Update app.py

app.py

@@ -2,8 +2,8 @@ import gradio as gr
 from huggingface_hub import InferenceClient
 import os
 
-# 🔹 Load HF token from secret
-HF_TOKEN = os.environ.get('telemedpro')  # store your HF token in Space Secrets
+# 🔹 Load HF token from Space Secrets
+HF_TOKEN = os.environ.get('telemedpro')  # Add your token in Space Settings → Secrets
 
 # 🔹 Fixed persona system message
 PERSONA_MESSAGE = (
@@ -13,14 +13,15 @@ PERSONA_MESSAGE = (
     "You behave politely, patiently, and with care, like a trusted family doctor."
 )
 
-# 🔹 Create InferenceClient once
+# 🔹 Initialize InferenceClient once
 client = InferenceClient(token=HF_TOKEN, model="m42-health/Llama3-Med42-70B")
 
-def respond(message, history, max_tokens, temperature, top_p):
+# 🔹 Respond function (non-streaming for stability)
+def respond(message, history, max_tokens=512, temperature=0.7, top_p=0.95):
     try:
         messages = [{"role": "system", "content": PERSONA_MESSAGE}]
 
-        # Add previous conversation
+        # Append previous conversation safely
         if history:
             for h in history:
                 user_msg = h[0] if h[0] else ""
@@ -28,25 +29,21 @@ def respond(message, history, max_tokens, temperature, top_p):
                 messages.append({"role": "user", "content": user_msg})
                 messages.append({"role": "assistant", "content": ai_msg})
 
-        # Add current user message
+        # Append current user message
         messages.append({"role": "user", "content": message})
 
-        response = ""
-        # Stream model output token by token
-        for msg in client.chat_completion(
+        # 🔹 Non-streaming call for stability
+        result = client.chat_completion(
             messages,
             max_tokens=max_tokens,
-            stream=True,
             temperature=temperature,
-            top_p=top_p,
-        ):
-            if msg.choices and msg.choices[0].delta.content:
-                token = msg.choices[0].delta.content
-                response += token
-                yield response
+            top_p=top_p
+        )
+
+        return result.choices[0].message.content
 
     except Exception as e:
-        yield f"⚠️ Space error: {e}"
+        return f"⚠️ Space error: {e}"
 
 # 🔹 Gradio Chat Interface
 chatbot = gr.ChatInterface(