Updated llama.cpp to b6c83aa and overhauled the demo app UI/UX.

* Updated llama.cpp to b6c83aad55a4ce17ec96fced7770cd1be8758193
* Refactored `ModelZoo` with new lightweight models (SmolLM2, Qwen2.5, DeepSeek-R1 Distill) and optimized inference configurations
* Added support for `defaultSystemPrompt` per model in `Model` data class
* Improved `ChatScreen` with expandable thinking blocks for reasoning models and "keep screen on" flag
* Implemented custom navigation transitions (fade, scale, slide, parallax) in `AppTransitions`
* Refactored `RootContainer` status bar with a dedicated loading progress bar and simplified state display
* Modularized `ConversationsScreen` and `ChatScreen` into smaller, reusable Composable components
* Added new utility methods to `ResourceState` for filtering and mapping loading/failure states
* Renamed various drawable resources to follow standard naming conventions (removed `px` suffix)

Author: whyisitworking

app/src/main/java/com/suhel/llamabro/demo/model/Model.kt

@@ -6,8 +6,9 @@ import com.suhel.llamabro.sdk.model.PromptFormat
 data class Model(
     val id: String,
     val name: String,
-    val description: String?,
+    val description: String? = null,
     val downloadUrl: String,
     val promptFormat: PromptFormat,
     val defaultInferenceConfig: InferenceConfig,
+    val defaultSystemPrompt: String? = null
 )

