Cloud Monitor:Integrate AgentScope (Python) Applications

Framework Introduction

AgentScope is an open-source, multi-agent application development framework from Alibaba Cloud. It provides various agent types, such as ReActAgent, and built-in model adapters for services such as DashScope and OpenAI. It also supports tool calling, memory management, and multi-agent collaboration.

After integration, the following capabilities are automatically monitored:

• Agent execution traces

• LLM calls, including token usage and the input/output content of model calls

• Tool calling traces, including the call details for each tool in the toolkit

• ReAct steps, including the observation of actions in each loop iteration

Integration methods

pip3 install aliyun-bootstrap

# Method 1: Add environment variables for all processes in the current shell.
export ARMS_APP_NAME=<your-app-name>        # Specifies a custom name for your application.
export ARMS_WORKSPACE=<your-workspace-name> # Specifies the workspace to which data is reported.
export ARMS_REGION_ID=<your-region-id>      # Specifies the Region ID of your workspace.
export ARMS_LICENSE_KEY=<your-license-key>  # Specifies the license key for authentication.

# Method 2: Add environment variables for a specific process.
ARMS_APP_NAME=<your-app-name> ARMS_WORKSPACE=<your-workspace-name> ARMS_REGION_ID=<your-region-id> ARMS_LICENSE_KEY=<your-license-key> aliyun-instrument <your-app-file>.py

# Add environment variables
ENV ARMS_APP_NAME={your-app-name}
ENV ARMS_REGION_ID={your-region-id}
ENV ARMS_LICENSE_KEY={your-license-key}
ENV ARMS_WORKSPACE={your-workspace-name}

## Add your original Dockerfile commands here

aliyun-instrument python app.py

Sample Code

import asyncio
import os
from agentscope.agent import ReActAgent
from agentscope.formatter import DashScopeChatFormatter
from agentscope.message import Msg, TextBlock
from agentscope.model import DashScopeChatModel
from agentscope.tool import Toolkit, ToolResponse

def get_weather(city: str) -> ToolResponse:
    weather_data = {
        "Beijing": "Sunny 25°C",
        "Shanghai": "Cloudy 22°C",
        "Hangzhou": "Light rain 20°C",
    }
    result = weather_data.get(city, f"{city}: No weather data available")
    return ToolResponse(content=[TextBlock(type="text", text=result)])

def search_info(keyword: str) -> ToolResponse:
    info = {
        "West Lake": "West Lake in Hangzhou is a famous freshwater lake, UNESCO World Heritage",
        "Great Wall": "Ancient Chinese defensive structure, UNESCO World Heritage",
    }
    result = info.get(keyword, f"No information found for '{keyword}'")
    return ToolResponse(content=[TextBlock(type="text", text=result)])

async def main():
    model = DashScopeChatModel(
        model_name=os.environ.get("MODEL_NAME", "qwen-plus"),
        api_key=os.environ.get("DASHSCOPE_API_KEY"),
    )
    toolkit = Toolkit()
    toolkit.register_tool_function(get_weather)
    toolkit.register_tool_function(search_info)
    agent = ReActAgent(
        name="TravelAgent",
        sys_prompt="You are a travel assistant. Use tools to query weather and travel info.",
        model=model,
        formatter=DashScopeChatFormatter(),
        toolkit=toolkit,
        max_iters=5,
    )
    msg = Msg(
        name="user",
        content="What's the weather in Hangzhou today? Is West Lake worth visiting?",
        role="user",
    )
    result = await agent(msg)
    print(result.content)

asyncio.run(main())

View Monitoring Details

1. Log on to the Cloud Monitor 2.0 console. Select the target workspace. In the navigation pane on the left, choose All Features > AI Application Observability.

2. On the AI Application List page, you can view the connected applications. Click an Application Name to view detailed monitoring data for the application.

On the AI Application Observability page, click an application name to open the details page. The details page shows the call chain of the application, with overview information such as the trace ID, start time, total duration, and total tokens. The timing diagram shows the complete agent call chain:

• invoke_agent (AGENT type, 4.59s total duration)

• react step (STEP type) -> chat qwen-plus (LLM type, with input and output tokens)

• execute_tool (TOOL type, such as get_weather and search_info)

You can filter spans by type: AGENT, STEP, LLM, and TOOL.

Click a span to view its details in the panel on the right. For an LLM span, Input Messages shows the system prompt and the user question, and Output Messages shows the model response together with the tool_call details.

More References

Troubleshoot common issues

Python Probe Usage FAQ