Update app.py

app.py

@@ -161,7 +161,6 @@ async def stream_answer(
                 ctype = chunk.get("type")
 
                 if ctype == "tool_log":
-                    # Example: {"type": "tool_log", "tool": "...", "status": "started/finished"}
                     name = chunk.get("tool", "tool")
                     status = chunk.get("status", "")
                     _append_log(tool_log, f"- {name} **{status}**")
@@ -174,7 +173,6 @@ async def stream_answer(
                     yield {"delta": chunk.get("text", ""), "tool_log": _format_tool_log(tool_log), "citations": _format_citations(citations)}
 
                 elif ctype == "tool_result":
-                    # Try to capture a useful citation label and optional URL if present
                     tool_name = chunk.get("tool", "tool")
                     content = chunk.get("content")
                     url = None
@@ -205,7 +203,6 @@ async def stream_answer(
                     }
 
             else:
-                # Fallback if provider yields plain strings
                 yield {"delta": str(chunk), "tool_log": _format_tool_log(tool_log), "citations": _format_citations(citations)}
 
     except Exception as e:
@@ -233,7 +230,7 @@ with gr.Blocks(fill_height=True) as demo:
             chat = gr.Chatbot(
                 label="Gradio Docs Assistant",
                 height=520,
-                type="messages",   # expects: [{"role": "...", "content": "..."}]
+                type="messages",
             )
             with gr.Row():
                 msg = gr.Textbox(
@@ -270,12 +267,10 @@ with gr.Blocks(fill_height=True) as demo:
           - tool activity
           - citations
         """
-        # Start a new assistant message for streaming
         history_msgs = (history_msgs or []) + [{"role": "user", "content": user_msg}]
         history_msgs.append({"role": "assistant", "content": ""})
         yield history_msgs, gr.update(value="_No tool activity yet._"), gr.update(value="_No citations captured yet._")
 
-        # Compose messages for LLM
         messages_for_llm = to_llm_messages(history_msgs[:-1], user_msg, style_choice)
 
         async for chunk in stream_answer(messages_for_llm, MODEL_ID, PROVIDER, HF_TOKEN):
@@ -284,22 +279,18 @@ with gr.Blocks(fill_height=True) as demo:
                 history_msgs[-1]["content"] += delta
             yield history_msgs, gr.update(value=chunk.get("tool_log", "")), gr.update(value=chunk.get("citations", ""))
 
-    # Wire both Enter and button click; also pass "style"
     msg.submit(on_submit, inputs=[msg, chat, style], outputs=[chat, tool_log_md, citations_md], queue=True)
     send_btn.click(on_submit, inputs=[msg, chat, style], outputs=[chat, tool_log_md, citations_md], queue=True)
 
 # ----------------------------
 # Gradio runtime (queue + launch)
 # ----------------------------
-# Enable Gradio’s task queue (important for async streaming)
-demo.queue(max_size=32)
-
-# On Spaces, Gradio auto-serves; only call launch() for local runs.
-if __name__ == "__main__":
-    on_spaces = bool(os.environ.get("SPACE_ID"))
-    if not on_spaces:
-        demo.launch(
-            ssr_mode=False,       # disable SSR to prevent Node server start/stop churn
-            server_name="0.0.0.0",
-            server_port=7860,
-        )
+# IMPORTANT: assign the queued app back to 'demo' for older Gradio versions.
+demo = demo.queue(max_size=32)
+
+# Always launch; Spaces runs this script directly.
+demo.launch(
+    ssr_mode=False,       # disable SSR to avoid Node helper churn
+    server_name="0.0.0.0",
+    server_port=7860,
+)