Fix streaming, add thread safety, settings UI, and request indicator

Fixes: - Streaming was bypassed for all requests (system prompt auto-inject made messages.size always > 1). Stream=true now routes directly to generateStreaming before multi-turn check - LiteRTModel: add Mutex to serialize Engine calls (not thread-safe) New features: - Settings card: temperature slider, max tokens, agent system prompt toggle (all saved to SharedPreferences, applied on server start) - Active request indicator: "⚡ Processing request…" shown in status card while inference is running - onActiveRequest callback from AIApiServer → ApiServerService → MainActivity Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
Upgrade to LiteRT-LM runtime for Gemma 3n E4B support
2026-02-28 23:17:41 +01:00 · 2026-02-28 22:49:22 +01:00
9 changed files with 416 additions and 65 deletions
@@ -51,9 +51,12 @@ dependencies {
    // 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.)
+    // MediaPipe LLM Inference — legacy fallback for .task/.bin models
    implementation("com.google.mediapipe:tasks-genai:0.10.24")

+    // LiteRT-LM — primary backend for Gemma 3n .litertlm models
+    implementation("com.google.ai.edge.litertlm:litertlm-android:0.9.0-alpha05")
+
    // Embedded HTTP server
    implementation("org.nanohttpd:nanohttpd:2.3.1")

@@ -0,0 +1,146 @@
+package com.pixel10.ai.inference
+
+import android.content.Context
+import android.util.Log
+import com.google.ai.edge.litertlm.Backend
+import com.google.ai.edge.litertlm.Engine
+import com.google.ai.edge.litertlm.EngineConfig
+import kotlinx.coroutines.Dispatchers
+import kotlinx.coroutines.flow.catch
+import kotlinx.coroutines.sync.Mutex
+import kotlinx.coroutines.sync.withLock
+import kotlinx.coroutines.withContext
+import java.io.File
+
+/**
+ * LiteRT-LM backend for Gemma 3n models (.litertlm format).
+ *
+ * This replaces MediaPipe for the newer Gemma 3n E4B/E2B models which use
+ * the LiteRT-LM runtime. Runs fully on-device using the Tensor G5 GPU.
+ *
+ * Model files must be placed in the app's files directory (see [ModelDownloader]).
+ */
+class LiteRTModel private constructor(
+    private val engine: Engine,
+    private val modelName: String
+) : OnDeviceModel {
+
+    override val backendName = "LiteRT-LM ($modelName)"
+
+    @Volatile
+    override var isReady: Boolean = true
+        private set
+
+    // LiteRT Engine is not thread-safe — serialize all inference calls
+    private val mutex = Mutex()
+
+    override suspend fun generate(
+        prompt: String,
+        maxTokens: Int,
+        temperature: Float
+    ): String = mutex.withLock {
+        withContext(Dispatchers.Default) {
+            val conversation = engine.createConversation()
+            try {
+                conversation.sendMessage(prompt).toString()
+            } catch (e: Exception) {
+                Log.e(TAG, "LiteRT inference error", e)
+                throw OnDeviceModel.InferenceException("Generation failed: ${e.message}", e)
+            } finally {
+                conversation.close()
+            }
+        }
+    }
+
+    override suspend fun generateStreaming(
+        prompt: String,
+        onToken: (String) -> Unit
+    ): String = mutex.withLock {
+        withContext(Dispatchers.Default) {
+            val conversation = engine.createConversation()
+            val sb = StringBuilder()
+            try {
+                conversation.sendMessageAsync(prompt)
+                    .catch { e ->
+                        throw OnDeviceModel.InferenceException("Streaming failed: ${e.message}", e)
+                    }
+                    .collect { message ->
+                        val token = message.toString()
+                        sb.append(token)
+                        onToken(token)
+                    }
+            } finally {
+                conversation.close()
+            }
+            sb.toString()
+        }
+    }
+
+    override fun close() {
+        isReady = false
+        engine.close()
+    }
+
+    companion object {
+        private const val TAG = "LiteRTModel"
+
+        private val MODEL_EXTENSIONS = listOf("litertlm")
+
+        suspend fun create(context: Context): LiteRTModel = withContext(Dispatchers.IO) {
+            val modelPath = findModelPath(context)
+                ?: throw OnDeviceModel.InferenceException(
+                    "No LiteRT-LM model file found.\n" +
+                    "Download a .litertlm model via the app or place one in:\n" +
+                    "  ${context.filesDir.absolutePath}/"
+                )
+
+            val modelName = File(modelPath).name
+            Log.i(TAG, "Loading LiteRT-LM model: $modelPath")
+
+            try {
+                val config = EngineConfig(
+                    modelPath = modelPath,
+                    backend = Backend.GPU
+                )
+                val engine = Engine(config)
+                withContext(Dispatchers.Default) {
+                    engine.initialize()
+                }
+                Log.i(TAG, "LiteRT-LM model loaded: $modelName")
+                LiteRTModel(engine, modelName)
+            } catch (gpuError: Exception) {
+                Log.w(TAG, "GPU backend failed, trying CPU: ${gpuError.message}")
+                try {
+                    val config = EngineConfig(
+                        modelPath = modelPath,
+                        backend = Backend.CPU
+                    )
+                    val engine = Engine(config)
+                    withContext(Dispatchers.Default) {
+                        engine.initialize()
+                    }
+                    Log.i(TAG, "LiteRT-LM model loaded on CPU: $modelName")
+                    LiteRTModel(engine, modelName)
+                } catch (e: Exception) {
+                    throw OnDeviceModel.InferenceException(
+                        "Failed to load LiteRT-LM model from $modelPath: ${e.message}", e
+                    )
+                }
+            }
+        }
+
+        private fun findModelPath(context: Context): String? {
+            val searchDirs = listOfNotNull(
+                context.filesDir,
+                File(context.filesDir, "models"),
+                context.getExternalFilesDir(null)
+            )
+            for (dir in searchDirs) {
+                if (!dir.exists()) continue
+                dir.listFiles()?.firstOrNull { it.extension in MODEL_EXTENSIONS }
+                    ?.let { return it.absolutePath }
+            }
+            return null
+        }
+    }
+}
@@ -23,7 +23,7 @@ object ModelDownloader {
    private const val TAG = "ModelDownloader"
    private const val HF_BASE = "https://huggingface.co"

