conversation history fix

frontend/src/app/chat/page.tsx

@@ -36,12 +36,23 @@ export default function ChatPage() {
     timestamp: Date
   } | null>(null)
   const [expandedFunctionCalls, setExpandedFunctionCalls] = useState<Set<string>>(new Set())
+  const [previousResponseIds, setPreviousResponseIds] = useState<{
+    chat: string | null
+    langflow: string | null
+  }>({ chat: null, langflow: null })
   const messagesEndRef = useRef<HTMLDivElement>(null)
 
   const scrollToBottom = () => {
     messagesEndRef.current?.scrollIntoView({ behavior: "smooth" })
   }
 
+  const handleEndpointChange = (newEndpoint: EndpointType) => {
+    setEndpoint(newEndpoint)
+    // Clear the conversation when switching endpoints to avoid response ID conflicts
+    setMessages([])
+    setPreviousResponseIds({ chat: null, langflow: null })
+  }
+
   useEffect(() => {
     scrollToBottom()
   }, [messages, streamingMessage])
@@ -50,15 +61,23 @@ export default function ChatPage() {
     const apiEndpoint = endpoint === "chat" ? "/api/chat" : "/api/langflow"
     
     try {
+      const requestBody: any = { 
+        prompt: userMessage.content,
+        stream: true 
+      }
+      
+      // Add previous_response_id if we have one for this endpoint
+      const currentResponseId = previousResponseIds[endpoint]
+      if (currentResponseId) {
+        requestBody.previous_response_id = currentResponseId
+      }
+      
       const response = await fetch(apiEndpoint, {
         method: "POST",
         headers: {
           "Content-Type": "application/json",
         },
-        body: JSON.stringify({ 
+        body: JSON.stringify(requestBody),
-          prompt: userMessage.content,
-          stream: true 
-        }),
       })
 
       if (!response.ok) {
@@ -74,6 +93,7 @@ export default function ChatPage() {
       let buffer = ""
       let currentContent = ""
       const currentFunctionCalls: FunctionCall[] = []
+      let newResponseId: string | null = null
       
       // Initialize streaming message
       setStreamingMessage({
@@ -100,6 +120,13 @@ export default function ChatPage() {
                 const chunk = JSON.parse(line)
                 console.log("Received chunk:", chunk.type || chunk.object, chunk)
                 
+                // Extract response ID if present
+                if (chunk.id) {
+                  newResponseId = chunk.id
+                } else if (chunk.response_id) {
+                  newResponseId = chunk.response_id
+                }
+                
                 // Handle OpenAI Chat Completions streaming format
                 if (chunk.object === "response.chunk" && chunk.delta) {
                   // Handle function calls in delta
@@ -336,6 +363,14 @@ export default function ChatPage() {
       setMessages(prev => [...prev, finalMessage])
       setStreamingMessage(null)
       
+      // Store the response ID for the next request for this endpoint
+      if (newResponseId) {
+        setPreviousResponseIds(prev => ({
+          ...prev,
+          [endpoint]: newResponseId
+        }))
+      }
+      
     } catch (error) {
       console.error("SSE Stream error:", error)
       setStreamingMessage(null)
@@ -369,12 +404,21 @@ export default function ChatPage() {
       // Original non-streaming logic
       try {
         const apiEndpoint = endpoint === "chat" ? "/api/chat" : "/api/langflow"
+        
+        const requestBody: any = { prompt: userMessage.content }
+        
+        // Add previous_response_id if we have one for this endpoint
+        const currentResponseId = previousResponseIds[endpoint]
+        if (currentResponseId) {
+          requestBody.previous_response_id = currentResponseId
+        }
+        
         const response = await fetch(apiEndpoint, {
           method: "POST",
           headers: {
             "Content-Type": "application/json",
           },
-          body: JSON.stringify({ prompt: userMessage.content }),
+          body: JSON.stringify(requestBody),
         })
 
