feat(demo): add cache toggle with local and cloud cache options

Adds ComboBox selector for semantic cache type:
- No Cache: Always call LLM
- Local Cache: Redis-based semantic cache using SemanticCache
- LangCache: Cloud-hosted semantic cache service

Changes:
- CacheType enum with NONE, LOCAL, LANGCACHE options
- ServiceFactory initializes both local SemanticCache and LangCache
- RAGService accepts CacheType parameter instead of boolean
- ChatController uses ComboBox instead of CheckBox
- CSS styling for cache ComboBox
- Tests updated to use CacheType enum

Author: bsbodden

demos/rag-multimodal/src/main/java/com/redis/vl/demo/rag/model/CacheType.java

@@ -0,0 +1,36 @@
+package com.redis.vl.demo.rag.model;
+
+/**
+ * Types of semantic caching available in the demo.
+ */
+public enum CacheType {
+  /** No caching - always call LLM */
+  NONE("No Cache", "Always call LLM, no caching"),
+
+  /** Local Redis-based semantic cache */
+  LOCAL("Local Cache", "Redis-based semantic cache (local)"),
+
+  /** Cloud LangCache service */
+  LANGCACHE("LangCache", "Cloud-hosted semantic cache service");
+
+  private final String displayName;
+  private final String description;
+
+  CacheType(String displayName, String description) {
+    this.displayName = displayName;
+    this.description = description;
+  }
+
+  public String getDisplayName() {
+    return displayName;
+  }
+
+  public String getDescription() {
+    return description;
+  }
+
+  @Override
+  public String toString() {
+    return displayName;
+  }
+}

demos/rag-multimodal/src/main/java/com/redis/vl/demo/rag/service/RAGService.java

@@ -1,8 +1,11 @@
 package com.redis.vl.demo.rag.service;
 
+import com.redis.vl.demo.rag.model.CacheType;
 import com.redis.vl.demo.rag.model.ChatMessage;
 import com.redis.vl.demo.rag.model.LLMConfig;
+import com.redis.vl.extensions.cache.CacheHit;
 import com.redis.vl.extensions.cache.LangCacheSemanticCache;
+import com.redis.vl.extensions.cache.SemanticCache;
 import com.redis.vl.langchain4j.RedisVLContentRetriever;
 import com.redis.vl.langchain4j.RedisVLDocumentStore;
 import dev.langchain4j.data.image.Image;
@@ -42,6 +45,7 @@ public class RAGService {
   private final ChatLanguageModel chatModel;
   private final CostTracker costTracker;
   private final LLMConfig config;
+  private final SemanticCache localCache;
   private final LangCacheSemanticCache langCache;
 
   private static final String SYSTEM_PROMPT =
@@ -61,38 +65,54 @@ public class RAGService {
    * @param chatModel Chat language model for generation
    * @param costTracker Cost tracker for calculating costs
    * @param config LLM configuration
-   * @param langCache LangCache wrapper for semantic caching (can be null)
+   * @param localCache Local Redis-based semantic cache
+   * @param langCache LangCache cloud service for semantic caching (can be null)
    */
   public RAGService(
       RedisVLContentRetriever contentRetriever,
       RedisVLDocumentStore documentStore,
       ChatLanguageModel chatModel,
       CostTracker costTracker,
       LLMConfig config,
+      SemanticCache localCache,
       LangCacheSemanticCache langCache) {
     this.contentRetriever = contentRetriever;
     this.documentStore = documentStore;
     this.chatModel = chatModel;
     this.costTracker = costTracker;
     this.config = config;
+    this.localCache = localCache;
     this.langCache = langCache;
   }
 
   /**
    * Queries the RAG system.
    *
    * @param userQuery User's question
-   * @param useCache Whether to use LangCache for semantic caching
+   * @param cacheType Type of caching to use (NONE, LOCAL, or LANGCACHE)
    * @return Assistant response with cost tracking
    * @throws IllegalArgumentException if userQuery is null or empty
    */
-  public ChatMessage query(String userQuery, boolean useCache) {
+  public ChatMessage query(String userQuery, CacheType cacheType) {
     if (userQuery == null || userQuery.trim().isEmpty()) {
       throw new IllegalArgumentException("User query cannot be null or empty");
     }
 
