Disable streaming for SK

Author: Milder Hernandez Cagua

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

@@ -95,6 +95,7 @@ public void runStreaming(
             ChatGPTConversation questionOrConversation,
             RAGOptions options,
             OutputStream outputStream) {
+        throw new IllegalStateException("Streaming not supported for this approach");
     }
 
     private List<ContentSource> formSourcesList(String result) {

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

@@ -102,48 +102,7 @@ public RAGResponse run(ChatGPTConversation questionOrConversation, RAGOptions op
 
     @Override
     public void runStreaming(ChatGPTConversation questionOrConversation, RAGOptions options, OutputStream outputStream) {
-        String question = ChatGPTUtils.getLastUserQuestion(questionOrConversation.getMessages());
-
-        // STEP 1: Build semantic kernel with Azure Cognitive Search as memory store. AnswerQuestion skill is imported from resources.
-        Kernel semanticKernel = buildSemanticKernel(options);
-
-        // STEP 2: Retrieve relevant documents using keywords extracted from the chat history
-        String conversation = ChatGPTUtils.formatAsChatML(questionOrConversation.toOpenAIChatMessages());
-        List<MemoryQueryResult> sourcesResult = getSourcesFromConversation(conversation, semanticKernel, options);
-
-        LOGGER.info("Total {} sources found in cognitive vector store for search query[{}]", sourcesResult.size(), question);
-
-        String sources = buildSourcesText(sourcesResult);
-        List<ContentSource> sourcesList = buildSources(sourcesResult);
-
-        // STEP 3: Generate a contextual and content specific answer using the search results and chat history
-        SKFunction answerConversation = semanticKernel.getFunction("RAG", "AnswerConversation");
-        SKContext skcontext = SKBuilders.context().build()
-                .setVariable("sources", sources)
-                .setVariable("conversation", conversation)
-                .setVariable("suggestions", String.valueOf(options.isSuggestFollowupQuestions()))
-                .setVariable("input",  question);
-
-        SKContext reply = (SKContext) answerConversation.invokeAsync(skcontext).block();
-
-        RAGResponse ragResponse =
-                new RAGResponse.Builder()
-                        .question(
-                                ChatGPTUtils.getLastUserQuestion(
-                                        questionOrConversation.getMessages()))
-                        .prompt("placeholders for prompt")
-                        .answer(reply.getResult())
-                        .sources(sourcesList)
-                        .sourcesAsText(sources)
-                        .build();
-
-        try {
-            String value = objectMapper.writeValueAsString(ChatResponse.buildChatResponse(ragResponse)) + "\n";
-            outputStream.write(value.getBytes());
-            outputStream.flush();
-        } catch (IOException e) {
-            throw new RuntimeException(e);
-        }
+        throw new IllegalStateException("Streaming not supported for this approach");
     }
 
     private List<MemoryQueryResult> getSourcesFromConversation (String conversation, Kernel kernel, RAGOptions options) {

app/frontend/src/pages/chat/Chat.tsx

@@ -35,6 +35,7 @@ const Chat = () => {
     const [retrievalMode, setRetrievalMode] = useState<RetrievalMode>(RetrievalMode.Hybrid);
     const [useSemanticRanker, setUseSemanticRanker] = useState<boolean>(true);
     const [shouldStream, setShouldStream] = useState<boolean>(true);
+    const [streamAvailable, setStreamAvailable] = useState<boolean>(true);
     const [useSemanticCaptions, setUseSemanticCaptions] = useState<boolean>(false);
     const [excludeCategory, setExcludeCategory] = useState<string>("");
     const [useSuggestFollowupQuestions, setUseSuggestFollowupQuestions] = useState<boolean>(false);
@@ -112,9 +113,10 @@ const Chat = () => {
                 { content: a[1].choices[0].message.content, role: "assistant" }
             ]);
 
+            const stream = streamAvailable && shouldStream;
             const request: ChatAppRequest = {
                 messages: [...messages, { content: question, role: "user" }],
-                stream: shouldStream,
+                stream: stream,
                 context: {
                     overrides: {
                         prompt_template: promptTemplate.length === 0 ? undefined : promptTemplate,
@@ -138,7 +140,7 @@ const Chat = () => {
             if (!response.body) {
                 throw Error("No response body");
             }
-            if (shouldStream) {
+            if (stream) {
                 const parsedResponse: ChatAppResponse = await handleAsyncRequest(question, answers, setAnswers, response.body);
                 setAnswers([...answers, [question, parsedResponse]]);
             } else {
@@ -186,7 +188,9 @@ const Chat = () => {
     };
 
     const onApproachChange = (_ev?: React.FormEvent<HTMLElement | HTMLInputElement>, option?: IChoiceGroupOption) => {
-        setApproach((option?.key as Approaches) || Approaches.JAVA_OPENAI_SDK);
+        const newApproach = (option?.key as Approaches);
+        setApproach(newApproach || Approaches.JAVA_OPENAI_SDK);
+        setStreamAvailable(newApproach === Approaches.JAVA_OPENAI_SDK);
     };
 
     const onUseSemanticRankerChange = (_ev?: React.FormEvent<HTMLElement | HTMLInputElement>, checked?: boolean) => {
@@ -450,12 +454,14 @@ const Chat = () => {
                         required
                         onChange={onRetrievalModeChange}
                     />
-                    <Checkbox
-                        className={styles.chatSettingsSeparator}
-                        checked={shouldStream}
-                        label="Stream chat completion responses"
-                        onChange={onShouldStreamChange}
-                    />
+                    {streamAvailable &&
+                        <Checkbox
+                            className={styles.chatSettingsSeparator}
+                            checked={shouldStream}
+                            label="Stream chat completion responses"
+                            onChange={onShouldStreamChange}
+                        />
+                    }
 
                     {useLogin && <TokenClaimsDisplay />}
                 </Panel>