         const result = await response.json()
@@ -386,6 +430,14 @@ export default function ChatPage() {
             timestamp: new Date()
           }
           setMessages(prev => [...prev, assistantMessage])
+          
+          // Store the response ID if present for this endpoint
+          if (result.response_id) {
+            setPreviousResponseIds(prev => ({
+              ...prev,
+              [endpoint]: result.response_id
+            }))
+          }
         } else {
           console.error("Chat failed:", result.error)
           const errorMessage: Message = {
@@ -526,7 +578,7 @@ export default function ChatPage() {
                 <Button
                   variant={endpoint === "chat" ? "default" : "ghost"}
                   size="sm"
-                  onClick={() => setEndpoint("chat")}
+                  onClick={() => handleEndpointChange("chat")}
                   className="h-7 text-xs"
                 >
                   Chat
@@ -534,7 +586,7 @@ export default function ChatPage() {
                 <Button
                   variant={endpoint === "langflow" ? "default" : "ghost"}
                   size="sm"
-                  onClick={() => setEndpoint("langflow")}
+                  onClick={() => handleEndpointChange("langflow")}
                   className="h-7 text-xs"
                 >
                   Langflow

src/agent.py

@@ -1,5 +1,11 @@
 messages = [{"role": "system", "content": "You are a helpful assistant. Always use the search_tools to answer questions."}]
 
+# Simple session store for conversation state
+conversation_state = {
+    "messages": messages,
+    "previous_response_id": None
+}
+
 # Generic async response function for streaming
 async def async_response_stream(client, prompt: str, model: str, previous_response_id: str = None, log_prefix: str = "response"):
     print(f"user ==> {prompt}")
@@ -79,7 +85,15 @@ async def async_response(client, prompt: str, model: str, previous_response_id:
     
     response_text = response.output_text
     print(f"{log_prefix} ==> {response_text}")
-    return response_text
+    
+    # Extract and store response_id if available
+    response_id = getattr(response, 'id', None) or getattr(response, 'response_id', None)
+    if response_id:
+        global conversation_state
+        conversation_state["previous_response_id"] = response_id
+        print(f"Stored response_id: {response_id}")
+    
+    return response_text, response_id
 
 # Unified streaming function for both chat and langflow
 async def async_stream(client, prompt: str, model: str, previous_response_id: str = None, log_prefix: str = "response"):
@@ -87,14 +101,15 @@ async def async_stream(client, prompt: str, model: str, previous_response_id: st
         yield chunk
 
 # Async langflow function (non-streaming only)
-async def async_langflow(langflow_client, flow_id: str, prompt: str):
+async def async_langflow(langflow_client, flow_id: str, prompt: str, previous_response_id: str = None):
-    return await async_response(langflow_client, prompt, flow_id, log_prefix="langflow")
+    response_text, response_id = await async_response(langflow_client, prompt, flow_id, previous_response_id=previous_response_id, log_prefix="langflow")
+    return response_text, response_id
 
 # Async langflow function for streaming (alias for compatibility)
-async def async_langflow_stream(langflow_client, flow_id: str, prompt: str):
+async def async_langflow_stream(langflow_client, flow_id: str, prompt: str, previous_response_id: str = None):
     print(f"[DEBUG] Starting langflow stream for prompt: {prompt}")
     try:
-        async for chunk in async_stream(langflow_client, prompt, flow_id, log_prefix="langflow"):
+        async for chunk in async_stream(langflow_client, prompt, flow_id, previous_response_id=previous_response_id, log_prefix="langflow"):
             print(f"[DEBUG] Yielding chunk from langflow_stream: {chunk[:100]}...")
             yield chunk
         print(f"[DEBUG] Langflow stream completed")
@@ -106,13 +121,47 @@ async def async_langflow_stream(langflow_client, flow_id: str, prompt: str):
 
 # Async chat function (non-streaming only)
 async def async_chat(async_client, prompt: str, model: str = "gpt-4.1-mini", previous_response_id: str = None):
-    global messages
+    global conversation_state
-    messages += [{"role": "user", "content": prompt}]
+    
-    return await async_response(async_client, prompt, model, previous_response_id=previous_response_id, log_prefix="agent")
+    # If no previous_response_id is provided, reset conversation state
+    if previous_response_id is None:
+        conversation_state["messages"] = [{"role": "system", "content": "You are a helpful assistant. Always use the search_tools to answer questions."}]
+        conversation_state["previous_response_id"] = None
+    
+    # Add user message to conversation
+    conversation_state["messages"].append({"role": "user", "content": prompt})
+    
+    response_text, response_id = await async_response(async_client, prompt, model, previous_response_id=previous_response_id, log_prefix="agent")
+    
+    # Add assistant response to conversation
+    conversation_state["messages"].append({"role": "assistant", "content": response_text})
+    
+    return response_text, response_id
 
