Update app.py

app.py

@@ -24,9 +24,6 @@ prompt = PromptTemplate(
     template=template
 )
 
-# Memory (Conversation history)
-memory = ConversationBufferMemory(memory_key="chat_history")
-
 # LLM with API Key
 llm = ChatOpenAI(
     temperature=0.5,
@@ -34,28 +31,33 @@ llm = ChatOpenAI(
     api_key=openai_api_key
 )
 
-# LLM Chain
-llm_chain = LLMChain(
-    llm=llm,
-    prompt=prompt,
-    verbose=True,
-    memory=memory
-)
+# Function with per-user memory
+def get_text_response(user_message, history, memory_state):
+    if memory_state is None:
+        memory_state = ConversationBufferMemory(memory_key="chat_history")
+
+    llm_chain = LLMChain(
+        llm=llm,
+        prompt=prompt,
+        verbose=True,
+        memory=memory_state
+    )
 
-# Chat Function
-def get_text_response(user_message, history):
     response = llm_chain.predict(user_message=user_message)
-    return response
-
-# Gradio Chat App
-demo = gr.ChatInterface(
-    fn=get_text_response,
-    examples=[
-        "How are you doing?",
-        "What are your interests?",
-        "Which places do you like to visit?"
-    ]
-)
+    return response, memory_state
+
+# Gradio Chat App with State (per-user memory)
+with gr.Blocks() as demo:
+    chatbot = gr.Chatbot()
+    msg = gr.Textbox(placeholder="Ask me anything...")
+    memory_state = gr.State()  # 👈 each user gets their own memory
+
+    def respond(message, history, memory_state):
+        reply, memory_state = get_text_response(message, history, memory_state)
+        history = history + [[message, reply]]
+        return history, history, memory_state
+
+    msg.submit(respond, [msg, chatbot, memory_state], [chatbot, chatbot, memory_state])
 
 if __name__ == "__main__":
     demo.launch(share=True)