AIApiServer.kt
+++ b/app/src/main/java/com/pixel10/ai/server/AIApiServer.kt
@@ -0,0 +1,254 @@
+package com.pixel10.ai.server
+
+import android.os.Build
+import android.util.Log
+import com.google.gson.Gson
+import com.pixel10.ai.inference.OnDeviceModel
+import fi.iki.elonen.NanoHTTPD
+import kotlinx.coroutines.runBlocking
+import java.util.UUID
+import java.util.concurrent.atomic.AtomicLong
+
+/**
+ * Embedded HTTP server that exposes the on-device AI model as a REST API.
+ *
+ * Provides OpenAI-compatible endpoints so existing tools (curl, Python openai
+ * library, etc.) can talk to this phone as if it were a cloud AI endpoint.
+ *
+ * Usage from any device on the same network:
+ *   curl http://<phone-ip>:8080/v1/chat/completions \
+ *     -H "Content-Type: application/json" \
+ *     -d '{"messages":[{"role":"user","content":"Hello!"}]}'
+ */
+class AIApiServer(
+    port: Int,
+    private val model: OnDeviceModel
+) : NanoHTTPD(port) {
+
+    private val gson = Gson()
+    private val startTime = System.currentTimeMillis()
+    val requestCount = AtomicLong(0)
+
+    var onRequestLogged: ((String) -> Unit)? = null
+
+    override fun serve(session: IHTTPSession): Response {
+        val method = session.method
+        val uri = session.uri
+        val count = requestCount.incrementAndGet()
+
+        log("[$count] ${method.name} $uri")
+
+        return try {
+            // Add CORS headers to all responses
+            when {
+                method == Method.OPTIONS -> corsPreflightResponse()
+                uri == "/" || uri == "/health" -> handleHealth()
+                uri == "/v1/models" && method == Method.GET -> handleModels()
+                uri == "/v1/chat/completions" && method == Method.POST -> handleChatCompletions(session)
+                uri == "/v1/completions" && method == Method.POST -> handleCompletions(session)
+                else -> errorResponse(404, "Not found: $uri")
+            }.also { addCorsHeaders(it) }
+        } catch (e: Exception) {
+            Log.e(TAG, "Request error", e)
+            log("ERROR: ${e.message}")
+            errorResponse(500, "Internal server error: ${e.message}")
+                .also { addCorsHeaders(it) }
+        }
+    }
+
+    // ── Endpoint Handlers ──────────────────────────────────────────────
+
+    private fun handleHealth(): Response {
+        val status = ServerStatus(
+            status = if (model.isReady) "ready" else "model_not_loaded",
+            model = model.backendName,
+            device = "${Build.MANUFACTURER} ${Build.MODEL} (${Build.SOC_MODEL})",
+            uptime_seconds = (System.currentTimeMillis() - startTime) / 1000,
+            requests_served = requestCount.get()
+        )
+        return jsonResponse(200, gson.toJson(status))
+    }
+
+    private fun handleModels(): Response {
+        return jsonResponse(200, gson.toJson(ModelList()))
+    }
+
+    private fun handleChatCompletions(session: IHTTPSession): Response {
+        val body = readBody(session)
+        val request = gson.fromJson(body, ChatRequest::class.java)
+
+        if (request.messages.isEmpty()) {
+            return errorResponse(400, "messages array is required and must not be empty")
+        }
+
+        // Build a prompt from the chat messages
+        val prompt = buildChatPrompt(request.messages)
+
+        log("Chat prompt (${request.messages.size} messages, ${prompt.length} chars)")
+
+        if (request.stream) {
+            return handleStreamingResponse(prompt, request)
+        }
+
+        // Synchronous generation
+        val responseText = runBlocking {
+            model.generate(prompt, request.max_tokens, request.temperature)
+        }
+
+        log("Response: ${responseText.take(80)}...")
+
+        val chatResponse = ChatResponse(
+            id = "chatcmpl-${UUID.randomUUID().toString().take(8)}",
+            choices = listOf(
+                Choice(
+                    message = Message(role = "assistant", content = responseText)
+                )
+            ),
+            usage = Usage(
+                prompt_tokens = estimateTokens(prompt),
+                completion_tokens = estimateTokens(responseText),
+                total_tokens = estimateTokens(prompt) + estimateTokens(responseText)
+            )
+        )
+
+        return jsonResponse(200, gson.toJson(chatResponse))
+    }
+
+    private fun handleCompletions(session: IHTTPSession): Response {
+        val body = readBody(session)
+        val request = gson.fromJson(body, ChatRequest::class.java)
+
+        val prompt = request.prompt
+            ?: request.messages.lastOrNull()?.content
+            ?: return errorResponse(400, "prompt or messages is required")
+
+        log("Completion prompt (${prompt.length} chars)")
+
+        val responseText = runBlocking {
+            model.generate(prompt, request.max_tokens, request.temperature)
+        }
+
+        log("Response: ${responseText.take(80)}...")