 # Async chat function for streaming (alias for compatibility)
 async def async_chat_stream(async_client, prompt: str, model: str = "gpt-4.1-mini", previous_response_id: str = None):
-    global messages
+    global conversation_state
-    messages += [{"role": "user", "content": prompt}]
+    
+    # If no previous_response_id is provided, reset conversation state
+    if previous_response_id is None:
+        conversation_state["messages"] = [{"role": "system", "content": "You are a helpful assistant. Always use the search_tools to answer questions."}]
+        conversation_state["previous_response_id"] = None
+    
+    # Add user message to conversation
+    conversation_state["messages"].append({"role": "user", "content": prompt})
+    
+    full_response = ""
     async for chunk in async_stream(async_client, prompt, model, previous_response_id=previous_response_id, log_prefix="agent"):
+        # Extract text content to build full response for history
+        try:
+            import json
+            chunk_data = json.loads(chunk.decode('utf-8'))
+            if 'delta' in chunk_data and 'content' in chunk_data['delta']:
+                full_response += chunk_data['delta']['content']
+        except:
+            pass
         yield chunk
+    
+    # Add the complete assistant response to message history
+    if full_response:
+        conversation_state["messages"].append({"role": "assistant", "content": full_response})

src/app.py

@@ -305,6 +305,7 @@ async def search_tool(query: str)-> dict[str, Any]:
 async def chat_endpoint(request):
     data = await request.json()
     prompt = data.get("prompt", "")
+    previous_response_id = data.get("previous_response_id")
     stream = data.get("stream", False)
 
     if not prompt:
@@ -313,7 +314,7 @@ async def chat_endpoint(request):
     if stream:
         from agent import async_chat_stream
         return StreamingResponse(
-            async_chat_stream(patched_async_client, prompt),
+            async_chat_stream(patched_async_client, prompt, previous_response_id=previous_response_id),
             media_type="text/event-stream",
             headers={
                 "Cache-Control": "no-cache",
@@ -323,12 +324,16 @@ async def chat_endpoint(request):
             }
         )
     else:
-        response = await async_chat(patched_async_client, prompt)
+        response_text, response_id = await async_chat(patched_async_client, prompt, previous_response_id=previous_response_id)
-        return JSONResponse({"response": response})
+        response_data = {"response": response_text}
+        if response_id:
+            response_data["response_id"] = response_id
+        return JSONResponse(response_data)
 
 async def langflow_endpoint(request):
     data = await request.json()
     prompt = data.get("prompt", "")
+    previous_response_id = data.get("previous_response_id")
     stream = data.get("stream", False)
     
     if not prompt:
@@ -341,7 +346,7 @@ async def langflow_endpoint(request):
         if stream:
             from agent import async_langflow_stream
             return StreamingResponse(
-                async_langflow_stream(langflow_client, flow_id, prompt),
+                async_langflow_stream(langflow_client, flow_id, prompt, previous_response_id=previous_response_id),
                 media_type="text/event-stream",
                 headers={
                     "Cache-Control": "no-cache",
@@ -351,8 +356,11 @@ async def langflow_endpoint(request):
                 }
             )
         else:
-            response = await async_langflow(langflow_client, flow_id, prompt)
+            response_text, response_id = await async_langflow(langflow_client, flow_id, prompt, previous_response_id=previous_response_id)
-            return JSONResponse({"response": response})
+            response_data = {"response": response_text}
+            if response_id:
+                response_data["response_id"] = response_id
+            return JSONResponse(response_data)
         
     except Exception as e:
         return JSONResponse({"error": f"Langflow request failed: {str(e)}"}, status_code=500)