-    /** Available model specs downloadable from HuggingFace. */
+    /** Available model specs downloadable from HuggingFace (requires token + license acceptance). */
    enum class ModelSpec(
        val displayName: String,
        val filename: String,
@@ -31,21 +31,27 @@ object ModelDownloader {
        val sizeMb: Int,
        val description: String
    ) {
-        /** Recommended: best size/quality trade-off, runs fast on Tensor G5. */
-        GEMMA_3_1B_Q4(
-            displayName = "Gemma 3 1B IT (Q4)",
-            filename = "gemma3-1b-it-int4.task",
-            repo = "litert-community/Gemma3-1B-IT",
-            sizeMb = 555,
-            description = "Best balance — fast & capable (~555 MB)"
+        /**
+         * Gemma 3n E4B INT4 — best quality, Tensor G5 optimised, background-safe.
+         * Accept license at: https://huggingface.co/google/gemma-3n-E4B-it-litert-lm
+         */
+        GEMMA_3N_E4B(
+            displayName = "Gemma 3n E4B",
+            filename = "gemma-3n-E4B-it-int4.litertlm",
+            repo = "google/gemma-3n-E4B-it-litert-lm",
+            sizeMb = 4920,
+            description = "Best quality — Tensor G5 optimised (~4.9 GB)"
        ),
-        /** Higher quality, slower. Good for complex reasoning. */
-        GEMMA_3_1B_Q8(
-            displayName = "Gemma 3 1B IT (Q8)",
-            filename = "gemma3-1b-it-int8-web.task",
-            repo = "litert-community/Gemma3-1B-IT",
-            sizeMb = 1010,
-            description = "Higher quality, slower (~1 GB)"
+        /**
+         * Gemma 3n E4B Web INT4 — smaller variant, slightly lower quality.
+         * Same license as above.
+         */
+        GEMMA_3N_E4B_WEB(
+            displayName = "Gemma 3n E4B (Web)",
+            filename = "gemma-3n-E4B-it-int4-Web.litertlm",
+            repo = "google/gemma-3n-E4B-it-litert-lm",
+            sizeMb = 4280,
+            description = "Slightly smaller variant (~4.3 GB)"
        )
    }