-    // Check cache first if enabled
-    if (useCache && langCache != null) {
+    // Check cache first based on cache type
+    if (cacheType == CacheType.LOCAL && localCache != null) {
+      System.out.println("→ Checking Local Cache for: " + userQuery);
+      Optional<CacheHit> cacheHit = localCache.check(userQuery);
+
+      if (cacheHit.isPresent()) {
+        String cachedResponse = cacheHit.get().getResponse();
+        int tokenCount = costTracker.countTokens(cachedResponse);
+
+        System.out.println("✓ Local Cache HIT for query: " + userQuery);
+        return ChatMessage.assistant(cachedResponse, tokenCount, 0.0, config.model(), true);
+      } else {
+        System.out.println("✗ Local Cache MISS for query: " + userQuery);
+      }
+    } else if (cacheType == CacheType.LANGCACHE && langCache != null) {
       try {
         System.out.println("→ Checking LangCache for: " + userQuery);
         List<Map<String, Object>> cacheHits = langCache.check(userQuery, null, 1, null, null, 0.9f);
@@ -103,10 +123,10 @@ public ChatMessage query(String userQuery, boolean useCache) {
           String cachedResponse = (String) hit.get("response");
           int tokenCount = costTracker.countTokens(cachedResponse);
 
-          System.out.println("✓ Cache HIT for query: " + userQuery);
+          System.out.println("✓ LangCache HIT for query: " + userQuery);
           return ChatMessage.assistant(cachedResponse, tokenCount, 0.0, config.model(), true);
         } else {
-          System.out.println("✗ Cache MISS for query: " + userQuery);
+          System.out.println("✗ LangCache MISS for query: " + userQuery);
         }
       } catch (IOException e) {
         System.err.println("LangCache check failed: " + e.getMessage());
@@ -201,12 +221,16 @@ public ChatMessage query(String userQuery, boolean useCache) {
     Response<AiMessage> response = chatModel.generate(messages);
     String responseText = response.content().text();
 
-    // 4. Store in cache if enabled
-    if (useCache && langCache != null) {
+    // 4. Store in cache based on cache type
+    if (cacheType == CacheType.LOCAL && localCache != null) {
+      System.out.println("→ Storing to Local Cache: " + userQuery);
+      localCache.store(userQuery, responseText);
+      System.out.println("✓ Cached response in Local Cache for query: " + userQuery);
+    } else if (cacheType == CacheType.LANGCACHE && langCache != null) {
       try {
         System.out.println("→ Storing to LangCache: " + userQuery);
         String entryId = langCache.store(userQuery, responseText, null);
-        System.out.println("✓ Cached response for query: " + userQuery + " (entry_id: " + entryId + ")");
+        System.out.println("✓ Cached response in LangCache for query: " + userQuery + " (entry_id: " + entryId + ")");
       } catch (IOException e) {
         System.err.println("LangCache store failed: " + e.getMessage());
         e.printStackTrace();

demos/rag-multimodal/src/main/java/com/redis/vl/demo/rag/service/ServiceFactory.java

@@ -3,11 +3,13 @@
 import com.redis.vl.demo.rag.config.AppConfig;
 import com.redis.vl.demo.rag.model.LLMConfig;
 import com.redis.vl.extensions.cache.LangCacheSemanticCache;
+import com.redis.vl.extensions.cache.SemanticCache;
 import com.redis.vl.index.SearchIndex;
 import com.redis.vl.langchain4j.RedisVLContentRetriever;
 import com.redis.vl.langchain4j.RedisVLDocumentStore;
 import com.redis.vl.langchain4j.RedisVLEmbeddingStore;
 import com.redis.vl.schema.IndexSchema;
+import com.redis.vl.utils.vectorize.LangChain4JVectorizer;
 import dev.langchain4j.model.chat.ChatLanguageModel;
 import dev.langchain4j.model.embedding.EmbeddingModel;
 import dev.langchain4j.model.embedding.onnx.allminilml6v2.AllMiniLmL6V2EmbeddingModel;
@@ -34,6 +36,7 @@ public class ServiceFactory {
   private EmbeddingModel embeddingModel;
   private CostTracker costTracker;
   private LangCacheSemanticCache langCache;
+  private SemanticCache localCache;
 
   /**
    * Initializes all services with default configuration.
@@ -59,6 +62,17 @@ public void initialize(String redisHost, int redisPort) throws Exception {
     embeddingStore = new RedisVLEmbeddingStore(searchIndex);
     documentStore = new RedisVLDocumentStore(jedis, "rag:docs:");
 
+    // Initialize local SemanticCache (always available)
+    LangChain4JVectorizer vectorizer = new LangChain4JVectorizer(
+        "all-minilm-l6-v2", embeddingModel, VECTOR_DIM);
+    localCache = new SemanticCache.Builder()
+        .name("rag_local_cache")
+        .redisClient(jedis)
+        .vectorizer(vectorizer)
+        .distanceThreshold(0.1f)  // 90% similarity for cache hit
+        .build();
+    System.out.println("Local SemanticCache initialized");
+
     // Initialize LangCache if enabled
     AppConfig config = AppConfig.getInstance();
     if (config.isLangCacheEnabled()) {
@@ -143,7 +157,7 @@ public RAGService createRAGService(LLMConfig config) {
     // Create chat model based on provider
     ChatLanguageModel chatModel = createChatModel(config);
 
