mirror of
https://github.com/ComposioHQ/composio.git
synced 2026-09-22 11:46:35 +08:00
Alberto Schiabel
41be258076
Spotted while fixing the provider-dependency failures in #4221. ## Problem `python/examples/tool_router/tool_router_mcp.py` guarded a missing credential like this: ```python api_key = os.environ.get("COMPOSIO_API_KEY") if not api_key: print("Error: COMPOSIO_API_KEY environment variable not set") print("Please set it using: export COMPOSIO_API_KEY='your_api_key'") return ``` `return` exits the coroutine normally, so `asyncio.run(main())` completes and the process exits **0**. The live-examples harness scores entries by exit code, so a run that did nothing at all was recorded **green** — the worst kind of failure for a canary, since it reports success while testing nothing. ## Fix Raise instead, using the `require_env` helper the sibling examples already use (`examples/tool_router/preload.py`): ```python def require_env(name: str) -> str: value = os.environ.get(name) if not value: raise RuntimeError(f"Set {name} before running this example.") return value ``` ## Verification Replaying the harness's own invocation with the variable unset: ``` $ unset COMPOSIO_API_KEY $ uv run --project python --with 'openai-agents>=0.19' \ --with ./python/providers/openai_agents --with ./python \ python python/examples/tool_router/tool_router_mcp.py before: exit=0 # silently "passes" after: exit=1 # RuntimeError: Set COMPOSIO_API_KEY before running this example. ``` `ruff check`, `ruff format --check`, and `nox -s chk_examples` all pass. ## Scope I checked whether other examples share the pattern. `examples/tool_router/authorize.py` prints the same message, but calls `exit(1)` at module level rather than returning from a coroutine, so it already fails correctly — left alone. This was the only instance.
79 lines
2.5 KiB
Python
79 lines
2.5 KiB
Python
"""
|
|
Tool Router - MCP (Model Context Protocol) Example with OpenAI Agents
|
|
|
|
This example demonstrates how to use Tool Router with MCP servers and OpenAI Agents provider.
|
|
The MCP server provides a standardized way to access tools across different platforms,
|
|
and OpenAI Agents provider allows you to use these tools with OpenAI's agent framework.
|
|
"""
|
|
|
|
import asyncio
|
|
import os
|
|
|
|
from agents import Agent, HostedMCPTool, Runner
|
|
from composio_openai_agents import OpenAIAgentsProvider
|
|
|
|
from composio import Composio
|
|
|
|
|
|
def require_env(name: str) -> str:
|
|
value = os.environ.get(name)
|
|
if not value:
|
|
raise RuntimeError(f"Set {name} before running this example.")
|
|
return value
|
|
|
|
|
|
async def main():
|
|
# Initialize Composio with OpenAI Agents provider
|
|
composio = Composio(
|
|
api_key=require_env("COMPOSIO_API_KEY"), provider=OpenAIAgentsProvider()
|
|
)
|
|
|
|
# Create a tool router session
|
|
session = composio.create(
|
|
# A user with connected GitHub and Gmail accounts (raises KeyError when unset)
|
|
user_id=os.environ["COMPOSIO_EXAMPLES_USER_ID"],
|
|
toolkits=["github", "gmail"],
|
|
# mcp=True surfaces session.mcp on the returned type
|
|
mcp=True,
|
|
)
|
|
|
|
mcpTool = HostedMCPTool(
|
|
tool_config={
|
|
"type": "mcp",
|
|
"server_label": "tool_router",
|
|
"server_url": session.mcp.url,
|
|
"require_approval": "never",
|
|
"headers": session.mcp.headers, # type: ignore[typeddict-item] # headers is typed Dict[str, Optional[str]]; the SDK always populates it with str values
|
|
}
|
|
)
|
|
|
|
print(f"Session created: {session.session_id}")
|
|
print(f"MCP Server URL: {session.mcp.url}")
|
|
print(f"MCP Server Type: {session.mcp.type}")
|
|
|
|
# Create an agent with the tools from tool router
|
|
agent = Agent(
|
|
name="MCP Tool Router Agent",
|
|
instructions=(
|
|
"You are a helpful assistant that can use GitHub and Gmail tools "
|
|
"through the MCP (Model Context Protocol) server. "
|
|
"Help users with their GitHub and email tasks."
|
|
),
|
|
tools=[mcpTool],
|
|
)
|
|
|
|
# Run the agent with a sample task
|
|
print("\n--- Running Agent with Tool Router Tools ---")
|
|
result = await Runner.run(
|
|
starting_agent=agent,
|
|
input=(
|
|
"List my recent GitHub repositories and tell me about them. "
|
|
"If successful, respond with a summary of what you found."
|
|
),
|
|
)
|
|
print(f"\nAgent Result: {result.final_output}")
|
|
|
|
|
|
if __name__ == "__main__":
|
|
asyncio.run(main())
|