+
+        val chatResponse = ChatResponse(
+            id = "cmpl-${UUID.randomUUID().toString().take(8)}",
+            choices = listOf(
+                Choice(
+                    message = Message(role = "assistant", content = responseText)
+                )
+            ),
+            usage = Usage(
+                prompt_tokens = estimateTokens(prompt),
+                completion_tokens = estimateTokens(responseText),
+                total_tokens = estimateTokens(prompt) + estimateTokens(responseText)
+            )
+        )
+
+        return jsonResponse(200, gson.toJson(chatResponse))
+    }
+
+    private fun handleStreamingResponse(prompt: String, request: ChatRequest): Response {
+        val id = "chatcmpl-${UUID.randomUUID().toString().take(8)}"
+
+        // For streaming, collect all tokens then return as SSE-formatted response.
+        // NanoHTTPD doesn't natively support chunked streaming in a clean way,
+        // so we buffer and return the full SSE payload.
+        val sseBuilder = StringBuilder()
+
+        // Initial role chunk
+        val roleChunk = StreamChunk(
+            id = id,
+            choices = listOf(StreamChoice(delta = Delta(role = "assistant")))
+        )
+        sseBuilder.append("data: ${gson.toJson(roleChunk)}\n\n")
+
+        val fullResponse = runBlocking {
+            model.generateStreaming(prompt) { token ->
+                val chunk = StreamChunk(
+                    id = id,
+                    choices = listOf(StreamChoice(delta = Delta(content = token)))
+                )
+                sseBuilder.append("data: ${gson.toJson(chunk)}\n\n")
+            }
+        }
+
+        // Final done chunk
+        val doneChunk = StreamChunk(
+            id = id,
+            choices = listOf(StreamChoice(delta = Delta(), finish_reason = "stop"))
+        )
+        sseBuilder.append("data: ${gson.toJson(doneChunk)}\n\n")
+        sseBuilder.append("data: [DONE]\n\n")
+
+        log("Streamed response: ${fullResponse.take(80)}...")
+
+        return newFixedLengthResponse(
+            Response.Status.OK,
+            "text/event-stream",
+            sseBuilder.toString()
+        )
+    }
+
+    // ── Helpers ─────────────────────────────────────────────────────────
+
+    private fun buildChatPrompt(messages: List<Message>): String {
+        val sb = StringBuilder()
+        for (msg in messages) {
+            when (msg.role) {
+                "system" -> sb.append("System: ${msg.content}\n\n")
+                "user" -> sb.append("User: ${msg.content}\n\n")
+                "assistant" -> sb.append("Assistant: ${msg.content}\n\n")
+            }
+        }
+        sb.append("Assistant: ")
+        return sb.toString()
+    }
+
+    private fun readBody(session: IHTTPSession): String {
+        val contentLength = session.headers["content-length"]?.toIntOrNull() ?: 0
+        val buffer = ByteArray(contentLength)
+        session.inputStream.read(buffer, 0, contentLength)
+        return String(buffer)
+    }
+
+    private fun estimateTokens(text: String): Int {
+        // Rough estimate: ~4 characters per token
+        return (text.length / 4).coerceAtLeast(1)
+    }
+
+    private fun jsonResponse(statusCode: Int, json: String): Response {
+        val status = when (statusCode) {
+            200 -> Response.Status.OK
+            400 -> Response.Status.BAD_REQUEST
+            404 -> Response.Status.NOT_FOUND
+            else -> Response.Status.INTERNAL_ERROR
+        }
+        return newFixedLengthResponse(status, "application/json", json)
+    }
+
+    private fun errorResponse(statusCode: Int, message: String): Response {
+        val error = ErrorResponse(
+            ErrorDetail(message = message, code = statusCode)
+        )
+        return jsonResponse(statusCode, gson.toJson(error))
+    }
+
+    private fun corsPreflightResponse(): Response {
+        return newFixedLengthResponse(Response.Status.OK, MIME_PLAINTEXT, "")
+    }
+
+    private fun addCorsHeaders(response: Response) {
+        response.addHeader("Access-Control-Allow-Origin", "*")
+        response.addHeader("Access-Control-Allow-Methods", "GET, POST, OPTIONS")
+        response.addHeader("Access-Control-Allow-Headers", "Content-Type, Authorization")
+    }
+
+    private fun log(message: String) {
+        Log.d(TAG, message)
+        onRequestLogged?.invoke(message)
+    }
+
+    companion object {
+        private const val TAG = "AIApiServer"
+    }
+}
--- a/app/src/main/java/com/pixel10/ai/server/ApiModels.kt
+++ b/app/src/main/java/com/pixel10/ai/server/ApiModels.kt
@@ -0,0 +1,91 @@
+package com.pixel10.ai.server
+
+/**
+ * Request/response models for the AI API.
