Try to change output to stream instead of wait for completion

app.py

@@ -19,16 +19,16 @@ def read_dataset(dataset):
 
 context = read_dataset(dataset)
 
-# Inject custom CSS to change the background color to yellow
+# Inject custom CSS
 st.markdown(
     """
     <style>
     .scrollable-div { 
-        max-height: 300px; 
+        max-height: 400px; 
         overflow-y: auto; 
         padding: 10px; 
         border: 1px solid #ccc; 
-        background-color: #f9f9f9; }
+    }
     </style>
     """,
     unsafe_allow_html=True
@@ -49,19 +49,22 @@ def handle_submit():
                 {"role": "user", "content": user_input}
             ]
 
-        completion = client.chat.completions.create(
+        response = client.chat.completions.create(
             model="Qwen/Qwen2.5-72B-Instruct",
             #model="Qwen/Qwen2.5-Coder-32B-Instruct",
             #model="mistralai/Mistral-7B-Instruct-v0.3",
             #model="meta-llama/Llama-3.1-8B-Instruct",
             messages=messages,
-            max_tokens=1000
+            max_tokens=1000,
+            stream=True
         )
         
-        answer = completion.choices[0].message['content']
+        answer = ""
         
-        #st.session_state.conversation = f"<p><strong>User:</strong> {user_input}</p><p><strong>Adrega AI:</strong> {answer}</p>" + st.session_state.conversation       
-        st.session_state.conversation = f"<p><strong>Adrega AI:</strong> {answer}</p>" + st.session_state.conversation       
+        for chunk in response: 
+            answer += chunk['choices'][0]['delta']['content'] 
+            st.session_state.conversation = f"<p><strong>Adrega AI:</strong> {answer}</p>" + st.session_state.conversation 
+            st.markdown(f'<div class="scrollable-div">{st.session_state.conversation}</div>', unsafe_allow_html=True)       
     else:
         st.session_state.conversation(f"<p><strong>Adrega AI:</strong>: Please enter a question.")