Handle overload error.

Signed-off-by: Aivin V. Solatorio <avsolatorio@gmail.com>

mcp_client.py

@@ -9,6 +9,7 @@ from gradio.components.chatbot import ChatMessage
 from mcp import ClientSession, StdioServerParameters
 from mcp.client.stdio import stdio_client
 from anthropic import Anthropic
+from anthropic._exceptions import OverloadedError
 from dotenv import load_dotenv
 
 load_dotenv()
@@ -150,14 +151,22 @@ class MCPClientWrapper:
 
         claude_messages.append({"role": "user", "content": message})
 
-        response = self.anthropic.messages.create(
-            # model="claude-3-5-sonnet-20241022",
-            model=LLM_MODEL,
-            system=SYSTEM_PROMPT,
-            max_tokens=1000,
-            messages=claude_messages,
-            tools=self.tools,
-        )
+        try:
+            response = self.anthropic.messages.create(
+                # model="claude-3-5-sonnet-20241022",
+                model=LLM_MODEL,
+                system=SYSTEM_PROMPT,
+                max_tokens=1000,
+                messages=claude_messages,
+                tools=self.tools,
+            )
+        except OverloadedError:
+            return [
+                {
+                    "role": "assistant",
+                    "content": "The LLM API is overloaded now, try again later...",
+                }
+            ]
 
         result_messages = []
 
@@ -272,12 +281,21 @@ class MCPClientWrapper:
                         ],
                     }
                 )
-                next_response = self.anthropic.messages.create(
-                    model=LLM_MODEL,
-                    system=SYSTEM_PROMPT,
-                    max_tokens=1000,
-                    messages=claude_messages,
-                )
+
+                try:
+                    next_response = self.anthropic.messages.create(
+                        model=LLM_MODEL,
+                        system=SYSTEM_PROMPT,
+                        max_tokens=1000,
+                        messages=claude_messages,
+                    )
+                except OverloadedError:
+                    return [
+                        {
+                            "role": "assistant",
+                            "content": "The LLM API is overloaded now, try again later...",
+                        }
+                    ]
 
                 print("next_response", next_response.content)