AutoGen (tracing)

AutoGen AgentChat is Microsoft’s framework for building event-driven single-agent and multi-agent applications. Respan gives you full observability over agent runs, team conversations, tool calls, and LLM generations.

Set up Respan

Create an account at platform.respan.ai and grab an API key .

Paste this into your coding agent:


Help me set up Respan tracing with the AutoGen integration.
Follow https://www.respan.ai/docs/integrations/autogen.md

Setup

Install packages

Terminal


pip install respan-ai respan-instrumentation-autogen

Set environment variables

Terminal


export OPENAI_API_KEY="YOUR_OPENAI_API_KEY"
export RESPAN_API_KEY="YOUR_RESPAN_API_KEY"

OPENAI_API_KEY is used for AutoGen model calls. RESPAN_API_KEY is used to export traces to Respan. Set OPENAI_MODEL to override the default gpt-4o-mini model.

Initialize and run

Python


import asyncio
import os

from autogen_agentchat.agents import AssistantAgent
from autogen_ext.models.openai import OpenAIChatCompletionClient
from respan import Respan
from respan_instrumentation_autogen import AutoGenInstrumentor

respan_api_key = os.environ["RESPAN_API_KEY"]
openai_api_key = os.environ["OPENAI_API_KEY"]
openai_model = os.getenv("OPENAI_MODEL", "gpt-4o-mini")

respan = Respan(
    api_key=respan_api_key,
    instrumentations=[AutoGenInstrumentor()],
)

model_client = OpenAIChatCompletionClient(
    model=openai_model,
    api_key=openai_api_key,
)

agent = AssistantAgent(
    name="assistant",
    model_client=model_client,
    system_message="You answer with concise, practical engineering advice.",
)

async def main():
    result = await agent.run(
        task="In one sentence, explain why tracing helps multi-agent apps."
    )
    print(result.messages[-1].content)
    await model_client.close()

asyncio.run(main())

View your trace

Open the Traces page to see your AutoGen workflow with agent spans, team spans, tool spans, and LLM generations.

Configuration

Parameter	Type	Default	Description
`api_key`	`str \| None`	`None`	Falls back to `RESPAN_API_KEY` env var.
`base_url`	`str \| None`	`None`	Falls back to `RESPAN_BASE_URL` env var.
`instrumentations`	`list`	`[]`	Plugin instrumentations to activate (e.g. `AutoGenInstrumentor()`).
`customer_identifier`	`str \| None`	`None`	Default customer identifier for all spans.
`metadata`	`dict \| None`	`None`	Default metadata attached to all spans.
`environment`	`str \| None`	`None`	Environment tag (e.g. `"production"`).

Attributes

In Respan()

Set defaults at initialization. These apply to all spans.

Python


from respan import Respan
from respan_instrumentation_autogen import AutoGenInstrumentor

respan = Respan(
    instrumentations=[AutoGenInstrumentor()],
    customer_identifier="user_123",
    metadata={"service": "autogen-api", "version": "1.0.0"},
)

With propagate_attributes

Override per-request using a context scope.

Python


from respan import Respan, propagate_attributes
from respan_instrumentation_autogen import AutoGenInstrumentor

respan = Respan(instrumentations=[AutoGenInstrumentor()])

async def handle_request(user_id: str, message: str):
    with propagate_attributes(
        customer_identifier=user_id,
        thread_identifier="conv_abc_123",
        metadata={"plan": "pro"},
    ):
        result = await agent.run(task=message)
        print(result.messages[-1].content)

Attribute	Type	Description
`customer_identifier`	`str`	Identifies the end user in Respan analytics.
`thread_identifier`	`str`	Groups related messages into a conversation.
`metadata`	`dict`	Custom key-value pairs. Merged with default metadata.

Examples

Tool calls

Tool calls are captured as spans with inputs, outputs, and timing.

Python


from autogen_agentchat.agents import AssistantAgent

async def estimate_latency(service: str, requests_per_minute: int) -> str:
    """Estimate API latency for a service under load."""
    return f"{service}: about 155 ms p95 latency"

agent = AssistantAgent(
    name="capacity_planner",
    model_client=model_client,
    tools=[estimate_latency],
    reflect_on_tool_use=True,
    system_message="Use the estimate_latency tool before answering.",
)

result = await agent.run(
    task="Estimate p95 latency for tracing-api at 240 requests per minute."
)
print(result.messages[-1].content)

Teams

Team conversations are traced with workflow, agent, and LLM spans.

Python


from autogen_agentchat.agents import AssistantAgent
from autogen_agentchat.conditions import MaxMessageTermination, TextMentionTermination
from autogen_agentchat.teams import RoundRobinGroupChat

planner = AssistantAgent(
    name="planner",
    model_client=model_client,
    system_message="Create a compact implementation plan.",
)
reviewer = AssistantAgent(
    name="reviewer",
    model_client=model_client,
    system_message="Review the plan. If clear, end with APPROVED.",
)

team = RoundRobinGroupChat(
    [planner, reviewer],
    termination_condition=TextMentionTermination("APPROVED") | MaxMessageTermination(4),
)
result = await team.run(task="Plan a small release checklist.")