-    return new RAGService(retriever, documentStore, chatModel, costTracker, config, langCache);
+    return new RAGService(retriever, documentStore, chatModel, costTracker, config, localCache, langCache);
   }
 
   /**

demos/rag-multimodal/src/main/java/com/redis/vl/demo/rag/ui/ChatController.java

@@ -1,6 +1,7 @@
 package com.redis.vl.demo.rag.ui;
 
 import com.redis.vl.demo.rag.config.AppConfig;
+import com.redis.vl.demo.rag.model.CacheType;
 import com.redis.vl.demo.rag.model.ChatMessage;
 import com.redis.vl.demo.rag.model.LLMConfig;
 import com.redis.vl.demo.rag.service.JTokKitCostTracker;
@@ -29,7 +30,7 @@ public class ChatController extends BorderPane {
   private final ListView<ChatMessage> messageListView;
   private final TextField inputField;
   private final Button sendButton;
-  private final CheckBox cacheCheckBox;
+  private final ComboBox<CacheType> cacheComboBox;
   private final Label statusLabel;
   private final Label providerLabel;
   private final Label modelLabel;
@@ -105,15 +106,17 @@ protected void updateItem(ChatMessage item, boolean empty) {
 
     Separator configSeparator = new Separator();
 
-    // LangCache section
+    // Cache section
     Label cacheTitle = new Label("Semantic Cache");
     cacheTitle.getStyleClass().add("section-title");
 
-    cacheCheckBox = new CheckBox("Enable LangCache");
-    cacheCheckBox.setSelected(config.isLangCacheEnabled());
-    cacheCheckBox.getStyleClass().add("cache-checkbox");
+    cacheComboBox = new ComboBox<>();
+    cacheComboBox.getItems().addAll(CacheType.values());
+    cacheComboBox.setValue(config.isLangCacheEnabled() ? CacheType.LANGCACHE : CacheType.NONE);
+    cacheComboBox.getStyleClass().add("cache-combobox");
+    cacheComboBox.setMaxWidth(Double.MAX_VALUE);
 
-    Label cacheInfo = new Label("Toggle to demonstrate\ncost savings with\nsemantic caching");
+    Label cacheInfo = new Label("Select caching mode:\n• No Cache: Always call LLM\n• Local: Redis-based cache\n• LangCache: Cloud cache");
     cacheInfo.getStyleClass().add("info-text");
     cacheInfo.setWrapText(true);
 
@@ -142,7 +145,7 @@ protected void updateItem(ChatMessage item, boolean empty) {
 
     rightPanel.getChildren().addAll(
         configTitle, providerLabel, modelLabel, loadedPdfLabel, configSeparator,
-        cacheTitle, cacheCheckBox, cacheInfo, cacheSeparator,
+        cacheTitle, cacheComboBox, cacheInfo, cacheSeparator,
         actionsTitle, uploadPdfButton, actionsSeparator,
         spacer,
         statusTitle, statusLabel
@@ -209,8 +212,8 @@ private void sendMessage() {
     executor.submit(
         () -> {
           try {
-            boolean useCache = cacheCheckBox.isSelected();
-            ChatMessage response = ragService.query(userInput, useCache);
+            CacheType cacheType = cacheComboBox.getValue();
+            ChatMessage response = ragService.query(userInput, cacheType);
 
             Platform.runLater(
                 () -> {

demos/rag-multimodal/src/main/resources/styles/app.css

@@ -337,3 +337,36 @@
     -fx-background-color: #E4E6EB;
     -fx-padding: 0 2px;
 }
+
+/* Cache ComboBox */
+.cache-combobox {
+    -fx-background-color: #FFFFFF;
+    -fx-border-color: #CED0D4;
+    -fx-border-width: 1px;
+    -fx-border-radius: 4px;
+    -fx-background-radius: 4px;
+    -fx-padding: 8px 12px;
+    -fx-font-size: 13px;
+}
+
+.cache-combobox:hover {
+    -fx-border-color: #1877F2;
+}
+
+.cache-combobox .list-cell {
+    -fx-padding: 8px 12px;
+}
+
+.cache-combobox-popup .list-view {
+    -fx-background-color: #FFFFFF;
+    -fx-border-color: #CED0D4;
+}
+
+.cache-combobox-popup .list-cell:hover {
+    -fx-background-color: #F0F2F5;
+}
+
+.cache-combobox-popup .list-cell:selected {
+    -fx-background-color: #E7F3FF;
+    -fx-text-fill: #1877F2;
+}

demos/rag-multimodal/src/test/java/com/redis/vl/demo/rag/service/MultimodalRAGIntegrationTest.java