@@ -66,10 +72,10 @@ object ModelDownloader {
    fun modelFile(context: Context, spec: ModelSpec): File =
        File(context.filesDir, spec.filename)

-    /** Legacy compat — returns the file of the installed model, or Q4 path as default. */
+    /** Returns the file of the installed model, or E4B path as default. */
    fun modelFile(context: Context): File =
        installedSpec(context)?.let { modelFile(context, it) }
-            ?: modelFile(context, ModelSpec.GEMMA_3_1B_Q4)
+            ?: modelFile(context, ModelSpec.GEMMA_3N_E4B)

    /**
     * Download [spec] from HuggingFace, using [hfToken] for authentication.
@@ -79,7 +85,7 @@ object ModelDownloader {
     */
    suspend fun download(
        context: Context,
-        spec: ModelSpec = ModelSpec.GEMMA_3_1B_Q4,
+        spec: ModelSpec = ModelSpec.GEMMA_3N_E4B,
        hfToken: String,
        onProgress: (Progress) -> Unit
    ) = withContext(Dispatchers.IO) {
@@ -112,9 +112,19 @@ interface OnDeviceModel {
         * Tap "Download Model" in the app UI to get the MediaPipe model automatically.
         */
        suspend fun create(context: Context): OnDeviceModel = withContext(Dispatchers.IO) {
-            // MediaPipe first — background-safe, GPU-accelerated via Tensor G5
+            // LiteRT-LM first — Gemma 3n .litertlm format, GPU-accelerated, background-safe
            try {
-                Log.i(TAG, "Attempting MediaPipe LLM with local model...")
+                Log.i(TAG, "Attempting LiteRT-LM with local .litertlm model...")
+                val litert = LiteRTModel.create(context)
+                Log.i(TAG, "LiteRT-LM model ready: ${litert.backendName}")
+                return@withContext litert
+            } catch (e: Exception) {
+                Log.w(TAG, "LiteRT-LM not available: ${e.message}")
+            }
+
+            // MediaPipe fallback — .task/.bin format, background-safe
+            try {
+                Log.i(TAG, "Attempting MediaPipe LLM with local .task model...")
                val mediapipe = MediaPipeModel.create(context)
                Log.i(TAG, "MediaPipe model ready: ${mediapipe.backendName}")
                return@withContext mediapipe
@@ -122,7 +132,7 @@ interface OnDeviceModel {
                Log.w(TAG, "MediaPipe not available: ${e.message}")
            }

-            // Gemini Nano fallback — only works when app is in foreground
+            // Gemini Nano last resort — foreground only
            try {
                Log.i(TAG, "Attempting Gemini Nano via ML Kit (foreground only)...")
                val nano = GeminiNanoModel.create(context)
@@ -134,11 +144,10 @@ interface OnDeviceModel {

            throw InferenceException(
                "No model loaded yet.\n\n" +
-                "Tap 'Download Model' in the app to download Gemma 2B (~1.3 GB).\n" +
+                "Tap 'Download Model' in the app to download Gemma 3n E4B.\n" +
                "Once downloaded the server works fully in the background.\n\n" +
-                "Or place a compatible model file in:\n" +
-                "  ${context.filesDir.absolutePath}/\n" +
-                "  Supported: gemma-2b-it-gpu-int4.bin, gemma-3n-E2B.task, etc."
+                "Or place a .litertlm file in:\n" +
+                "  ${context.filesDir.absolutePath}/"
            )
        }
    }
@@ -27,9 +27,16 @@ import java.util.concurrent.atomic.AtomicLong
 *     -H "Content-Type: application/json" \
 *     -d '{"messages":[{"role":"user","content":"Hello!"}]}'
 */
+data class ServerConfig(
+    val defaultTemperature: Float = 0.7f,
+    val defaultMaxTokens: Int = 1024,
+    val autoSystemPrompt: Boolean = true
+)
+
 class AIApiServer(
    port: Int,
-    private val model: OnDeviceModel
+    private val model: OnDeviceModel,
+    private val config: ServerConfig = ServerConfig()
 ) : NanoHTTPD(port) {