+ * Follows an OpenAI-compatible schema for easy integration.
+ */
+
+data class ChatRequest(
+    val messages: List<Message> = emptyList(),
+    val prompt: String? = null,
+    val max_tokens: Int = 1024,
+    val temperature: Float = 0.7f,
+    val stream: Boolean = false
+)
+
+data class Message(
+    val role: String = "user",
+    val content: String = ""
+)
+
+data class ChatResponse(
+    val id: String,
+    val model: String = "pixel10-on-device",
+    val choices: List<Choice>,
+    val usage: Usage
+)
+
+data class Choice(
+    val index: Int = 0,
+    val message: Message,
+    val finish_reason: String = "stop"
+)
+
+data class Usage(
+    val prompt_tokens: Int,
+    val completion_tokens: Int,
+    val total_tokens: Int
+)
+
+data class StreamChunk(
+    val id: String,
+    val model: String = "pixel10-on-device",
+    val choices: List<StreamChoice>
+)
+
+data class StreamChoice(
+    val index: Int = 0,
+    val delta: Delta,
+    val finish_reason: String? = null
+)
+
+data class Delta(
+    val role: String? = null,
+    val content: String? = null
+)
+
+data class ModelInfo(
+    val id: String = "pixel10-on-device",
+    val object_type: String = "model",
+    val owned_by: String = "local-device",
+    val description: String = "On-device AI model running on Pixel 10 Tensor G5 chip"
+)
+
+data class ModelList(
+    val data: List<ModelInfo> = listOf(ModelInfo())
+)
+
+data class ErrorResponse(
+    val error: ErrorDetail
+)
+
+data class ErrorDetail(
+    val message: String,
+    val type: String = "server_error",
+    val code: Int = 500
+)
+
+data class ServerStatus(
+    val status: String,
+    val model: String,
+    val device: String,
+    val uptime_seconds: Long,
+    val requests_served: Long,
+    val endpoints: List<String> = listOf(
+        "POST /v1/chat/completions",
+        "POST /v1/completions",
+        "GET  /v1/models",
+        "GET  /health",
+        "GET  /"
+    )
+)
--- a/app/src/main/java/com/pixel10/ai/server/ApiServerService.kt
+++ b/app/src/main/java/com/pixel10/ai/server/ApiServerService.kt
@@ -0,0 +1,165 @@
+package com.pixel10.ai.server
+
+import android.app.Notification
+import android.app.PendingIntent
+import android.app.Service
+import android.content.Intent
+import android.os.Binder
+import android.os.IBinder
+import android.os.PowerManager
+import android.util.Log
+import androidx.core.app.NotificationCompat
+import com.pixel10.ai.Pixel10AIApp
+import com.pixel10.ai.R
+import com.pixel10.ai.inference.OnDeviceModel
+import com.pixel10.ai.ui.MainActivity
+import kotlinx.coroutines.CoroutineScope
+import kotlinx.coroutines.Dispatchers
+import kotlinx.coroutines.SupervisorJob
+import kotlinx.coroutines.cancel
+import kotlinx.coroutines.launch
+
+/**
+ * Foreground service that keeps the AI API server running even when the
+ * app is in the background. Shows a persistent notification with server status.
+ */
+class ApiServerService : Service() {
+
+    private val binder = LocalBinder()
+    private val scope = CoroutineScope(SupervisorJob() + Dispatchers.Main)
+
+    private var server: AIApiServer? = null
+    private var model: OnDeviceModel? = null
+    private var wakeLock: PowerManager.WakeLock? = null
+
+    var onStatusChanged: ((ServerState) -> Unit)? = null
+    var onLog: ((String) -> Unit)? = null
+
+    val isRunning: Boolean get() = server != null
+    val requestCount: Long get() = server?.requestCount?.get() ?: 0
+
+    inner class LocalBinder : Binder() {
+        val service: ApiServerService get() = this@ApiServerService
+    }
+
+    override fun onBind(intent: Intent?): IBinder = binder
+
+    override fun onStartCommand(intent: Intent?, flags: Int, startId: Int): Int {
+        when (intent?.action) {
+            ACTION_START -> {
+                val port = intent.getIntExtra(EXTRA_PORT, DEFAULT_PORT)
+                startServer(port)
+            }
+            ACTION_STOP -> stopServer()
+        }
+        return START_STICKY
+    }
+
+    private fun startServer(port: Int) {
+        if (server != null) return
+
+        startForeground(NOTIFICATION_ID, buildNotification(port))
+        acquireWakeLock()
+        notifyStatus(ServerState.LOADING_MODEL)
+
+        scope.launch {
+            try {
+                // Load the on-device AI model
+                notifyLog("Loading AI model...")
