refactor: Simplify chat message handling in the Gradio application by consolidating user message updates and tool feedback processing, enhancing clarity in message history and improving the overall structure of the chat response pipeline.

app.py

@@ -4,6 +4,7 @@ Gradio + OpenAI MCP Connector — Clean, Fast, Streaming, With File Upload
 
 import os
 import shutil
+
 import gradio as gr
 from openai import OpenAI
 
@@ -62,33 +63,27 @@ def handle_upload(file_obj, request: gr.Request):
 # MAIN CHAT HANDLER
 # ---------------------
 def chat_send_stream(user_msg, history, file_url):
-    """
-    2-phase pipeline:
-        PHASE 1 ➜ Non-streaming tool resolution using gpt-5-mini
-        PHASE 2 ➜ Streaming final output using gpt-5
-    """
-
+    # history always starts as list of dicts
     if history is None:
         history = []
 
-    # Build message history for OpenAI
+    # Append the user's message
+    history.append({"role": "user", "content": user_msg})
+
+    # Build OpenAI message history
     messages = [{"role": "system", "content": SYSTEM_PROMPT}]
-    for u, a in history:
-        messages.append({"role": "user", "content": u})
-        if a:
-            messages.append({"role": "assistant", "content": a})
+    messages.extend(history)
 
     # Inject file context
+    final_user_msg = user_msg
     if file_url:
-        user_msg_full = f"[Uploaded CSV file: {file_url}]\n\n{user_msg}"
-    else:
-        user_msg_full = user_msg
+        final_user_msg = f"[Uploaded CSV file: {file_url}]\n\n{user_msg}"
 
-    messages.append({"role": "user", "content": user_msg_full})
+    messages[-1] = {"role": "user", "content": final_user_msg}
 
-    # -----------------------------
+    # ----------------------------------
     # PHASE 1 — TOOL RESOLUTION
-    # -----------------------------
+    # ----------------------------------
     tool_phase = client.responses.create(
         model=MODEL_FAST,
         reasoning={"effort": "low"},
@@ -99,13 +94,12 @@ def chat_send_stream(user_msg, history, file_url):
 
     tool_feedback = []
 
-    # Detect tool calls (if any)
     if tool_phase.output:
         for item in tool_phase.output:
             if item.type == "tool_call":
                 tool_feedback.append(f"🛠️ Used tool `{item.name}`.")
             elif item.type == "tool_result":
-                tool_feedback.append(f"{item.content}")
+                tool_feedback.append(str(item.content))
 
         if not tool_feedback:
             tool_feedback.append("No MCP tools needed.")
@@ -113,29 +107,29 @@ def chat_send_stream(user_msg, history, file_url):
     else:
         tool_feedback.append("No MCP tools needed.")
 
-    # Append tool results to messages before final generation
-    messages.append({"role": "assistant", "content": "\n".join(tool_feedback)})
+    # Add assistant message with tool feedback
+    history.append({"role": "assistant", "content": "\n".join(tool_feedback)})
 
-    # Yield intermediate tool feedback to the UI
-    history.append((user_msg, "\n".join(tool_feedback)))
     yield history
 
-    # -----------------------------
+    # ----------------------------------
     # PHASE 2 — STREAMING FINAL ANSWER
-    # -----------------------------
+    # ----------------------------------
+    final_msg = history[-1]["content"] + "\n\n"
+    history[-1]["content"] = final_msg
+
     stream = client.responses.create(
         model=MODEL_STREAM,
         reasoning={"effort": "low"},
         instructions=SYSTEM_PROMPT,
-        input=messages,
+        input=messages + [history[-1]],
         stream=True,
     )
 
-    final_text = ""
     for ev in stream:
         if ev.type == "response.output_text.delta":
-            final_text += ev.delta
-            history[-1] = (user_msg, "\n".join(tool_feedback) + "\n\n" + final_text)
+            final_msg += ev.delta
+            history[-1]["content"] = final_msg
             yield history
 
         elif ev.type == "response.completed":