    private val gson = Gson()
@@ -37,6 +44,7 @@ class AIApiServer(
    val requestCount = AtomicLong(0)

    var onRequestLogged: ((String) -> Unit)? = null
+    var onActiveRequest: ((Boolean) -> Unit)? = null

    override fun serve(session: IHTTPSession): Response {
        val method = session.method
@@ -113,31 +121,48 @@ class AIApiServer(
            return errorResponse(400, "messages array is required and must not be empty")
        }

-        // Auto-inject agent system prompt if the conversation has no system message.
-        // Auto-inject default tools if the request provides none.
-        // This makes the server zero-config as a coding agent for any OpenAI-compatible client.
-        val messages = if (raw.messages.none { it.role == "system" }) {
+        // Auto-inject agent system prompt if enabled and no system message present
+        val messages = if (config.autoSystemPrompt && raw.messages.none { it.role == "system" }) {
            listOf(Message(role = "system", content = AgentConfig.SYSTEM_PROMPT)) + raw.messages
        } else {
            raw.messages
        }
        val request = raw.copy(
            messages = messages,
-            tools = raw.tools.takeUnless { it.isNullOrEmpty() } ?: AgentConfig.DEFAULT_TOOLS
+            tools = raw.tools.takeUnless { it.isNullOrEmpty() }
+                ?: if (config.autoSystemPrompt) AgentConfig.DEFAULT_TOOLS else null,
+            temperature = if (raw.temperature == 0.7f) config.defaultTemperature else raw.temperature,
+            max_tokens = if (raw.max_tokens == 8192) config.defaultMaxTokens else raw.max_tokens
        )

        val id = "chatcmpl-${UUID.randomUUID().toString().take(8)}"
        val hasTools = !request.tools.isNullOrEmpty()

-        log("Chat: ${request.messages.size} messages, tools=${request.tools?.size ?: 0}, stream=${request.stream}")
+        log("Chat: ${request.messages.size} messages, tools=${request.tools?.size ?: 0}, stream=${request.stream}, temp=${request.temperature}")
+
+        // ── Streaming — always uses flat prompt + generateStreaming ────────────
+        if (request.stream) {
+            val prompt = buildFlatPrompt(request.messages)
+            onActiveRequest?.invoke(true)
+            return try {
+                handleStreamingResponse(id, prompt, request)
+            } finally {
+                onActiveRequest?.invoke(false)
+            }
+        }

        // ── Tool calling / multi-turn chat ─────────────────────────────────────
        if (hasTools || request.messages.size > 1 || request.messages.any { it.role == "system" }) {
            val convMessages = request.messages.map { it.toConvMessage() }
            val toolDefs = request.tools?.map { it.toToolDef() } ?: emptyList()

-            val result = runBlocking {
-                model.chat(convMessages, toolDefs, request.max_tokens, request.temperature)
+            onActiveRequest?.invoke(true)
+            val result = try {
+                runBlocking {
+                    model.chat(convMessages, toolDefs, request.max_tokens, request.temperature)
+                }
+            } finally {
+                onActiveRequest?.invoke(false)
            }

            if (result.toolCalls != null) {
@@ -34,6 +34,7 @@ class ApiServerService : Service() {

    var onStatusChanged: ((ServerState) -> Unit)? = null
    var onLog: ((String) -> Unit)? = null
+    var onActiveRequest: ((Boolean) -> Unit)? = null

    val isRunning: Boolean get() = server != null
    val requestCount: Long get() = server?.requestCount?.get() ?: 0
@@ -70,8 +71,15 @@ class ApiServerService : Service() {

