Update app.py

app.py

@@ -31,7 +31,8 @@ try:
             model_path=MODEL_PATH,
             n_ctx=512,  # ✅ Lower memory usage, speeds up responses
             n_threads=2,  # Matches available vCPUs
-            numa=True  # Optimize CPU memory access
+            numa=True,
+            n_batch=32  # ✅ Faster token processing
         )
         st.write("✅ Model loaded successfully!")
 except Exception as e:
@@ -64,24 +65,27 @@ if st.button("Send") and user_input:
     # ✅ Use a minimal prompt format (no system message)
     formatted_messages = [{"role": "user", "content": user_input}]
 
-    # ✅ Streamed response for faster user experience
+    # ✅ Disable streaming for debugging
     response_data = st.session_state["model"].create_chat_completion(
         messages=formatted_messages, 
-        max_tokens=256, temperature=0.7, top_p=0.9,
-        stream=True  # ✅ Enables real-time streaming
+        max_tokens=128, temperature=0.7, top_p=0.9,
+        stream=False  # ❌ Disabled streaming for debugging
     )
 
-    response_text = ""
-    response_container = st.empty()  # Placeholder for live updates
-
-    for chunk in response_data:
-        if "choices" in chunk and len(chunk["choices"]) > 0:
-            choice = chunk["choices"][0]
-            if "message" in choice:
-                response_text += choice["message"]["content"]
-                response_container.markdown(f"**AI:** {response_text}")
-            if choice.get("finish_reason") == "stop":
-                break
+    # ✅ Debugging output
+    st.write("🔍 Debug: Raw Model Response:", response_data)
+
+    if "choices" in response_data and len(response_data["choices"]) > 0:
+        choice = response_data["choices"][0]
+        if "message" in choice and "content" in choice["message"]:
+            response_text = choice["message"]["content"].strip()
+            st.session_state["messages"].append(("assistant", response_text))
+            st.chat_message("assistant").write(response_text)
+        else:
+            st.error("⚠️ No valid response content found.")
+    else:
+        st.error("⚠️ Model did not return any choices.")
+