Align Gemma4 config with gallery multimodal backend setup

Author: Android-PowerUser

app/src/main/kotlin/com/google/ai/sample/GenerativeAiViewModelFactory.kt

@@ -60,6 +60,7 @@ enum class ModelOption(
         "gemma-4-e4b-it",
         ApiProvider.GOOGLE,
         "https://huggingface.co/litert-community/gemma-4-E4B-it-litert-lm/resolve/main/gemma-4-E4B-it.litertlm?download=true",
+        supportsScreenshot = false,
         isOfflineModel = true,
         offlineModelFilename = "gemma-4-E4B-it.litertlm"
     ),

app/src/main/kotlin/com/google/ai/sample/feature/multimodal/PhotoReasoningViewModel.kt

@@ -345,21 +345,27 @@ class PhotoReasoningViewModel(
                     )
                     if (liteRtEngine == null) {
                         val liteRtBackend = if (backend == InferenceBackend.GPU) Backend.GPU() else Backend.CPU()
-                        val visionBackend = if (currentModel.supportsScreenshot) Backend.GPU() else null
+                        val visionBackend = if (currentModel.supportsScreenshot) Backend.CPU() else null
                         val audioBackend = null
+                        val cacheDir =
+                            if (modelFile.absolutePath.startsWith("/data/local/tmp")) {
+                                context.getExternalFilesDir(null)?.absolutePath
+                            } else {
+                                null
+                            }
                         val engineConfig = EngineConfig(
                             modelPath = modelFile.absolutePath,
                             backend = liteRtBackend,
                             visionBackend = visionBackend,
                             audioBackend = audioBackend,
                             maxNumTokens = null,
-                            cacheDir = context.cacheDir.absolutePath
+                            cacheDir = cacheDir
                         )
                         Log.i(
                             TAG,
                             "Creating LiteRT engine with backend=$liteRtBackend, " +
                                 "visionBackend=$visionBackend, audioBackend=$audioBackend, " +
-                                "cacheDir=${context.cacheDir.absolutePath}"
+                                "cacheDir=$cacheDir"
                         )
                         liteRtEngine = Engine(engineConfig).also { it.initialize() }
                         Log.d(TAG, "Offline model initialized with LiteRT-LM Engine backend=$liteRtBackend")