@@ -3,6 +3,7 @@
 import static org.junit.jupiter.api.Assertions.*;
 
 import com.redis.vl.BaseIntegrationTest;
+import com.redis.vl.demo.rag.model.CacheType;
 import com.redis.vl.demo.rag.model.ChatMessage;
 import com.redis.vl.demo.rag.model.LLMConfig;
 import com.redis.vl.index.SearchIndex;
@@ -126,7 +127,7 @@ void setUp() throws Exception {
 
     JTokKitCostTracker costTracker = new JTokKitCostTracker();
 
-    ragService = new RAGService(retriever, documentStore, chatModel, costTracker, config, null);
+    ragService = new RAGService(retriever, documentStore, chatModel, costTracker, config, null, null);
 
     // Ingest Attention.pdf
     File attentionPdf =
@@ -163,7 +164,7 @@ void testVisionModelReceivesAndProcessesImages() {
     String query = "Describe the Transformer architecture diagram. What are the main components?";
 
     // When: Query the RAG system
-    ChatMessage response = ragService.query(query, false);
+    ChatMessage response = ragService.query(query, CacheType.NONE);
 
     // Then: Response should mention visual elements from the diagram
     String responseText = response.content().toLowerCase();
@@ -203,7 +204,7 @@ void testImagesRetrievedFromDocumentStore() {
     String query = "What does the attention visualization show?";
 
     // When: Query the RAG system
-    ChatMessage response = ragService.query(query, false);
+    ChatMessage response = ragService.query(query, CacheType.NONE);
 
     // Then: Response should indicate it saw the actual visualization image
     String responseText = response.content().toLowerCase();
@@ -232,7 +233,7 @@ void testGPT4VisionDescribesVisualElements() {
     String query = "How many attention heads are shown in the Multi-Head Attention diagram?";
 
     // When: Query the RAG system
-    ChatMessage response = ragService.query(query, false);
+    ChatMessage response = ragService.query(query, CacheType.NONE);
 
     // Then: Response should provide specific count from diagram
     String responseText = response.content().toLowerCase();

demos/rag-multimodal/src/test/java/com/redis/vl/demo/rag/service/RAGServiceTest.java

@@ -4,6 +4,7 @@
 import static org.mockito.ArgumentMatchers.any;
 import static org.mockito.Mockito.*;
 
+import com.redis.vl.demo.rag.model.CacheType;
 import com.redis.vl.demo.rag.model.ChatMessage;
 import com.redis.vl.demo.rag.model.LLMConfig;
 import com.redis.vl.langchain4j.RedisVLContentRetriever;
@@ -36,7 +37,7 @@ class RAGServiceTest {
   void setUp() {
     MockitoAnnotations.openMocks(this);
     config = LLMConfig.defaultConfig(LLMConfig.Provider.OPENAI, "test-key");
-    ragService = new RAGService(contentRetriever, documentStore, chatModel, costTracker, config, null);
+    ragService = new RAGService(contentRetriever, documentStore, chatModel, costTracker, config, null, null);
   }
 
   @Test
@@ -61,7 +62,7 @@ void testQueryWithContext() {
         .thenReturn(0.0001);
 
     // When
-    ChatMessage result = ragService.query(userQuery, false);
+    ChatMessage result = ragService.query(userQuery, CacheType.NONE);
 
     // Then
     assertNotNull(result);
@@ -89,7 +90,7 @@ void testQueryWithCache() {
     when(costTracker.calculateCost(any(), any(), anyInt())).thenReturn(0.0);
 
     // When
-    ChatMessage result = ragService.query(userQuery, true);
+    ChatMessage result = ragService.query(userQuery, CacheType.LANGCACHE);
 
     // Then
     assertNotNull(result);
@@ -113,7 +114,7 @@ void testQueryWithEmptyContext() {
     when(costTracker.calculateCost(any(), any(), anyInt())).thenReturn(0.00005);
 
     // When
-    ChatMessage result = ragService.query(userQuery, false);
+    ChatMessage result = ragService.query(userQuery, CacheType.NONE);
 
     // Then
     assertNotNull(result);
@@ -123,12 +124,12 @@ void testQueryWithEmptyContext() {
   @Test
   void testQueryWithNullInput() {
     // When/Then
-    assertThrows(IllegalArgumentException.class, () -> ragService.query(null, false));
+    assertThrows(IllegalArgumentException.class, () -> ragService.query(null, CacheType.NONE));
   }
 
   @Test
   void testQueryWithEmptyInput() {
     // When/Then
-    assertThrows(IllegalArgumentException.class, () -> ragService.query("", false));
+    assertThrows(IllegalArgumentException.class, () -> ragService.query("", CacheType.NONE));
   }
 }