added support for cognitive search vector database. App insights monitoring added as well. README updated

Author: dantelmomsft

app/backend/pom.xml

@@ -16,7 +16,7 @@
 	<properties>
 		<java.version>17</java.version>
 		<spring-cloud-azure.version>4.9.0</spring-cloud-azure.version>
-		<azure-search.version>11.6.0-beta.6</azure-search.version>
+		<azure-search.version>11.6.0-beta.8</azure-search.version>
 		<azure-openai.version>1.0.0-beta.2</azure-openai.version>
 		<semantic-kernel.version>0.2.8-alpha</semantic-kernel.version>
 		<mockito-inline.version>4.5.1</mockito-inline.version>

app/backend/src/main/java/com/microsoft/openai/samples/rag/approaches/RAGApproachFactorySpringBootImpl.java

@@ -1,17 +1,19 @@
 package com.microsoft.openai.samples.rag.approaches;
 
-import com.microsoft.openai.samples.rag.ask.approaches.RetrieveThenReadApproach;
-import com.microsoft.openai.samples.rag.ask.approaches.semantickernel.ReadRetrieveReadApproach;
-import com.microsoft.openai.samples.rag.chat.approaches.ChatReadRetrieveReadApproach;
+import com.microsoft.openai.samples.rag.ask.approaches.PlainJavaAskApproach;
+import com.microsoft.openai.samples.rag.ask.approaches.semantickernel.JavaSemanticKernelAskApproach;
+import com.microsoft.openai.samples.rag.chat.approaches.PlainJavaChatApproach;
 import org.springframework.context.ApplicationContext;
 import org.springframework.context.ApplicationContextAware;
 import org.springframework.stereotype.Component;
 
 @Component
 public class RAGApproachFactorySpringBootImpl implements RAGApproachFactory, ApplicationContextAware {
 
-    private static final String READ_RETRIEVE_READ = "rrr";
-    private static final String RETRIEVE_THEN_READ = "rtr";
+    private static final String JAVA_OPENAI_SDK = "jos";
+    private static final String JAVA_SEMANTIC_KERNEL = "jsk";
+
+    private static final String JAVA_SEMANTIC_KERNEL_VECTORS = "jskv";
     private ApplicationContext applicationContext;
 
     /**
@@ -23,14 +25,14 @@ public class RAGApproachFactorySpringBootImpl implements RAGApproachFactory, App
     @Override
     public RAGApproach createApproach(String approachName, RAGType ragType) {
 
-        if (ragType.equals(RAGType.CHAT) && READ_RETRIEVE_READ.equals(approachName)) {
-            return applicationContext.getBean(ChatReadRetrieveReadApproach.class);
+        if (ragType.equals(RAGType.CHAT) && JAVA_OPENAI_SDK.equals(approachName)) {
+            return applicationContext.getBean(PlainJavaChatApproach.class);
 
         } else if (ragType.equals(RAGType.ASK)) {
-            if (RETRIEVE_THEN_READ.equals(approachName))
-                return applicationContext.getBean(RetrieveThenReadApproach.class);
-            else if (READ_RETRIEVE_READ.equals(approachName))
-                return applicationContext.getBean(ReadRetrieveReadApproach.class);
+            if (JAVA_OPENAI_SDK.equals(approachName))
+                return applicationContext.getBean(PlainJavaAskApproach.class);
+            else if (JAVA_SEMANTIC_KERNEL.equals(approachName))
+                return applicationContext.getBean(JavaSemanticKernelAskApproach.class);
         }
         //if this point is reached then the combination of approach and rag type is not supported
         throw new IllegalArgumentException("Invalid combination for approach[%s] and rag type[%s]: ".formatted(approachName, ragType));

app/backend/src/main/java/com/microsoft/openai/samples/rag/approaches/RAGOptions.java

@@ -2,6 +2,8 @@
 
 public class RAGOptions {
 
+    private RetrievalMode retrievalMode;
+
     private boolean semanticRanker;
     private boolean semanticCaptions;
     private boolean suggestFollowupQuestions;
@@ -12,6 +14,9 @@ public class RAGOptions {
     private RAGOptions() {
     }
 
+    public RetrievalMode getRetrievalMode() {
+        return retrievalMode;
+    }
     public boolean isSemanticRanker() {
         return semanticRanker;
     }
@@ -37,6 +42,7 @@ public boolean isSuggestFollowupQuestions() {
     }
 
     public static class Builder {
+        private RetrievalMode retrievalMode;
         private boolean semanticRanker;
         private boolean semanticCaptions;
         private String excludeCategory;
@@ -45,6 +51,10 @@ public static class Builder {
 
         private boolean suggestFollowupQuestions;
 