                // Start the HTTP server
                notifyLog("Starting API server on port $port...")
-                val apiServer = AIApiServer(port, model!!)
+                val prefs = getSharedPreferences("pixel10_prefs", MODE_PRIVATE)
+                val serverConfig = ServerConfig(
+                    defaultTemperature = prefs.getFloat("temperature", 0.7f),
+                    defaultMaxTokens = prefs.getInt("max_tokens", 1024),
+                    autoSystemPrompt = prefs.getBoolean("auto_system_prompt", true)
+                )
+                val apiServer = AIApiServer(port, model!!, serverConfig)
                apiServer.onRequestLogged = { msg -> notifyLog(msg) }
+                apiServer.onActiveRequest = { active -> onActiveRequest?.invoke(active) }
                apiServer.start()
                server = apiServer

@@ -13,6 +13,7 @@ import android.os.Build
 import android.os.Bundle
 import android.os.IBinder
 import android.view.View
+import android.widget.SeekBar
 import androidx.activity.result.contract.ActivityResultContracts
 import androidx.appcompat.app.AppCompatActivity
 import androidx.core.content.ContextCompat
@@ -53,6 +54,16 @@ class MainActivity : AppCompatActivity() {
            service?.onLog = { message ->
                runOnUiThread { appendLog(message) }
            }
+            service?.onActiveRequest = { active ->
+                runOnUiThread {
+                    if (active) {
+                        binding.tvActiveRequest.text = "⚡ Processing request…"
+                        binding.tvActiveRequest.visibility = View.VISIBLE
+                    } else {
+                        binding.tvActiveRequest.visibility = View.GONE
+                    }
+                }
+            }

            if (service?.isRunning == true) {
                updateStatus(ApiServerService.ServerState.RUNNING)
@@ -73,8 +84,21 @@ class MainActivity : AppCompatActivity() {
        prefs = getSharedPreferences("pixel10_prefs", MODE_PRIVATE)
        requestNotificationPermission()

-        // Restore saved HF token
+        // Restore saved settings
        binding.etHfToken.setText(prefs.getString("hf_token", ""))
+        val savedTemp = (prefs.getFloat("temperature", 0.7f) * 100).toInt()
+        binding.seekTemperature.progress = savedTemp
+        binding.tvTemperatureValue.text = "%.1f".format(savedTemp / 100f)
+        binding.etMaxTokens.setText(prefs.getInt("max_tokens", 1024).toString())
+        binding.switchSystemPrompt.isChecked = prefs.getBoolean("auto_system_prompt", true)
+
+        binding.seekTemperature.setOnSeekBarChangeListener(object : SeekBar.OnSeekBarChangeListener {
+            override fun onProgressChanged(seekBar: SeekBar, progress: Int, fromUser: Boolean) {
+                binding.tvTemperatureValue.text = "%.1f".format(progress / 100f)
+            }
+            override fun onStartTrackingTouch(seekBar: SeekBar) {}
+            override fun onStopTrackingTouch(seekBar: SeekBar) {}
+        })

        binding.btnToggle.setOnClickListener {
            if (service?.isRunning == true) stopServer() else startServer()
@@ -82,11 +106,11 @@ class MainActivity : AppCompatActivity() {

        binding.btnDownloadModel.setOnClickListener {
            saveHfToken()
-            startModelDownload(ModelSpec.GEMMA_3_1B_Q4)
+            startModelDownload(ModelSpec.GEMMA_3N_E4B)
        }
        binding.btnDownloadGemma3Q8.setOnClickListener {
            saveHfToken()
-            startModelDownload(ModelSpec.GEMMA_3_1B_Q8)
+            startModelDownload(ModelSpec.GEMMA_3N_E4B_WEB)
        }

        updateModelCard()
@@ -201,7 +225,16 @@ class MainActivity : AppCompatActivity() {
        }
    }

