Update app.py

app.py

@@ -2,25 +2,29 @@ import os
 import gradio as gr
 from openai import OpenAI
 
+# API endpoints
 API_HOSTS = {
     "Domestic": "https://api.chatanywhere.tech/v1",
     "Overseas": "https://api.chatanywhere.org/v1"
 }
 
+# Model info
 MODELS_INFO = {
-    "gpt-3.5-turbo": {"input_price": "0.0035", "output_price": "0.0105", "features": "Default, fast, affordable"},
+    "gpt-3.5-turbo": {"input_price": "0.0035", "output_price": "0.0105", "features": "Fast, affordable"},
     "gpt-4o": {"input_price": "0.0175", "output_price": "0.07", "features": "Cheaper & faster GPT-4O"},
     "gpt-4-turbo": {"input_price": "0.07", "output_price": "0.21", "features": "Multimodal, tool use"},
     "gpt-4o-ca": {"input_price": "0.01", "output_price": "0.04", "features": "CA variant, daily free limit"},
 }
 
 def create_client(host):
+    """Create OpenAI API client with given host."""
     key = os.getenv("OPENAI_API_KEY")
     if not key:
         raise ValueError("Missing environment variable: OPENAI_API_KEY")
     return OpenAI(api_key=key, base_url=host)
 
 def get_model_card(model_name):
+    """Return markdown info for the selected model."""
     info = MODELS_INFO.get(model_name, {})
     if not info:
         return "Model info not available."
@@ -32,7 +36,9 @@ def get_model_card(model_name):
     )
 
 def respond(user, history, host_choice, model_name, temperature, top_p, max_tokens, sys_prompt):
+    """Main chat handler with streaming and error handling."""
     history = history or []
+
     if not user.strip():
         yield history + [("", "⚠️ Please enter a message.")]
         return
@@ -61,12 +67,25 @@ def respond(user, history, host_choice, model_name, temperature, top_p, max_toke
 
         partial = ""
         history.append((user, partial))
-        yield history
+        yield history  # initial blank assistant message
 
         for chunk in stream:
-            delta = chunk.choices[0].delta.content or ""
-            partial += delta
-            history[-1] = (user, partial)
+            try:
+                if not chunk.choices:
+                    continue
+                choice = chunk.choices[0]
+                if not hasattr(choice, "delta") or not choice.delta:
+                    continue
+                delta = getattr(choice.delta, "content", "") or ""
+                if delta:
+                    partial += delta
+                    history[-1] = (user, partial)
+                    yield history
+            except Exception:
+                continue  # skip malformed chunks
+
+        if not partial.strip():
+            history[-1] = (user, "⚠️ No response received from the model.")
             yield history
 
     except Exception as e:
@@ -76,11 +95,16 @@ def respond(user, history, host_choice, model_name, temperature, top_p, max_toke
                 "🚫 Daily quota reached for this model.\n"
                 "Please try again after 00:00 China time or switch model/host."
             )
+        elif "401" in err or "invalid_api_key" in err.lower():
+            out = "❌ Invalid or missing API key. Check your OPENAI_API_KEY."
+        elif "timed out" in err.lower():
+            out = "⌛ Request timed out. Please try again."
         else:
-            out = f"❌ API Error: {e}"
+            out = f"❌ API Error: {err}"
         history.append((user, out))
         yield history
 
+
 with gr.Blocks(title="ChatAnywhere Realtime Chatbot", theme=gr.themes.Soft()) as demo:
     gr.Markdown("## 💬 ChatAnywhere Realtime Chatbot\nPowered by GPT-5 via ChatAnywhere API")
     with gr.Row():