standalone_mcp_call.py•4.85 kB
# /home/rj/Code/mcp-generated/test.py
import os
import json
from openai import OpenAI
from dotenv import load_dotenv
import asyncio # Needed for async operations
from mcp import ClientSession, StdioServerParameters, types # Import new MCP components
from mcp.client.stdio import stdio_client # Import the stdio client connector
# --- Configuration ---
load_dotenv() # Load environment variables from .env file
# Ensure your OPENAI_API_KEY is set in your environment or .env file
if not os.getenv("OPENAI_API_KEY"):
print("Error: OPENAI_API_KEY environment variable not set.")
exit(1)
openai_client = OpenAI()
mcp_server_script = "/home/rj/Code/mcp-generated/whatsapp_server.py" # Path to the server script
# Get the python executable from the virtual environment for consistency
python_executable = "/home/rj/Code/mcp-generated/.venv/bin/python3"
# --- Define the Tool for OpenAI ---
# This structure mirrors the tool defined in whatsapp_server.py
# It tells the OpenAI model what tool is available.
whatsapp_tool_definition = {
"type": "function",
"function": {
"name": "send_whatsapp",
"description": "Sends a WhatsApp message to the specified recipient number using Twilio. "
"The recipient number MUST be in E.164 format, including the 'whatsapp:' prefix "
"(e.g., 'whatsapp:+15551234567').",
"parameters": {
"type": "object",
"properties": {
"to_number": {
"type": "string",
"description": "Recipient's WhatsApp number in E.164 format, including the 'whatsapp:' prefix (e.g., whatsapp:+15551234567)"
},
"message": {
"type": "string",
"description": "The message content to send"
}
},
"required": ["to_number", "message"]
}
}
}
# --- User Request ---
# Define what you want the AI to do
user_prompt = "Please send a WhatsApp message to whatsapp:+919080573869 saying 'Hello from the OpenAI test client!'"
# --- Main Async Function ---
async def run():
print(f"--- Sending request to OpenAI ---")
print(f"Prompt: {user_prompt}")
try:
# --- Call OpenAI API ---
response = openai_client.chat.completions.create(
model="gpt-4o", # Or another model that supports tool calling like gpt-3.5-turbo
messages=[{"role": "user", "content": user_prompt}],
tools=[whatsapp_tool_definition],
tool_choice="auto" # Let the model decide if it needs to use the tool
)
response_message = response.choices[0].message
tool_calls = response_message.tool_calls
# --- Process Response ---
if tool_calls:
print("\n--- OpenAI requested tool call ---")
for tool_call in tool_calls:
if tool_call.function.name == "send_whatsapp":
function_args = json.loads(tool_call.function.arguments)
tool_name = tool_call.function.name
print(f"Tool to call: {tool_name}")
print(f"Arguments: {json.dumps(function_args, indent=2)}")
print(f"\n--- Connecting to MCP Server ({mcp_server_script}) via stdio ---")
# Define server parameters for stdio connection
server_params = StdioServerParameters(
command=python_executable,
args=[mcp_server_script],
)
# Connect using stdio_client and manage session
async with stdio_client(server_params) as (read, write):
async with ClientSession(read, write) as session:
# Initialize the connection (optional but good practice)
# await session.initialize() # Might not be strictly needed if just calling a tool
print(f"--- Calling tool '{tool_name}' on server ---")
# Call the tool on the server with the arguments from OpenAI
# Note: The mcp library expects arguments nested under 'arguments' key
result = await session.call_tool(tool_name, arguments=function_args)
print("\n--- Response from MCP Server ---")
print(result) # The result object might have more structure, adjust printing if needed
else:
print("\n--- OpenAI Response (no tool call requested) ---")
print(response_message.content)
except Exception as e:
print(f"\n--- An error occurred ---")
print(e)
# --- Run the async function ---
if __name__ == "__main__":
asyncio.run(run())