+    private fun saveSettings() {
+        prefs.edit()
+            .putFloat("temperature", binding.seekTemperature.progress / 100f)
+            .putInt("max_tokens", binding.etMaxTokens.text.toString().toIntOrNull() ?: 1024)
+            .putBoolean("auto_system_prompt", binding.switchSystemPrompt.isChecked)
+            .apply()
+    }
+
    private fun startServer() {
+        saveSettings()
        val port = binding.etPort.text.toString().toIntOrNull() ?: 8080
        val intent = Intent(this, ApiServerService::class.java).apply {
            action = ApiServerService.ACTION_START
@@ -227,6 +260,9 @@ class MainActivity : AppCompatActivity() {
        val canEdit = state == ApiServerService.ServerState.STOPPED ||
                      state == ApiServerService.ServerState.ERROR
        binding.etPort.isEnabled = canEdit
+        binding.seekTemperature.isEnabled = canEdit
+        binding.etMaxTokens.isEnabled = canEdit
+        binding.switchSystemPrompt.isEnabled = canEdit

        when (state) {
            ApiServerService.ServerState.STOPPED -> {
@@ -102,6 +102,16 @@
                android:textColor="@color/log_text"
                android:textSize="13sp"
                android:layout_marginTop="2dp" />
+
+            <TextView
+                android:id="@+id/tvActiveRequest"
+                android:layout_width="wrap_content"
+                android:layout_height="wrap_content"
+                android:textColor="@color/primary"
+                android:textSize="13sp"
+                android:textStyle="bold"
+                android:layout_marginTop="2dp"
+                android:visibility="gone" />
        </LinearLayout>
    </com.google.android.material.card.MaterialCardView>

@@ -176,38 +186,143 @@
        </LinearLayout>
    </com.google.android.material.card.MaterialCardView>

-    <!-- Port Config -->
-    <LinearLayout
+    <!-- Settings Card -->
+    <com.google.android.material.card.MaterialCardView
        android:id="@+id/layoutPort"
        android:layout_width="0dp"
        android:layout_height="wrap_content"
-        android:orientation="horizontal"
-        android:gravity="center_vertical"
-        android:layout_marginTop="16dp"
+        android:layout_marginTop="12dp"
+        app:cardBackgroundColor="@color/surface_variant"
+        app:cardCornerRadius="16dp"
+        app:cardElevation="0dp"
+        app:strokeWidth="0dp"
        app:layout_constraintTop_toBottomOf="@id/cardModel"
        app:layout_constraintStart_toStartOf="parent"
        app:layout_constraintEnd_toEndOf="parent">

-        <TextView
-            android:layout_width="wrap_content"
+        <LinearLayout
+            android:layout_width="match_parent"
            android:layout_height="wrap_content"
-            android:text="@string/port_label"
-            android:textColor="@color/on_surface"
-            android:textSize="16sp" />
+            android:orientation="vertical"
+            android:padding="16dp">

