openinference

OpenInference AG2 Examples

Runnable examples that send AG2 traces to a local Phoenix instance at http://localhost:6006. Each example sets its own Phoenix project through the openinference.project.name resource attribute, so traces stay separated as you work through them.

Setup

pip install arize-phoenix
phoenix serve            # in a separate terminal
pip install -r requirements.txt
export OPENAI_API_KEY=<your-key>

Examples

Example	Phoenix project	Requires a key	What it traces
no_llm_multi_agent.py	ag2-no-llm-multi-agent	No	A two-agent chat and a tool call, using canned replies
sessions_and_metadata.py	ag2-sessions-and-metadata	No	Two chats grouped into one Phoenix session, with user, metadata, and tags
conversable_agent_run.py	ag2-conversable-agent	Yes	The quickstart agent driven by run() and response.process()
tool_calling.py	ag2-tool-calling	Yes	An LLM-driven tool call, split across a calling and an executing agent
group_chat.py	ag2-group-chat	Yes	An AutoPattern group chat where a manager routes between specialists
sequential_chats.py	ag2-sequential-chats	Yes	A queue of chats passing carryover forward via initiate_chats
structured_output.py	ag2-structured-output	Yes	An agent replying with schema-validated JSON via response_format
async_tool_calling.py	ag2-async-tool-calling	Yes	The async paths: a_initiate_chat, a_generate_reply, a_execute_function

If you have no API key handy, start with no_llm_multi_agent.py — it runs offline and is the quickest way to confirm traces are reaching Phoenix.

python no_llm_multi_agent.py

group_chat.py produces the most detailed trace, including the manager’s speaker selection:

_User.initiate_chat [AGENT]
  chat_manager.generate_reply [AGENT]
    finance_bot.generate_reply [AGENT]
      ChatCompletion [LLM]
    checking_agent.initiate_chat [AGENT]
      speaker_selection_agent.generate_reply [AGENT]
        ChatCompletion [LLM]
    summary_bot.generate_reply [AGENT]
      ChatCompletion [LLM]

Span kinds

AG2 method	Span name	OpenInference span kind
initiate_chat / a_initiate_chat (also used by run and initiate_chats)	<agent>.initiate_chat	AGENT
generate_reply / a_generate_reply	<agent>.generate_reply	AGENT
execute_function / a_execute_function	<tool>	TOOL

The LLM spans come from instrumenting the OpenAI client alongside AG2, which the LLM-backed examples do with OpenAIInstrumentor.

To send traces to Phoenix Cloud instead, point the exporter at your collector endpoint and add your API key as described in the Phoenix docs.

This site is open source. Improve this page.