+                model = OnDeviceModel.create(applicationContext)
+                notifyLog("Model ready: ${model!!.backendName}")
+
+                // Start the HTTP server
+                notifyLog("Starting API server on port $port...")
+                val apiServer = AIApiServer(port, model!!)
+                apiServer.onRequestLogged = { msg -> notifyLog(msg) }
+                apiServer.start()
+                server = apiServer
+
+                notifyStatus(ServerState.RUNNING)
+                notifyLog("Server running on port $port")
+                notifyLog("Endpoints:")
+                notifyLog("  POST /v1/chat/completions")
+                notifyLog("  POST /v1/completions")
+                notifyLog("  GET  /v1/models")
+                notifyLog("  GET  /health")
+            } catch (e: Exception) {
+                Log.e(TAG, "Failed to start server", e)
+                notifyLog("ERROR: ${e.message}")
+                notifyStatus(ServerState.ERROR)
+                stopServer()
+            }
+        }
+    }
+
+    fun stopServer() {
+        server?.stop()
+        server = null
+        model?.close()
+        model = null
+        releaseWakeLock()
+        notifyStatus(ServerState.STOPPED)
+        notifyLog("Server stopped")
+        stopForeground(STOP_FOREGROUND_REMOVE)
+        stopSelf()
+    }
+
+    private fun acquireWakeLock() {
+        val pm = getSystemService(POWER_SERVICE) as PowerManager
+        wakeLock = pm.newWakeLock(
+            PowerManager.PARTIAL_WAKE_LOCK,
+            "Pixel10AI::ServerWakeLock"
+        ).apply { acquire(4 * 60 * 60 * 1000L) } // 4 hours max
+    }
+
+    private fun releaseWakeLock() {
+        wakeLock?.let {
+            if (it.isHeld) it.release()
+        }
+        wakeLock = null
+    }
+
+    private fun buildNotification(port: Int): Notification {
+        val pendingIntent = PendingIntent.getActivity(
+            this, 0,
+            Intent(this, MainActivity::class.java),
+            PendingIntent.FLAG_IMMUTABLE
+        )
+
+        return NotificationCompat.Builder(this, Pixel10AIApp.CHANNEL_ID)
+            .setContentTitle(getString(R.string.notification_title))
+            .setContentText(getString(R.string.notification_text, port))
+            .setSmallIcon(android.R.drawable.ic_menu_share)
+            .setContentIntent(pendingIntent)
+            .setOngoing(true)
+            .build()
+    }
+
+    private fun notifyStatus(state: ServerState) {
+        onStatusChanged?.invoke(state)
+    }
+
+    private fun notifyLog(message: String) {
+        Log.d(TAG, message)
+        onLog?.invoke(message)
+    }
+
+    override fun onDestroy() {
+        stopServer()
+        scope.cancel()
+        super.onDestroy()
+    }
+
+    enum class ServerState {
+        STOPPED, LOADING_MODEL, RUNNING, ERROR
+    }
+
+    companion object {
+        private const val TAG = "ApiServerService"
+        const val ACTION_START = "com.pixel10.ai.START_SERVER"
+        const val ACTION_STOP = "com.pixel10.ai.STOP_SERVER"
+        const val EXTRA_PORT = "port"
+        const val DEFAULT_PORT = 8080
+        private const val NOTIFICATION_ID = 1
+    }
+}
--- a/app/src/main/java/com/pixel10/ai/ui/MainActivity.kt
+++ b/app/src/main/java/com/pixel10/ai/ui/MainActivity.kt
@@ -0,0 +1,221 @@
+package com.pixel10.ai.ui
+
+import android.Manifest
+import android.content.ComponentName
+import android.content.Context
+import android.content.Intent
+import android.content.ServiceConnection
+import android.content.pm.PackageManager
+import android.net.wifi.WifiManager
+import android.os.Build
+import android.os.Bundle
+import android.os.IBinder
+import android.view.View
+import androidx.activity.result.contract.ActivityResultContracts
+import androidx.appcompat.app.AppCompatActivity
+import androidx.core.content.ContextCompat
+import com.pixel10.ai.R
+import com.pixel10.ai.databinding.ActivityMainBinding
+import com.pixel10.ai.server.ApiServerService
+import java.text.SimpleDateFormat
+import java.util.Date
+import java.util.Locale
+
+class MainActivity : AppCompatActivity() {
+
+    private lateinit var binding: ActivityMainBinding
+    private var service: ApiServerService? = null
+    private var bound = false
+
+    private val logBuffer = StringBuilder()
+
+    private val notificationPermission = registerForActivityResult(
+        ActivityResultContracts.RequestPermission()
+    ) { /* proceed regardless */ }
+
+    private val serviceConnection = object : ServiceConnection {
+        override fun onServiceConnected(name: ComponentName?, binder: IBinder?) {
+            val localBinder = binder as ApiServerService.LocalBinder
+            service = localBinder.service
+            bound = true
+
