changes to the model names

templates/chatbot.py

@@ -42,9 +42,10 @@ if prompt := st.chat_input(CHAT_PLACEHOLDER):
 
         def get_stream():
             for chunk in stream:
-                content = chunk.choices[0].delta.content
-                if content:
-                    yield content
+                delta = chunk.choices[0].delta
+                # Skip reasoning_content from thinking models — yield only final answer
+                if delta.content:
+                    yield delta.content
 
         response = st.write_stream(get_stream())
         st.session_state.messages.append({{"role": "assistant", "content": response}})

templates/comparison.py

@@ -1,7 +1,6 @@
 COMPARISON_TEMPLATE = '''import streamlit as st
 from openai import OpenAI
 import os
-import threading
 
 TITLE = "{title}"
 DESCRIPTION = "{description}"
@@ -21,51 +20,70 @@ st.title(TITLE)
 st.caption(DESCRIPTION)
 st.divider()
 
-prompt = st.text_area("Enter your prompt:", height=120, placeholder="Ask anything to compare both models...")
+with st.sidebar:
+    st.caption("Powered by [Crusoe](https://crusoe.ai)")
+    st.markdown(f"**Model A:** `{{MODEL_A}}`")
+    st.markdown(f"**Model B:** `{{MODEL_B}}`")
+
+prompt = st.text_area(
+    "Enter your prompt:",
+    height=120,
+    placeholder="Ask anything to compare both models...",
+)
+
+
+def _stream_response(model: str) -> str:
+    """Stream a response from a model, handling both content and reasoning_content."""
+    full_response = ""
+    thinking_placeholder = st.empty()
+    answer_placeholder = st.empty()
+    thinking_text = ""
+    answer_text = ""
+
+    stream = client.chat.completions.create(
+        model=model,
+        messages=[
+            {{"role": "system", "content": SYSTEM_PROMPT}},
+            {{"role": "user", "content": prompt}},
+        ],
+        stream=True,
+    )
+
+    for chunk in stream:
+        delta = chunk.choices[0].delta
+
+        # Handle reasoning/thinking content (DeepSeek R1, Kimi-K2-Thinking, etc.)
+        reasoning = getattr(delta, "reasoning_content", None)
+        if reasoning:
+            thinking_text += reasoning
+            thinking_placeholder.markdown(
+                f"<details><summary>💭 Thinking...</summary>\n\n{thinking_text}\n\n</details>",
+                unsafe_allow_html=True,
+            )
+
+        # Handle final answer content
+        if delta.content:
+            answer_text += delta.content
+            answer_placeholder.markdown(answer_text)
+            full_response += delta.content
+
+    return full_response
+
 
 if st.button("⚡ Compare Models", type="primary", disabled=not prompt):
     col1, col2 = st.columns(2)
 
     with col1:
         st.subheader(f"🤖 {MODEL_A_LABEL}")
-        stream_a = client.chat.completions.create(
-            model=MODEL_A,
-            messages=[
-                {{"role": "system", "content": SYSTEM_PROMPT}},
-                {{"role": "user", "content": prompt}},
-            ],
-            stream=True,
-        )
-
-        def stream_a_content():
-            for chunk in stream_a:
-                content = chunk.choices[0].delta.content
-                if content:
-                    yield content
-
-        st.write_stream(stream_a_content())
+        try:
+            _stream_response(MODEL_A)
+        except Exception as e:
+            st.error(f"Error: {{e}}")
 
     with col2:
         st.subheader(f"🤖 {MODEL_B_LABEL}")
-        stream_b = client.chat.completions.create(
-            model=MODEL_B,
-            messages=[
-                {{"role": "system", "content": SYSTEM_PROMPT}},
-                {{"role": "user", "content": prompt}},
-            ],
-            stream=True,
-        )
-
-        def stream_b_content():
-            for chunk in stream_b:
-                content = chunk.choices[0].delta.content
-                if content:
-                    yield content
-
-        st.write_stream(stream_b_content())
-
-with st.sidebar:
-    st.caption("Powered by [Crusoe](https://crusoe.ai)")
-    st.markdown(f"**Model A:** `{MODEL_A}`")
-    st.markdown(f"**Model B:** `{MODEL_B}`")
+        try:
+            _stream_response(MODEL_B)
+        except Exception as e:
+            st.error(f"Error: {{e}}")
 '''

templates/dashboard.py

@@ -40,9 +40,10 @@ with col_output:
 
         def get_stream():
             for chunk in stream:
-                content = chunk.choices[0].delta.content
-                if content:
-                    yield content
+                delta = chunk.choices[0].delta
+                # Skip reasoning_content from thinking models — yield only final answer
+                if delta.content:
+                    yield delta.content
 
         st.write_stream(get_stream())
     else: