Skip to content
OpenObserve
Get Demo Star
OpenObserve
Get Demo Star
Open Menu Close Menu

AutoGen → OpenObserve

Automatically capture agent conversations, LLM calls, and tool executions for every AutoGen multi-agent workflow in your Python application.

Prerequisites

  • Python 3.9+
  • An OpenObserve account (cloud or self-hosted)
  • Your OpenObserve organisation ID and Base64-encoded auth token
  • An OpenAI API key (or whichever LLM backend your agents use)

Installation

pip install openobserve-telemetry-sdk openinference-instrumentation-autogen "ag2==0.9.0" python-dotenv

The ag2 package is the maintained fork of pyautogen and imports as autogen.

Configuration

Create a .env file in your project root:

# OpenObserve instance URL
# Default for self-hosted: http://localhost:5080
OPENOBSERVE_URL=https://api.openobserve.ai/

# Your OpenObserve organisation slug or ID
OPENOBSERVE_ORG=your_org_id

# Basic auth token — Base64-encoded "email:password"
OPENOBSERVE_AUTH_TOKEN=Basic <your_base64_token>

# LLM provider key
OPENAI_API_KEY=your-openai-key

Instrumentation

Call AutogenInstrumentor().instrument() before importing AutoGen.

from dotenv import load_dotenv
load_dotenv()

from openinference.instrumentation.autogen import AutogenInstrumentor
from openobserve import openobserve_init

AutogenInstrumentor().instrument()
openobserve_init()

import os
import autogen

config_list = [{"model": "gpt-4o-mini", "api_key": os.environ["OPENAI_API_KEY"]}]

assistant = autogen.AssistantAgent(
    name="assistant",
    llm_config={"config_list": config_list},
    system_message="You are a helpful assistant. Keep answers brief.",
)

user = autogen.UserProxyAgent(
    name="user",
    human_input_mode="NEVER",
    max_consecutive_auto_reply=1,
    is_termination_msg=lambda x: True,
)

user.initiate_chat(assistant, message="Explain OpenTelemetry in one sentence.")

Two-agent conversation

researcher = autogen.AssistantAgent(
    name="researcher",
    llm_config={"config_list": config_list},
    system_message="You research questions and provide factual answers.",
)

critic = autogen.AssistantAgent(
    name="critic",
    llm_config={"config_list": config_list},
    system_message="You review answers and confirm if they are correct. Reply TERMINATE when done.",
)

user = autogen.UserProxyAgent(
    name="user",
    human_input_mode="NEVER",
    max_consecutive_auto_reply=0,
)

user.initiate_chat(
    researcher,
    message="What is the capital of France?",
    max_turns=2,
)

What Gets Captured

Each initiate_chat() call produces a root span with child AGENT spans per agent and LLM spans for each model call.

Agent span

Attribute Description
openinference_span_kind AGENT
operation_name Agent class name (e.g. AssistantAgent)
agent_type Agent class name (e.g. AssistantAgent)
llm_observation_type AGENT
input_mime_type application/json
output_mime_type application/json
duration Agent turn latency
span_status UNSET on success, ERROR on failure

Token counts and model name appear on the child LLM spans beneath each agent span.

Viewing Traces

  1. Log in to OpenObserve and navigate to Traces in the left sidebar
  2. Click any root conversation span to open the waterfall view
  3. Expand the tree to see each AssistantAgent span and its child LLM calls
  4. Filter by operation_name = AssistantAgent to find all agent spans

AutoGen trace in OpenObserve

Next Steps

With AutoGen instrumented, every multi-agent conversation is recorded in OpenObserve with the full sequence of LLM calls. From here you can track how many turns a conversation takes, measure token usage per agent, and identify which agents drive the most cost.

Read More