+            service?.onStatusChanged = { state ->
+                runOnUiThread { updateStatus(state) }
+            }
+            service?.onLog = { message ->
+                runOnUiThread { appendLog(message) }
+            }
+
+            if (service?.isRunning == true) {
+                updateStatus(ApiServerService.ServerState.RUNNING)
+            }
+        }
+
+        override fun onServiceDisconnected(name: ComponentName?) {
+            service = null
+            bound = false
+        }
+    }
+
+    override fun onCreate(savedInstanceState: Bundle?) {
+        super.onCreate(savedInstanceState)
+        binding = ActivityMainBinding.inflate(layoutInflater)
+        setContentView(binding.root)
+
+        requestNotificationPermission()
+
+        binding.btnToggle.setOnClickListener {
+            if (service?.isRunning == true) {
+                stopServer()
+            } else {
+                startServer()
+            }
+        }
+
+        updateStatus(ApiServerService.ServerState.STOPPED)
+        appendLog("Pixel10 AI Server ready")
+        appendLog("Device: ${Build.MANUFACTURER} ${Build.MODEL}")
+        appendLog("SoC: ${Build.SOC_MODEL}")
+        appendLog("Android ${Build.VERSION.RELEASE} (API ${Build.VERSION.SDK_INT})")
+        appendLog("")
+        appendLog("Tap 'Start Server' to begin serving AI inference")
+    }
+
+    override fun onStart() {
+        super.onStart()
+        Intent(this, ApiServerService::class.java).also { intent ->
+            bindService(intent, serviceConnection, Context.BIND_AUTO_CREATE)
+        }
+    }
+
+    override fun onStop() {
+        super.onStop()
+        if (bound) {
+            service?.onStatusChanged = null
+            service?.onLog = null
+            unbindService(serviceConnection)
+            bound = false
+        }
+    }
+
+    private fun requestNotificationPermission() {
+        if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.TIRAMISU) {
+            if (ContextCompat.checkSelfPermission(
+                    this, Manifest.permission.POST_NOTIFICATIONS
+                ) != PackageManager.PERMISSION_GRANTED
+            ) {
+                notificationPermission.launch(Manifest.permission.POST_NOTIFICATIONS)
+            }
+        }
+    }
+
+    private fun startServer() {
+        val port = binding.etPort.text.toString().toIntOrNull() ?: 8080
+
+        val intent = Intent(this, ApiServerService::class.java).apply {
+            action = ApiServerService.ACTION_START
+            putExtra(ApiServerService.EXTRA_PORT, port)
+        }
+        startForegroundService(intent)
+
+        // Bind if not already bound
+        if (!bound) {
+            bindService(
+                Intent(this, ApiServerService::class.java),
+                serviceConnection,
+                Context.BIND_AUTO_CREATE
+            )
+        }
+
+        updateStatus(ApiServerService.ServerState.LOADING_MODEL)
+    }
+
+    private fun stopServer() {
+        service?.stopServer()
+        updateStatus(ApiServerService.ServerState.STOPPED)
+    }
+
+    private fun updateStatus(state: ApiServerService.ServerState) {
+        when (state) {
+            ApiServerService.ServerState.STOPPED -> {
+                binding.tvServerStatus.text = getString(R.string.server_status_stopped)
+                binding.viewStatusDot.setBackgroundColor(getColor(R.color.status_stopped))
+                binding.tvServerUrl.text = "http://—"
+                binding.tvModelStatus.text = "Model: not loaded"
+                binding.btnToggle.text = getString(R.string.btn_start)
+                binding.btnToggle.isEnabled = true
+                binding.etPort.isEnabled = true
+            }
+            ApiServerService.ServerState.LOADING_MODEL -> {
+                binding.tvServerStatus.text = getString(R.string.server_status_starting)
+                binding.viewStatusDot.setBackgroundColor(getColor(R.color.primary))
+                binding.tvModelStatus.text = getString(R.string.model_loading)
+                binding.btnToggle.isEnabled = false
+                binding.etPort.isEnabled = false
+            }
+            ApiServerService.ServerState.RUNNING -> {
+                val port = binding.etPort.text.toString()
+                val ip = getLocalIpAddress()
+                binding.tvServerStatus.text = getString(R.string.server_status_running)
+                binding.viewStatusDot.setBackgroundColor(getColor(R.color.status_running))
+                binding.tvServerUrl.text = "http://$ip:$port"
+                binding.tvModelStatus.text = getString(R.string.model_ready)
+                binding.btnToggle.text = getString(R.string.btn_stop)
+                binding.btnToggle.isEnabled = true
+                binding.etPort.isEnabled = false