+        public Builder retrievialMode(String retrievialMode) {
+            this.retrievalMode = RetrievalMode.valueOf(retrievialMode);
+            return this;
+        }
         public Builder semanticRanker(boolean semanticRanker) {
             this.semanticRanker = semanticRanker;
             return this;
@@ -77,12 +87,12 @@ public Builder top(Integer top) {
 
         public RAGOptions build() {
             RAGOptions ragOptions = new RAGOptions();
+            ragOptions.retrievalMode = this.retrievalMode;
             ragOptions.semanticRanker = this.semanticRanker;
             ragOptions.semanticCaptions = this.semanticCaptions;
             ragOptions.suggestFollowupQuestions = this.suggestFollowupQuestions;
             ragOptions.excludeCategory = this.excludeCategory;
             ragOptions.promptTemplate = this.promptTemplate;
-
             ragOptions.top = this.top;
             return ragOptions;
         }

app/backend/src/main/java/com/microsoft/openai/samples/rag/approaches/RetrievalMode.java

@@ -0,0 +1,6 @@
+package com.microsoft.openai.samples.rag.approaches;
+
+public enum RetrievalMode {
+    hybrid, vectors, text;
+
+}

app/backend/src/main/java/com/microsoft/openai/samples/rag/ask/approaches/RetrieveThenReadApproach.java renamed to app/backend/src/main/java/com/microsoft/openai/samples/rag/ask/approaches/PlainJavaAskApproach.java

@@ -6,32 +6,29 @@
 import com.microsoft.openai.samples.rag.approaches.RAGOptions;
 import com.microsoft.openai.samples.rag.approaches.RAGResponse;
 import com.microsoft.openai.samples.rag.common.ChatGPTUtils;
+import com.microsoft.openai.samples.rag.retrieval.FactsRetrieverProvider;
 import com.microsoft.openai.samples.rag.proxy.OpenAIProxy;
-import com.microsoft.openai.samples.rag.retrieval.CognitiveSearchRetriever;
 import com.microsoft.openai.samples.rag.retrieval.Retriever;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
-import org.springframework.context.ApplicationContext;
-import org.springframework.context.ApplicationContextAware;
 import org.springframework.stereotype.Component;
 
 import java.util.List;
 
 /**
- * Simple retrieve-then-read implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
+ * Simple retrieve-then-read java implementation, using the Cognitive Search and OpenAI APIs directly. It first retrieves
  *     top documents from search, then constructs a prompt with them, and then uses OpenAI to generate a completion
  *     (answer) with that prompt.
  */
 @Component
-public class RetrieveThenReadApproach implements RAGApproach<String, RAGResponse>, ApplicationContextAware {
+public class PlainJavaAskApproach implements RAGApproach<String, RAGResponse> {
 
-    private static final Logger LOGGER = LoggerFactory.getLogger(RetrieveThenReadApproach.class);
-    private ApplicationContext applicationContext;
+    private static final Logger LOGGER = LoggerFactory.getLogger(PlainJavaAskApproach.class);
     private final OpenAIProxy openAIProxy;
-    private final Retriever factsRetriever;
+    private final FactsRetrieverProvider factsRetrieverProvider;
 
-    public RetrieveThenReadApproach(Retriever factsRetriever, OpenAIProxy openAIProxy) {
-        this.factsRetriever = factsRetriever;
+    public PlainJavaAskApproach(FactsRetrieverProvider factsRetrieverProvider, OpenAIProxy openAIProxy) {
+        this.factsRetrieverProvider = factsRetrieverProvider;
         this.openAIProxy = openAIProxy;
     }
 
@@ -44,7 +41,8 @@ public RetrieveThenReadApproach(Retriever factsRetriever, OpenAIProxy openAIProx
     public RAGResponse run(String question, RAGOptions options) {
         //TODO exception handling
 
-        Retriever factsRetriever = getFactsRetriever(options);
+        //Get instance of retriever based on the retrieval mode: hybryd, text, vectors
+        Retriever factsRetriever = factsRetrieverProvider.getFactsRetriever(options);
         List<ContentSource> sources = factsRetriever.retrieveFromQuestion(question, options);
         LOGGER.info("Total {} sources found in cognitive search for keyword search query[{}]", sources.size(),
                 question);
@@ -80,19 +78,5 @@ public RAGResponse run(String question, RAGOptions options) {
 