app/src/main/java/com/suhel/llamabro/demo/model/ModelZoo.kt

@@ -5,46 +5,94 @@ import com.suhel.llamabro.sdk.model.PromptFormats
 
 val ModelZoo = listOf(
     Model(
-        id = "gemma-3n-2b",
-        name = "Gemma 3n 2B",
-        description = "Google’s Gemma 3n multimodal model handles image, audio, video, and text inputs. Available in 2B and 4B sizes, it supports 140 languages for text and multimodal tasks.",
-        downloadUrl = "https://huggingface.co/unsloth/gemma-3n-E2B-it-GGUF/resolve/main/gemma-3n-E2B-it-Q4_K_M.gguf",
-        promptFormat = PromptFormats.Gemma3,
+        id = "smollm2-135m-instruct",
+        name = "SmolLM2 135M Instruct",
+        description = "Ultra-lightweight model for absolute maximum tokens-per-second. Perfect for baseline speed tests.",
+        downloadUrl = "https://huggingface.co/unsloth/SmolLM2-135M-Instruct-GGUF/resolve/main/SmolLM2-135M-Instruct-F16.gguf",
+        promptFormat = PromptFormats.ChatML,
         defaultInferenceConfig = InferenceConfig(
-            temperature = 1.0f,
-            topK = 64,
-            minP = 0.01f,
-            topP = 0.95f,
+            temperature = 0.6f,
+            repeatPenalty = 1.15f,
+            presencePenalty = 0.15f,
+            minP = 0.05f,
+            topP = 0.9f,
+            topK = 40,
         )
     ),
-
     Model(
-        id = "qwen3.5-4b",
-        name = "Qwen3.5 4B",
-        description = "Qwen3.5 is Alibaba’s new model family. The multimodal hybrid reasoning LLMs deliver the strongest performances for their sizes",
-        downloadUrl = "https://huggingface.co/unsloth/Qwen3.5-4B-GGUF/resolve/main/Qwen3.5-4B-Q4_K_M.gguf",
+        id = "smollm2-360m-instruct",
+        name = "SmolLM2 360M Instruct",
+        description = "Highly efficient sub-0.5B model balancing sheer speed with improved coherence.",
+        downloadUrl = "https://huggingface.co/HuggingFaceTB/SmolLM2-360M-Instruct-GGUF/resolve/main/smollm2-360m-instruct-q8_0.gguf",
         promptFormat = PromptFormats.ChatML,
         defaultInferenceConfig = InferenceConfig(
-            temperature = 1.0f,
-            topP = 0.95f,
-            topK = 20,
-            minP = 0.0f,
-            presencePenalty = 1.5f,
-            repeatPenalty = 1.0f,
+            temperature = 0.7f,
+            repeatPenalty = 1.15f,
+            presencePenalty = 0.15f,
+            minP = 0.05f,
+            topP = 0.9f,
+            topK = 40,
+        )
+    ),
+    Model(
+        id = "qwen2.5-0.5b-instruct",
+        name = "Qwen2.5 0.5B Instruct",
+        description = "Exceptional speed with strong multilingual support and structured JSON formatting capabilities.",
+        downloadUrl = "https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct-GGUF/resolve/main/qwen2.5-0.5b-instruct-q8_0.gguf",
+        promptFormat = PromptFormats.ChatML,
+        defaultInferenceConfig = InferenceConfig(
+            temperature = 0.7f,
+            repeatPenalty = 1.05f,
+            presencePenalty = 0.15f,
+            minP = 0.1f,
+            topP = 0.8f,
+            topK = 40,
         )
     ),
-
     Model(
         id = "llama-3.2-1b-instruct",
-        name = "Llama 3.2 1B Instruct",
-        description = "The Llama 3.2 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models.",
+        name = "Llama-3.2 1B Instruct",
+        description = "Meta's highly optimized 1B mobile model. The industry standard for reliable on-device chat.",
         downloadUrl = "https://huggingface.co/bartowski/Llama-3.2-1B-Instruct-GGUF/resolve/main/Llama-3.2-1B-Instruct-Q5_K_M.gguf",
         promptFormat = PromptFormats.Llama3,
         defaultInferenceConfig = InferenceConfig(
             temperature = 0.6f,
+            repeatPenalty = 1.1f,
+            presencePenalty = 0.15f,
+            minP = 0.05f,
+            topP = 0.9f,
+            topK = 40,
+        )
+    ),
+    Model(
+        id = "deepseek-r1-distill-qwen-1.5b",
+        name = "DeepSeek-R1 1.5B (Distilled)",
+        description = "Advanced reasoning capabilities on-device. Uses chain-of-thought processing.",
+        downloadUrl = "https://huggingface.co/unsloth/DeepSeek-R1-Distill-Qwen-1.5B-GGUF/resolve/main/DeepSeek-R1-Distill-Qwen-1.5B-Q8_0.gguf",
+        promptFormat = PromptFormats.ChatML,
+        defaultInferenceConfig = InferenceConfig(
+            temperature = 0.6f,
+            repeatPenalty = 1.0f,
+            presencePenalty = 0.0f,
+            minP = null,
+            topP = 0.95f,
+            topK = 40,
+        ),
+        defaultSystemPrompt = "You are a helpful and harmless assistant. You are Llama Bro."
+    ),
+    Model(
+        id = "smollm2-1.7b-instruct",
+        name = "SmolLM2 1.7B Instruct",
+        description = "High-quality, nuanced text generation that punches above its weight class.",
+        downloadUrl = "https://huggingface.co/bartowski/SmolLM2-1.7B-Instruct-GGUF/resolve/main/SmolLM2-1.7B-Instruct-Q5_K_M.gguf",
+        promptFormat = PromptFormats.ChatML,
+        defaultInferenceConfig = InferenceConfig(
+            temperature = 0.7f,
+            repeatPenalty = 1.1f,
+            presencePenalty = 0.15f,
+            minP = 0.1f,
             topP = 0.9f,
-            topK = 50,
-            repeatPenalty = 1.2f,
+            topK = 40,
         )
     )
 )