+            }
+            ApiServerService.ServerState.ERROR -> {
+                binding.tvServerStatus.text = getString(R.string.server_status_error)
+                binding.viewStatusDot.setBackgroundColor(getColor(R.color.error))
+                binding.tvModelStatus.text = getString(R.string.model_error)
+                binding.btnToggle.text = getString(R.string.btn_start)
+                binding.btnToggle.isEnabled = true
+                binding.etPort.isEnabled = true
+            }
+        }
+    }
+
+    private fun appendLog(message: String) {
+        val timestamp = SimpleDateFormat("HH:mm:ss", Locale.US).format(Date())
+        logBuffer.append("[$timestamp] $message\n")
+        binding.tvLog.text = logBuffer.toString()
+
+        // Auto-scroll to bottom
+        binding.scrollLog.post {
+            binding.scrollLog.fullScroll(View.FOCUS_DOWN)
+        }
+
+        // Update request count
+        service?.let {
+            binding.tvRequestCount.text = "Requests served: ${it.requestCount}"
+        }
+    }
+
+    @Suppress("DEPRECATION")
+    private fun getLocalIpAddress(): String {
+        try {
+            val wifiManager = applicationContext.getSystemService(WIFI_SERVICE) as WifiManager
+            val ip = wifiManager.connectionInfo.ipAddress
+            if (ip != 0) {
+                return "${ip and 0xFF}.${ip shr 8 and 0xFF}.${ip shr 16 and 0xFF}.${ip shr 24 and 0xFF}"
+            }
+        } catch (_: Exception) {}
+
+        // Fallback: iterate network interfaces
+        try {
+            val interfaces = java.net.NetworkInterface.getNetworkInterfaces()
+            while (interfaces.hasMoreElements()) {
+                val iface = interfaces.nextElement()
+                val addresses = iface.inetAddresses
+                while (addresses.hasMoreElements()) {
+                    val addr = addresses.nextElement()
+                    if (!addr.isLoopbackAddress && addr is java.net.Inet4Address) {
+                        return addr.hostAddress ?: "0.0.0.0"
+                    }
+                }
+            }
+        } catch (_: Exception) {}
+
+        return "0.0.0.0"
+    }
+}
--- a/app/src/main/res/drawable/status_dot.xml
+++ b/app/src/main/res/drawable/status_dot.xml
@@ -0,0 +1,6 @@
+<?xml version="1.0" encoding="utf-8"?>
+<shape xmlns:android="http://schemas.android.com/apk/res/android"
+    android:shape="oval">
+    <solid android:color="@color/status_stopped" />
+    <size android:width="12dp" android:height="12dp" />
+</shape>
--- a/app/src/main/res/layout/activity_main.xml
+++ b/app/src/main/res/layout/activity_main.xml
@@ -0,0 +1,189 @@
+<?xml version="1.0" encoding="utf-8"?>
+<androidx.constraintlayout.widget.ConstraintLayout
+    xmlns:android="http://schemas.android.com/apk/res/android"
+    xmlns:app="http://schemas.android.com/apk/res-auto"
+    android:layout_width="match_parent"
+    android:layout_height="match_parent"
+    android:background="@color/surface"
+    android:padding="24dp">
+
+    <!-- Title -->
+    <TextView
+        android:id="@+id/tvTitle"
+        android:layout_width="wrap_content"
+        android:layout_height="wrap_content"
+        android:text="Pixel10 AI Server"
+        android:textColor="@color/on_surface"
+        android:textSize="28sp"
+        android:textStyle="bold"
+        android:layout_marginTop="16dp"
+        app:layout_constraintTop_toTopOf="parent"
+        app:layout_constraintStart_toStartOf="parent" />
+
+    <TextView
+        android:id="@+id/tvSubtitle"
+        android:layout_width="wrap_content"
+        android:layout_height="wrap_content"
+        android:text="On-device AI inference API"
+        android:textColor="@color/log_text"
+        android:textSize="14sp"
+        android:layout_marginTop="4dp"
+        app:layout_constraintTop_toBottomOf="@id/tvTitle"
+        app:layout_constraintStart_toStartOf="parent" />
+
+    <!-- Status Card -->
+    <com.google.android.material.card.MaterialCardView
+        android:id="@+id/cardStatus"
+        android:layout_width="0dp"
+        android:layout_height="wrap_content"
+        android:layout_marginTop="24dp"
+        app:cardBackgroundColor="@color/surface_variant"
+        app:cardCornerRadius="16dp"
+        app:cardElevation="0dp"
+        app:strokeWidth="0dp"
+        app:layout_constraintTop_toBottomOf="@id/tvSubtitle"
+        app:layout_constraintStart_toStartOf="parent"
+        app:layout_constraintEnd_toEndOf="parent">
+
+        <LinearLayout
+            android:layout_width="match_parent"
+            android:layout_height="wrap_content"
+            android:orientation="vertical"
+            android:padding="20dp">
+
+            <LinearLayout
+                android:layout_width="match_parent"