     }
 
-    /**
-     *
-     * @param options rag options containing search types(Cognitive Semantic Search, Cognitive Vector Search, Cognitive Hybrid Search, Semantic Kernel Memory) )
-     * @return retriever implementation
-     */
-    private CognitiveSearchRetriever getFactsRetriever(RAGOptions options) {
-        //default to Cognitive Semantic Search for MVP.
-        return this.applicationContext.getBean(CognitiveSearchRetriever.class);
-
-    }
-
-    public void setApplicationContext(ApplicationContext applicationContext) {
-        this.applicationContext = applicationContext;
-    }
 
 }

app/backend/src/main/java/com/microsoft/openai/samples/rag/ask/approaches/semantickernel/ReadRetrieveReadApproach.java renamed to app/backend/src/main/java/com/microsoft/openai/samples/rag/ask/approaches/semantickernel/JavaSemanticKernelAskApproach.java

@@ -25,9 +25,9 @@
  *    Accomplish the same task as in the Retrieve-then-read approach but using Semantic Kernel framework and Planner goal oriented concept.
  */
 @Component
-public class ReadRetrieveReadApproach implements RAGApproach<String, RAGResponse> {
+public class JavaSemanticKernelAskApproach implements RAGApproach<String, RAGResponse> {
 
-    private static final Logger LOGGER = LoggerFactory.getLogger(ReadRetrieveReadApproach.class);
+    private static final Logger LOGGER = LoggerFactory.getLogger(JavaSemanticKernelAskApproach.class);
     private static final String PLAN_PROMPT = """
             Take the input as a question and answer it finding any information needed
             """;
@@ -37,7 +37,7 @@ public class ReadRetrieveReadApproach implements RAGApproach<String, RAGResponse
     @Value("${openai.chatgpt.deployment}")
     private String gptChatDeploymentModelId;
 
-    public ReadRetrieveReadApproach(CognitiveSearchProxy cognitiveSearchProxy, OpenAIAsyncClient openAIAsyncClient) {
+    public JavaSemanticKernelAskApproach(CognitiveSearchProxy cognitiveSearchProxy, OpenAIAsyncClient openAIAsyncClient) {
         this.cognitiveSearchProxy = cognitiveSearchProxy;
         this.openAIAsyncClient = openAIAsyncClient;
     }
@@ -92,6 +92,7 @@ private Kernel buildSemanticKernel( RAGOptions options) {
                         .build())
                 .build();
 
+        //TODO: should be refactored to reuse the facts retriever provider injected by spring
         kernel.importSkill(new CognitiveSearchPlugin(this.cognitiveSearchProxy, CognitiveSearchPlugin.buildSearchOptions(options),options), "CognitiveSearchPlugin");
 
         kernel.importSkillFromResources(

app/backend/src/main/java/com/microsoft/openai/samples/rag/ask/controller/AskController.java

@@ -46,6 +46,7 @@ public ResponseEntity<AskResponse> openAIAsk(@RequestBody AskRequest askRequest)
         }
 
         var ragOptions = new RAGOptions.Builder()
+            .retrievialMode(askRequest.getOverrides().getRetrievalMode())
             .semanticRanker(askRequest.getOverrides().isSemanticRanker())
             .semanticCaptions(askRequest.getOverrides().isSemanticCaptions())
             .excludeCategory(askRequest.getOverrides().getExcludeCategory())

app/backend/src/main/java/com/microsoft/openai/samples/rag/chat/approaches/ChatReadRetrieveReadApproach.java renamed to app/backend/src/main/java/com/microsoft/openai/samples/rag/chat/approaches/PlainJavaChatApproach.java

@@ -7,27 +7,32 @@
 import com.microsoft.openai.samples.rag.approaches.RAGResponse;
 import com.microsoft.openai.samples.rag.common.ChatGPTConversation;
 import com.microsoft.openai.samples.rag.common.ChatGPTUtils;
+import com.microsoft.openai.samples.rag.retrieval.FactsRetrieverProvider;
 import com.microsoft.openai.samples.rag.proxy.OpenAIProxy;
-import com.microsoft.openai.samples.rag.retrieval.CognitiveSearchRetriever;
 import com.microsoft.openai.samples.rag.retrieval.Retriever;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 import org.springframework.context.ApplicationContext;
-import org.springframework.context.ApplicationContextAware;
 import org.springframework.stereotype.Component;
 
 import java.util.*;
 
