streamlit_client.py•3.42 kB
import traceback
from contextlib import AsyncExitStack
import os
import json
# MCP Client Imports
from mcp import ClientSession, StdioServerParameters
from mcp.client.stdio import stdio_client
# Langchain Imports
from langchain_mcp_adapters.tools import load_mcp_tools
from langgraph.prebuilt import create_react_agent
from langchain_anthropic import ChatAnthropic
class CustomEncoder(json.JSONEncoder):
def default(self, o):
if hasattr(o, 'content'):
return {'type': o.__class__.__name__, 'content': o.content}
return super().default(o)
# Ensure correct config path
os.environ["mcp_config_2"] = os.path.join(os.path.dirname(__file__), "mcp_config_2.json")
def read_config_json():
config_path = os.getenv('mcp_config_2')
if not config_path:
script_dir = os.path.dirname(os.path.abspath(__file__))
config_path = os.path.join(script_dir, 'mcp_config_2.json')
with open(config_path, 'r') as f:
return json.load(f)
async def run_agent(logs):
# OpenAI GPT 4 LLM Integration
llm = ChatAnthropic(
model='claude-sonnet-4-20250514',
temperature=0,
max_retries=2,
anthropic_api_key='Your-API-Key'
)
configu = read_config_json()
mcp_servers = configu.get('mcpServers', {})
try:
async with AsyncExitStack() as stack:
tools = []
for server_name, server_info in mcp_servers.items():
print(f"🔌 Connecting to MCP server")
server_params = StdioServerParameters(
command=server_info['command'],
args=server_info['args']
)
try:
read, write = await stack.enter_async_context(stdio_client(server_params))
session = await stack.enter_async_context(ClientSession(read, write))
print("✅ MCP subprocess started, waiting for session initialization...")
await session.initialize()
server_tools = await load_mcp_tools(session)
except Exception as e:
print(f"🚨 MCP Connection Failed for {server_name}")
traceback.print_exception(type(e), e, e.__traceback__)
raise e
for tool in server_tools:
print(f"✅ Loaded tool: {tool.name}")
tools.append(tool)
print(f"📦 {len(server_tools)} tools loaded from {server_name}.")
agent = create_react_agent(llm, tools)
try:
print("🤖 Invoking agent...")
response = await agent.ainvoke({
"messages": [
{"role": "user",
"content": f"Please analyze these logs: {json.dumps(logs, indent=2)} and suggest fixes."}
]
})
print("✅ Agent response received")
return response
except Exception as e:
print("❌ Agent invocation failed:", e)
traceback.print_exc()
return {"error": "Agent invocation failed", "details": traceback.format_exc()}
except Exception as e:
print("🚨 Unexpected exception in run_agent()")
traceback.print_exc()
return {"error": str(e), "details": traceback.format_exc()}