Update app.py

app.py

@@ -264,12 +264,11 @@ def chatbot_respond(
     uploaded_files,
 ) -> Generator[List, None, None]:
     """
-    Called on every user message.  history is the Gradio chatbot history
-    (list of [user_msg, bot_msg] pairs).  uploaded_files is the current
-    file list from gr.File.
+    Called on every user message.  history is in Gradio 6 "messages" format:
+    list of {"role": "user"|"assistant", "content": "..."} dicts.
+    uploaded_files is the current file list from gr.File.
 
-    Yields the updated history list after each token chunk, giving a
-    streaming effect in the Gradio Chatbot.
+    Yields the updated history list after each token chunk for streaming.
     """
     # ---- 1. Parse any newly uploaded files ----
     file_context = parse_all_files(uploaded_files)
@@ -277,7 +276,7 @@ def chatbot_respond(
     # ---- 2. Build the message list for llama-cpp-python (ChatML format) ----
     messages = [{"role": "system", "content": SYSTEM_PROMPT}]
 
-    # If files were uploaded, inject their content as a system-level note
+    # If files were uploaded, inject their content
     if file_context:
         messages.append({
             "role": "user",
@@ -295,21 +294,21 @@ def chatbot_respond(
             ),
         })
 
-    # Append the real conversation history
-    for user_msg, bot_msg in history:
-        if user_msg:
-            messages.append({"role": "user", "content": user_msg})
-        if bot_msg:
-            messages.append({"role": "assistant", "content": bot_msg})
+    # Append the real conversation history (Gradio 6 "messages" format)
+    for entry in history:
+        role = entry.get("role", "user")
+        content = entry.get("content", "")
+        if content:
+            messages.append({"role": role, "content": content})
 
-    # Append the current message
+    # Append the current user message
     messages.append({"role": "user", "content": message})
 
     # ---- 3. Stream the completion ----
     llm = _get_llm()
     stream = llm.create_chat_completion(
         messages=messages,
-        temperature=0.6,      # lower = more precise for coding
+        temperature=0.6,
         top_p=0.8,
         top_k=20,
         max_tokens=4096,
@@ -325,11 +324,13 @@ def chatbot_respond(
             content = delta.get("content", "")
             if content:
                 partial += content
-        # YIELD the updated history so Gradio streams into the chat UI
-        yield history + [[message, partial]]
+        # YIELD in Gradio 6 "messages" format
+        yield history + [{"role": "user", "content": message},
+                         {"role": "assistant", "content": partial}]
 
-    # Final yield to guarantee the full message is rendered
-    yield history + [[message, partial]]
+    # Final yield
+    yield history + [{"role": "user", "content": message},
+                     {"role": "assistant", "content": partial}]
 
 
 # ===========================================================================
@@ -337,13 +338,7 @@ def chatbot_respond(
 # ===========================================================================
 
 def create_demo() -> gr.Blocks:
-    css = """
-    .file-upload-col { background: var(--background-fill-secondary); border-radius: 12px; padding: 16px; }
-    footer { display: none !important; }
-    """
     with gr.Blocks(
-        css=css,
-        theme=gr.themes.Soft(primary_hue="violet", secondary_hue="slate"),
         title="Qwen 3.6 27B CoderBot",
     ) as demo:
 
@@ -361,7 +356,7 @@ def create_demo() -> gr.Blocks:
                 chatbot = gr.Chatbot(
                     label="Chat",
                     height=580,
-                    bubble_full_width=False,
+                    type="messages",  # Gradio 6 uses "messages" or "tuples"
                     avatar_images=(
                         None,
                         "https://huggingface.co/front/assets/huggingface_logo-noborder.svg",
@@ -441,8 +436,8 @@ def create_demo() -> gr.Blocks:
             outputs=[chatbot],
         ).then(lambda: "", None, [msg])
 
-        # Clear chat
-        clear_btn.click(lambda: None, None, chatbot, queue=False)
+        # Clear chat — return empty list (Gradio 6 "messages" format)
+        clear_btn.click(lambda: [], None, chatbot, queue=False)
 
         # File upload feedback
         files.change(update_file_info, files, [uploaded_info, file_content_state])
@@ -457,4 +452,9 @@ def create_demo() -> gr.Blocks:
 if __name__ == "__main__":
     demo = create_demo()
     demo.queue(default_concurrency_limit=1, max_size=4)
-    demo.launch()
+    demo.launch(
+        css="""
+        .file-upload-col { background: var(--background-fill-secondary); border-radius: 12px; padding: 16px; }
+        footer { display: none !important; }
+        """,
+    )