app/src/main/java/com/suhel/llamabro/demo/ui/AppScaffold.kt

@@ -24,11 +24,13 @@ import com.suhel.llamabro.demo.R
 @Composable
 fun AppScaffold(
     title: String,
+    modifier: Modifier = Modifier,
     onBack: (() -> Unit)? = null,
     floatingActionButton: @Composable () -> Unit = {},
     content: @Composable ColumnScope.() -> Unit
 ) {
     Scaffold(
+        modifier = modifier,
         topBar = {
             TopAppBar(
                 title = {
@@ -38,7 +40,7 @@ fun AppScaffold(
                     if (onBack != null) {
                         IconButton(onBack) {
                             Icon(
-                                painter = painterResource(R.drawable.arrow_back_24px),
+                                painter = painterResource(R.drawable.arrow_back_24),
                                 contentDescription = "Back"
                             )
                         }

app/src/main/java/com/suhel/llamabro/demo/ui/screens/chat/ChatScreen.kt

@@ -1,7 +1,10 @@
 package com.suhel.llamabro.demo.ui.screens.chat
 
 import androidx.compose.animation.AnimatedContent
+import androidx.compose.animation.animateContentSize
+import androidx.compose.foundation.Image
 import androidx.compose.foundation.background
+import androidx.compose.foundation.clickable
 import androidx.compose.foundation.layout.Arrangement
 import androidx.compose.foundation.layout.Box
 import androidx.compose.foundation.layout.Column
@@ -29,6 +32,9 @@ import androidx.compose.runtime.remember
 import androidx.compose.runtime.setValue
 import androidx.compose.ui.Alignment
 import androidx.compose.ui.Modifier
+import androidx.compose.ui.draw.clip
+import androidx.compose.ui.draw.rotate
+import androidx.compose.ui.keepScreenOn
 import androidx.compose.ui.res.painterResource
 import androidx.compose.ui.text.input.KeyboardCapitalization
 import androidx.compose.ui.unit.dp
@@ -52,6 +58,7 @@ fun ChatScreen(
 ) {
     AppScaffold(
         title = "Chat",
+        modifier = Modifier.keepScreenOn(),
         onBack = onBack
     ) {
         val messages = viewModel.messages.collectAsLazyPagingItems()
@@ -160,12 +167,12 @@ private fun InputBar(
                 ) { generating ->
                     if (generating) {
                         Icon(
-                            painter = painterResource(R.drawable.stop_circle_24px),
+                            painter = painterResource(R.drawable.stop_circle_24),
                             contentDescription = "Stop"
                         )
                     } else {
                         Icon(
-                            painter = painterResource(R.drawable.arrow_circle_up_24px),
+                            painter = painterResource(R.drawable.arrow_circle_up_24),
                             contentDescription = "Send"
                         )
                     }
@@ -186,56 +193,117 @@ private fun MessageBubble(message: UiChatMessage) {
         Column(
             modifier = if (isUser) Modifier.widthIn(max = 300.dp) else Modifier.fillMaxWidth(),
             horizontalAlignment = if (isUser) Alignment.End else Alignment.Start,
-            verticalArrangement = Arrangement.spacedBy(8.dp)
+            verticalArrangement = Arrangement.spacedBy(16.dp)
         ) {
-            if (message.isProcessing && (message.content.isNullOrBlank() && message.thinking.isNullOrBlank())) {
-                Text(
-                    text = "Processing...",
-                    style = MaterialTheme.typography.bodySmall,
-                    color = MaterialTheme.colorScheme.onSurfaceVariant
-                )
+            if (message.isProcessing && message.content.isNullOrBlank() && message.thinking.isNullOrBlank()) {
+                ProcessingIndicator()
             }
 
             message.thinking?.let { thinkingText ->
-                Text(
-                    text = thinkingText,
-                    style = MaterialTheme.typography.bodySmall,
-                    color = MaterialTheme.colorScheme.primaryFixedDim,
-                )
+                ExpandableThinkingBlock(thinkingText = thinkingText)
             }
 
             message.content?.let { contentText ->
                 if (isUser) {
-                    Box(
-                        modifier = Modifier
-                            .background(
-                                MaterialTheme.colorScheme.primary,
-                                MaterialTheme.shapes.medium
-                            )
-                            .padding(16.dp)
-                    ) {
-                        Text(
-                            text = contentText,
-                            style = MaterialTheme.typography.bodyMedium,
-                            color = MaterialTheme.colorScheme.onPrimary
-                        )
-                    }
+                    UserMessageContent(contentText = contentText)
                 } else {
-                    MarkdownText(
-                        markdown = contentText,
-                        style = MaterialTheme.typography.bodyMedium,
-                        modifier = Modifier.fillMaxWidth()
-                    )
+                    AssistantMessageContent(contentText = contentText)
                 }
             }
 
             if (!isUser && message.tokensPerSecond != null) {
-                Text(
-                    text = "%.1f tok/s".format(message.tokensPerSecond),
-                    style = MaterialTheme.typography.labelSmall,
-                    color = MaterialTheme.colorScheme.onSurfaceVariant
-                )
+                GenerationMetrics(tokensPerSecond = message.tokensPerSecond)
             }
         }
     }
 }
+
+@Composable
+private fun ProcessingIndicator() {
+    Text(
+        text = "Processing...",
+        style = MaterialTheme.typography.bodySmall,
+        color = MaterialTheme.colorScheme.onSurfaceVariant
+    )
+}
+
+@Composable
+private fun ExpandableThinkingBlock(thinkingText: String) {
+    var isExpanded by remember { mutableStateOf(false) }
+
+    Column(
+        modifier = Modifier
+            .fillMaxWidth()
+            .background(
+                MaterialTheme.colorScheme.surfaceVariant.copy(alpha = 0.5f),
+                MaterialTheme.shapes.small
+            )
+            .animateContentSize()
+    ) {
+        Row(
+            modifier = Modifier
+                .fillMaxWidth()
+                .clip(MaterialTheme.shapes.small)
+                .clickable { isExpanded = !isExpanded }
+                .padding(8.dp),
+            horizontalArrangement = Arrangement.spacedBy(4.dp),
+            verticalAlignment = Alignment.CenterVertically
+        ) {
+            Text(
+                text = if (isExpanded) "Hide thought process" else "Show thought process",
+                style = MaterialTheme.typography.labelSmall,
+                color = MaterialTheme.colorScheme.onSurfaceVariant,
+                modifier = Modifier.weight(1f)
+            )
+            Image(
+                painter = painterResource(R.drawable.keyboard_arrow_down_24),
+                contentDescription = "chevron-thinking",
+                modifier = Modifier.rotate(if (isExpanded) 180f else 0f)
+            )
+        }
+
+        if (isExpanded) {
+            MarkdownText(
+                markdown = thinkingText,
+                style = MaterialTheme.typography.bodySmall,
+                modifier = Modifier.padding(start = 8.dp, end = 8.dp, bottom = 8.dp)
+            )
+        }
+    }
+}
+
+@Composable
+private fun UserMessageContent(contentText: String) {
+    Box(
+        modifier = Modifier
+            .background(
+                MaterialTheme.colorScheme.primary,
+                MaterialTheme.shapes.medium
+            )
+            .padding(16.dp)
+    ) {
+        Text(
+            text = contentText,
+            style = MaterialTheme.typography.bodyMedium,
+            color = MaterialTheme.colorScheme.onPrimary
+        )
+    }
+}
+
+@Composable
+private fun AssistantMessageContent(contentText: String) {
+    MarkdownText(
+        markdown = contentText,
+        style = MaterialTheme.typography.bodyMedium,
+        modifier = Modifier.fillMaxWidth()
+    )
+}
+
+@Composable
+private fun GenerationMetrics(tokensPerSecond: Float) {
+    Text(
+        text = "%.1f tok/s".format(tokensPerSecond),
+        style = MaterialTheme.typography.labelSmall,
+        color = MaterialTheme.colorScheme.onSurfaceVariant
+    )
+}