+                android:layout_height="wrap_content"
+                android:orientation="horizontal"
+                android:gravity="center_vertical">
+
+                <View
+                    android:id="@+id/viewStatusDot"
+                    android:layout_width="12dp"
+                    android:layout_height="12dp"
+                    android:background="@drawable/status_dot" />
+
+                <TextView
+                    android:id="@+id/tvServerStatus"
+                    android:layout_width="wrap_content"
+                    android:layout_height="wrap_content"
+                    android:text="@string/server_status_stopped"
+                    android:textColor="@color/on_surface"
+                    android:textSize="18sp"
+                    android:textStyle="bold"
+                    android:layout_marginStart="12dp" />
+            </LinearLayout>
+
+            <TextView
+                android:id="@+id/tvServerUrl"
+                android:layout_width="wrap_content"
+                android:layout_height="wrap_content"
+                android:text="http://—"
+                android:textColor="@color/log_text"
+                android:textSize="14sp"
+                android:fontFamily="monospace"
+                android:layout_marginTop="8dp"
+                android:textIsSelectable="true" />
+
+            <TextView
+                android:id="@+id/tvModelStatus"
+                android:layout_width="wrap_content"
+                android:layout_height="wrap_content"
+                android:text="Model: not loaded"
+                android:textColor="@color/log_text"
+                android:textSize="13sp"
+                android:layout_marginTop="4dp" />
+
+            <TextView
+                android:id="@+id/tvRequestCount"
+                android:layout_width="wrap_content"
+                android:layout_height="wrap_content"
+                android:text="Requests served: 0"
+                android:textColor="@color/log_text"
+                android:textSize="13sp"
+                android:layout_marginTop="2dp" />
+        </LinearLayout>
+    </com.google.android.material.card.MaterialCardView>
+
+    <!-- Port Config -->
+    <LinearLayout
+        android:id="@+id/layoutPort"
+        android:layout_width="0dp"
+        android:layout_height="wrap_content"
+        android:orientation="horizontal"
+        android:gravity="center_vertical"
+        android:layout_marginTop="16dp"
+        app:layout_constraintTop_toBottomOf="@id/cardStatus"
+        app:layout_constraintStart_toStartOf="parent"
+        app:layout_constraintEnd_toEndOf="parent">
+
+        <TextView
+            android:layout_width="wrap_content"
+            android:layout_height="wrap_content"
+            android:text="Port:"
+            android:textColor="@color/on_surface"
+            android:textSize="16sp" />
+
+        <com.google.android.material.textfield.TextInputEditText
+            android:id="@+id/etPort"
+            android:layout_width="100dp"
+            android:layout_height="48dp"
+            android:layout_marginStart="12dp"
+            android:text="8080"
+            android:inputType="number"
+            android:textColor="@color/on_surface"
+            android:backgroundTint="@color/primary"
+            android:fontFamily="monospace"
+            android:textSize="16sp"
+            android:gravity="center" />
+    </LinearLayout>
+
+    <!-- Start/Stop Button -->
+    <com.google.android.material.button.MaterialButton
+        android:id="@+id/btnToggle"
+        android:layout_width="0dp"
+        android:layout_height="56dp"
+        android:layout_marginTop="16dp"
+        android:text="@string/btn_start"
+        android:textSize="16sp"
+        app:cornerRadius="12dp"
+        app:layout_constraintTop_toBottomOf="@id/layoutPort"
+        app:layout_constraintStart_toStartOf="parent"
+        app:layout_constraintEnd_toEndOf="parent" />
+
+    <!-- Log Output -->
+    <TextView
+        android:id="@+id/tvLogLabel"
+        android:layout_width="wrap_content"
+        android:layout_height="wrap_content"
+        android:text="Request Log"
+        android:textColor="@color/on_surface"
+        android:textSize="14sp"
+        android:textStyle="bold"
+        android:layout_marginTop="20dp"
+        app:layout_constraintTop_toBottomOf="@id/btnToggle"
+        app:layout_constraintStart_toStartOf="parent" />
+
+    <ScrollView
+        android:id="@+id/scrollLog"
+        android:layout_width="0dp"
+        android:layout_height="0dp"
+        android:layout_marginTop="8dp"
+        android:background="@color/log_bg"
+        android:padding="12dp"
+        app:layout_constraintTop_toBottomOf="@id/tvLogLabel"
