Update app.py

app.py

@@ -47,7 +47,8 @@ SERPER_API_KEY      = os.getenv("SERPER_API_KEY")
 llm = HuggingFaceInferenceAPI(
     model_name="Qwen/Qwen2.5-Coder-32B-Instruct",
     token=HF_TOKEN, 
-    task="conversational"
+    task="conversational",
+    streaming=True
 )
 
 memory = ChatMemoryBuffer.from_defaults(token_limit=8192)
@@ -246,6 +247,7 @@ tools = [
 web_agent = AgentWorkflow.from_tools_or_functions(
     tools, 
     llm=llm,
+    streaming=True,
     system_prompt="""You are a helpful assistant with access to specialized tools for retrieving information about weather, and news.
     AVAILABLE TOOLS:
     1. current_weather - Get current weather conditions for a location
@@ -310,24 +312,40 @@ async def run_query(query: str):
             session_id="web-agent-session",
             user_id=ANON_USER_ID,
         ):
-            return await web_agent.run(query, ctx=ctx)
+            response_gen = await web_agent.astream(query, ctx=ctx)
+            return response_gen
     finally:
         instrumentor.flush()
 
 # Gradio interface function
-async def gradio_query_stream(user_input, chat_history=None):
+async def gradio_query(user_input, chat_history=None):
     history = chat_history or []
     history.append({"role": "user", "content": user_input})
-    # Start streaming response from the agent
-    result = await web_agent.run(user_input, ctx=ctx, streaming=True)
-    partial = ""
-    async for chunk in result.response_gen:
-        # If chunk is an object, use chunk.delta or chunk.text
-        partial += chunk
-        if len(history) > 1 and history[-1]["role"] == "assistant":
-            history[-1]["content"] = partial
-        else:
-            history.append({"role": "assistant", "content": partial})
+    
+    # Add initial assistant message
+    history.append({"role": "assistant", "content": ""})
+    yield history, history
+    
+    # Get streaming response generator
+    response_gen = await run_query(user_input)
+    
+    # Accumulated response
+    full_response = ""
+    
+    # Process the streaming response
+    async for delta in response_gen:
+        if hasattr(delta, 'delta'):
+            # For streaming text chunks
+            full_response += delta.delta
+        elif hasattr(delta, 'response') and delta.response:
+            # For final response or tool outputs
+            if isinstance(delta.response, str):
+                full_response = delta.response
+            else:
+                full_response = str(delta.response)
+        
+        # Update the last message with accumulated text
+        history[-1]["content"] = full_response
         yield history, history
 
 # Build and launch Gradio app
@@ -344,13 +362,28 @@ with grb:
     )
     chatbot = gr.Chatbot(type="messages")
     txt = gr.Textbox(placeholder="Ask me anything...", show_label=False)
+    
+    # Modify the submit handler to work with streaming
     txt.submit(
-    gradio_query_stream, 
-    inputs=[txt, chatbot], 
-    outputs=[chatbot, chatbot],
-    stream=True
-    )   
-    gr.Button("Send").click(gradio_query_stream, [txt, chatbot], [chatbot, chatbot])
+        gradio_query, 
+        inputs=[txt, chatbot], 
+        outputs=[chatbot, chatbot]
+    ).then(
+        lambda: gr.update(value=""),  # Clear the textbox after submission
+        None,
+        [txt]
+    )
+    
+    # Also update the button click handler
+    gr.Button("Send").click(
+        gradio_query, 
+        [txt, chatbot], 
+        [chatbot, chatbot]
+    ).then(
+        lambda: gr.update(value=""),  # Clear the textbox after submission
+        None,
+        [txt]
+    )
 
 if __name__ == "__main__":
     grb.launch()