-        <com.google.android.material.textfield.TextInputEditText
-            android:id="@+id/etPort"
-            android:layout_width="100dp"
-            android:layout_height="48dp"
-            android:layout_marginStart="12dp"
-            android:text="@string/port_default"
-            android:inputType="number"
-            android:textColor="@color/on_surface"
-            android:backgroundTint="@color/primary"
-            android:fontFamily="monospace"
-            android:textSize="16sp"
-            android:gravity="center" />
-    </LinearLayout>
+            <!-- Port row -->
+            <LinearLayout
+                android:layout_width="match_parent"
+                android:layout_height="wrap_content"
+                android:orientation="horizontal"
+                android:gravity="center_vertical">
+
+                <TextView
+                    android:layout_width="0dp"
+                    android:layout_height="wrap_content"
+                    android:layout_weight="1"
+                    android:text="@string/port_label"
+                    android:textColor="@color/on_surface"
+                    android:textSize="14sp" />
+
+                <com.google.android.material.textfield.TextInputEditText
+                    android:id="@+id/etPort"
+                    android:layout_width="80dp"
+                    android:layout_height="40dp"
+                    android:text="@string/port_default"
+                    android:inputType="number"
+                    android:textColor="@color/on_surface"
+                    android:backgroundTint="@color/primary"
+                    android:fontFamily="monospace"
+                    android:textSize="14sp"
+                    android:gravity="center" />
+            </LinearLayout>
+
+            <!-- Temperature row -->
+            <LinearLayout
+                android:layout_width="match_parent"
+                android:layout_height="wrap_content"
+                android:orientation="horizontal"
+                android:gravity="center_vertical"
+                android:layout_marginTop="12dp">
+
+                <TextView
+                    android:layout_width="0dp"
+                    android:layout_height="wrap_content"
+                    android:layout_weight="1"
+                    android:text="@string/setting_temperature"
+                    android:textColor="@color/on_surface"
+                    android:textSize="14sp" />
+
+                <TextView
+                    android:id="@+id/tvTemperatureValue"
+                    android:layout_width="36dp"
+                    android:layout_height="wrap_content"
+                    android:text="0.7"
+                    android:textColor="@color/log_text"
+                    android:textSize="13sp"
+                    android:fontFamily="monospace"
+                    android:gravity="end" />
+
+                <SeekBar
+                    android:id="@+id/seekTemperature"
+                    android:layout_width="120dp"
+                    android:layout_height="wrap_content"
+                    android:layout_marginStart="8dp"
+                    android:max="100"
+                    android:progress="70" />
+            </LinearLayout>
+
+            <!-- Max tokens row -->
+            <LinearLayout
+                android:layout_width="match_parent"
+                android:layout_height="wrap_content"
+                android:orientation="horizontal"
+                android:gravity="center_vertical"
+                android:layout_marginTop="8dp">
+
+                <TextView
+                    android:layout_width="0dp"
+                    android:layout_height="wrap_content"
+                    android:layout_weight="1"
+                    android:text="@string/setting_max_tokens"
+                    android:textColor="@color/on_surface"
+                    android:textSize="14sp" />
+
+                <com.google.android.material.textfield.TextInputEditText
+                    android:id="@+id/etMaxTokens"
+                    android:layout_width="80dp"
+                    android:layout_height="40dp"
+                    android:text="1024"
+                    android:inputType="number"
+                    android:textColor="@color/on_surface"
+                    android:backgroundTint="@color/primary"
+                    android:fontFamily="monospace"
+                    android:textSize="14sp"
+                    android:gravity="center" />
+            </LinearLayout>
+
+            <!-- System prompt toggle -->
+            <LinearLayout
+                android:layout_width="match_parent"
+                android:layout_height="wrap_content"
+                android:orientation="horizontal"
+                android:gravity="center_vertical"
+                android:layout_marginTop="8dp">
+
+                <TextView
+                    android:layout_width="0dp"
+                    android:layout_height="wrap_content"
+                    android:layout_weight="1"
+                    android:text="@string/setting_auto_system_prompt"
+                    android:textColor="@color/on_surface"
+                    android:textSize="14sp" />
+
+                <com.google.android.material.switchmaterial.SwitchMaterial
+                    android:id="@+id/switchSystemPrompt"
+                    android:layout_width="wrap_content"
+                    android:layout_height="wrap_content"
+                    android:checked="true" />
+            </LinearLayout>
+
+        </LinearLayout>
+    </com.google.android.material.card.MaterialCardView>

    <!-- Start/Stop Button -->
    <com.google.android.material.button.MaterialButton
@@ -19,13 +19,16 @@

    <!-- Controls -->
    <string name="hf_token_hint">HuggingFace token (huggingface.co/settings/tokens)</string>
-    <string name="btn_download_gemma3_q4">⭐ Gemma 3 1B IT Q4 — Fast (~555 MB)</string>
-    <string name="btn_download_gemma3_q8">Gemma 3 1B IT Q8 — Higher quality (~1 GB)</string>
+    <string name="btn_download_gemma3_q4">⭐ Gemma 3n E4B — Best quality (~4.9 GB)</string>
+    <string name="btn_download_gemma3_q8">Gemma 3n E4B Web — Smaller (~4.3 GB)</string>
    <string name="model_downloaded">✓ %s ready — background inference enabled</string>
    <string name="model_not_downloaded">No local model. Enter HuggingFace token and download.</string>
    <string name="btn_start">Start Server</string>
    <string name="btn_stop">Stop Server</string>
-    <string name="port_label">Port:</string>
+    <string name="port_label">Port</string>
+    <string name="setting_temperature">Temperature</string>
+    <string name="setting_max_tokens">Max tokens</string>
+    <string name="setting_auto_system_prompt">Agent system prompt</string>
    <string name="port_default">8080</string>
    <string name="requests_served">Requests served: %d</string>
    <string name="request_log_label">Request Log</string>