+        app:layout_constraintBottom_toBottomOf="parent"
+        app:layout_constraintStart_toStartOf="parent"
+        app:layout_constraintEnd_toEndOf="parent">
+
+        <TextView
+            android:id="@+id/tvLog"
+            android:layout_width="match_parent"
+            android:layout_height="wrap_content"
+            android:textColor="@color/log_text"
+            android:textSize="12sp"
+            android:fontFamily="monospace"
+            android:textIsSelectable="true" />
+    </ScrollView>
+
+</androidx.constraintlayout.widget.ConstraintLayout>
--- a/app/src/main/res/mipmap-hdpi/ic_launcher.xml
+++ b/app/src/main/res/mipmap-hdpi/ic_launcher.xml
@@ -0,0 +1,5 @@
+<?xml version="1.0" encoding="utf-8"?>
+<adaptive-icon xmlns:android="http://schemas.android.com/apk/res/android">
+    <background android:drawable="@color/primary" />
+    <foreground android:drawable="@color/on_primary" />
+</adaptive-icon>
--- a/app/src/main/res/values/colors.xml
+++ b/app/src/main/res/values/colors.xml
@@ -0,0 +1,15 @@
+<?xml version="1.0" encoding="utf-8"?>
+<resources>
+    <color name="primary">#1A73E8</color>
+    <color name="primary_container">#D2E3FC</color>
+    <color name="on_primary">#FFFFFF</color>
+    <color name="secondary">#34A853</color>
+    <color name="surface">#0F0F0F</color>
+    <color name="on_surface">#E8EAED</color>
+    <color name="surface_variant">#1A1A2E</color>
+    <color name="error">#EA4335</color>
+    <color name="status_running">#34A853</color>
+    <color name="status_stopped">#9AA0A6</color>
+    <color name="log_bg">#0D1117</color>
+    <color name="log_text">#8B949E</color>
+</resources>
--- a/app/src/main/res/values/strings.xml
+++ b/app/src/main/res/values/strings.xml
@@ -0,0 +1,17 @@
+<?xml version="1.0" encoding="utf-8"?>
+<resources>
+    <string name="app_name">Pixel10 AI Server</string>
+    <string name="server_status_stopped">Server Stopped</string>
+    <string name="server_status_starting">Starting Server…</string>
+    <string name="server_status_running">Server Running</string>
+    <string name="server_status_error">Server Error</string>
+    <string name="btn_start">Start Server</string>
+    <string name="btn_stop">Stop Server</string>
+    <string name="model_loading">Loading AI model…</string>
+    <string name="model_ready">AI model ready</string>
+    <string name="model_error">AI model failed to load</string>
+    <string name="notification_channel_name">AI Server</string>
+    <string name="notification_channel_desc">Pixel10 AI API Server status</string>
+    <string name="notification_title">Pixel10 AI Server</string>
+    <string name="notification_text">Serving AI inference on port %d</string>
+</resources>
--- a/app/src/main/res/values/styles.xml
+++ b/app/src/main/res/values/styles.xml
@@ -0,0 +1,11 @@
+<?xml version="1.0" encoding="utf-8"?>
+<resources>
+    <style name="Theme.Pixel10AI" parent="Theme.Material3.DayNight.NoActionBar">
+        <item name="colorPrimary">@color/primary</item>
+        <item name="colorPrimaryContainer">@color/primary_container</item>
+        <item name="colorOnPrimary">@color/on_primary</item>
+        <item name="colorSecondary">@color/secondary</item>
+        <item name="android:statusBarColor">@color/surface</item>
+        <item name="android:navigationBarColor">@color/surface</item>
+    </style>
+</resources>
--- a/app/src/main/res/xml/network_security_config.xml
+++ b/app/src/main/res/xml/network_security_config.xml
@@ -0,0 +1,4 @@
+<?xml version="1.0" encoding="utf-8"?>
+<network-security-config>
+    <base-config cleartextTrafficPermitted="true" />
+</network-security-config>
--- a/build.gradle.kts
+++ b/build.gradle.kts
@@ -0,0 +1,4 @@
+plugins {
+    id("com.android.application") version "8.7.3" apply false
+    id("org.jetbrains.kotlin.android") version "2.0.21" apply false
+}
--- a/gradle.properties
+++ b/gradle.properties
@@ -0,0 +1,4 @@
+org.gradle.jvmargs=-Xmx2048m -Dfile.encoding=UTF-8
+android.useAndroidX=true
+android.nonTransitiveRClass=true
+kotlin.code.style=official
--- a/gradle/wrapper/gradle-wrapper.properties
+++ b/gradle/wrapper/gradle-wrapper.properties
@@ -0,0 +1,7 @@
+distributionBase=GRADLE_USER_HOME
+distributionPath=wrapper/dists
+distributionUrl=https\://services.gradle.org/distributions/gradle-8.9-bin.zip
+networkTimeout=10000
+validateDistributionUrl=true
+zipStoreBase=GRADLE_USER_HOME
+zipStorePath=wrapper/dists
--- a/settings.gradle.kts
+++ b/settings.gradle.kts
@@ -0,0 +1,17 @@
+pluginManagement {
+    repositories {
+        google()
+        mavenCentral()
+        gradlePluginPortal()
+    }
+}
+
+dependencyResolution {
+    repositories {
+        google()
+        mavenCentral()
+    }
+}
+
+rootProject.name = "Pixel10AI"
+include(":app")