core[patch]: add token counting callback handler

[ccurme]

Stripped-down version of OpenAICallbackHandler that just tracks AIMessage.usage_metadata.

from langchain_core.callbacks import get_usage_metadata_callback
from langgraph.prebuilt import create_react_agent

def get_weather(location: str) -> str:
    """Get the weather at a location."""
    return "It's sunny."

tools = [get_weather]
agent = create_react_agent("openai:gpt-4o-mini", tools)

with get_usage_metadata_callback() as cb:
    result = await agent.ainvoke({"messages": "What's the weather in Boston?"})
    print(cb.usage_metadata)

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
langchain	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Mar 25, 2025 10:14pm

[eyurtsev]

implementation looks good, needs more api reference documentation

[eyurtsev]

printing the callback isn't that useful I think.

instead

print(callback.usage_metadata) and consider showing example output

[ccurme]

the callback does implement __repr__, so it renders OK but I think it's better to do callback.usage_metadata as you suggest to show people how to access the underlying dict.