+/**
+ * Simple chat-read-retrieve-read java implementation, using the Cognitive Search and OpenAI APIs directly.
+ * It uses the ChatGPT API to turn the user question into a good search query.
+ * It queries Azure Cognitive Search for search results for that query (optionally using the vector embeddings for that query).
+ * It then combines the search results and original user question, and asks ChatGPT API to answer the question based on the sources. It includes the last 4K of message history as well (or however many tokens are allowed by the deployed model).
+ */
 @Component
-public class ChatReadRetrieveReadApproach implements RAGApproach<ChatGPTConversation, RAGResponse>, ApplicationContextAware {
+public class PlainJavaChatApproach implements RAGApproach<ChatGPTConversation, RAGResponse> {
 
-    private static final Logger LOGGER = LoggerFactory.getLogger(ChatReadRetrieveReadApproach.class);
+    private static final Logger LOGGER = LoggerFactory.getLogger(PlainJavaChatApproach.class);
     private ApplicationContext applicationContext;
     private final OpenAIProxy openAIProxy;
-    private final Retriever factsRetriever;
+    private final FactsRetrieverProvider factsRetrieverProvider;
 
-    public ChatReadRetrieveReadApproach(Retriever factsRetriever, OpenAIProxy openAIProxy) {
-        this.factsRetriever = factsRetriever;
+    public PlainJavaChatApproach(FactsRetrieverProvider factsRetrieverProvider, OpenAIProxy openAIProxy) {
+        this.factsRetrieverProvider = factsRetrieverProvider;
         this.openAIProxy = openAIProxy;
     }
 
@@ -39,7 +44,7 @@ public ChatReadRetrieveReadApproach(Retriever factsRetriever, OpenAIProxy openAI
     @Override
     public RAGResponse run(ChatGPTConversation questionOrConversation, RAGOptions options) {
 
-        Retriever factsRetriever = getFactsRetriever(options);
+        Retriever factsRetriever = factsRetrieverProvider.getFactsRetriever(options);
         List<ContentSource> sources = factsRetriever.retrieveFromConversation(questionOrConversation, options);
         LOGGER.info("Total {} sources retrieved", sources.size());
 
@@ -65,18 +70,4 @@ public RAGResponse run(ChatGPTConversation questionOrConversation, RAGOptions op
     }
 
 
-    /**
-     *
-     * @param options rag options containing search types(Cognitive Semantic Search, Cognitive Vector Search, Cognitive Hybrid Search, Semantic Kernel Memory) )
-     * @return retriever implementation
-     */
-    private CognitiveSearchRetriever getFactsRetriever(RAGOptions options) {
-        //default to Cognitive Semantic Search for MVP.
-        return this.applicationContext.getBean(CognitiveSearchRetriever.class);
-
-    }
-    public void setApplicationContext(ApplicationContext applicationContext) {
-        this.applicationContext = applicationContext;
-    }
-
 }

app/backend/src/main/java/com/microsoft/openai/samples/rag/chat/approaches/SemanticSearchChat.java

@@ -19,11 +19,11 @@ public class SemanticSearchChat {
     private Boolean replacePrompt = false;
 
     private static final String FOLLOW_UP_QUESTIONS_TEMPLATE = """
-    Generate three very brief follow-up questions that the user would likely ask next about their healthcare plan and employee handbook.
+    After answering question, also generate three very brief follow-up questions that the user would likely ask next.
     Use double angle brackets to reference the questions, e.g. <<Are there exclusions for prescriptions?>>.
     Try not to repeat questions that have already been asked.
     Only generate questions and do not generate any text before or after the questions, such as 'Next Questions'
-    """;
+            """;
     private static final String SYSTEM_CHAT_MESSAGE_TEMPLATE = """
      Assistant helps the company employees with their healthcare plan questions, and questions about the employee handbook. Be brief in your answers.
      Answer ONLY with the facts listed in the list of sources below. If there isn't enough information below, say you don't know. Do not generate answers that don't use the sources below. If asking a clarifying question to the user would help, ask the question.

app/backend/src/main/java/com/microsoft/openai/samples/rag/chat/controller/ChatController.java

@@ -50,6 +50,7 @@ public ResponseEntity<ChatResponse> openAIAsk(@RequestBody ChatRequest chatReque
         }
 
         var ragOptions = new RAGOptions.Builder()
+            .retrievialMode(chatRequest.getOverrides().getRetrievalMode())
             .semanticRanker(chatRequest.getOverrides().isSemanticRanker())
             .semanticCaptions(chatRequest.getOverrides().isSemanticCaptions())
             .suggestFollowupQuestions(chatRequest.getOverrides().isSuggestFollowupQuestions())