Implemented Asynccallbackhandler. Able to get LLM START, LLM END, TOOL START, TOOL end events. Not getting ON_AGENT_FINISH, ON _AGENT_ACTION ..
Referred this doc
https://python.langchain.com/api_reference/core/callbacks/langchain_core.callbacks.base.BaseCallbackHandler.html#langchain_core.callbacks.base.BaseCallbackHandler.on_agent_action
Passing config as runnable object still not sure why few events not getting triggered… any suggestions
could you please provide some relevant snippets of your code? That would make it easier to understand what might happen there.
Thanks in advance 
I am looking forward to seeing the snippets 
Here is the snippet. Agent finish and Agent Action is not getting triggered
from langgraph.prebuilt import create_react_agent
from langchain_anthropic import ChatAnthropic
import os
from langchain_core.callbacks import BaseCallbackHandler
from typing import Any, Dict, List,Optional
from uuid import UUID
from langchain_core.agents import AgentAction, AgentFinish
from langchain_anthropic import ChatAnthropic
from langchain_core.callbacks import BaseCallbackHandler
from langchain_core.messages import BaseMessage
from langchain_core.outputs import LLMResult
from langchain_core.prompts import ChatPromptTemplate
class LoggingHandler(BaseCallbackHandler):
def on_agent_finish(
self,
finish: AgentFinish,
*,
run_id: UUID,
parent_run_id: Optional[UUID] = None,
**kwargs: Any,
) -> Any:
print("agent finish")
def on_agent_action(
action: AgentAction,
*,
run_id: UUID,
parent_run_id: UUID | None = None,
**kwargs: Any,
)-> Any:
print("agent started")
def on_chat_model_start(
self, serialized: Dict[str, Any], messages: List[List[BaseMessage]], **kwargs
) -> None:
print("Chat model started")
def on_llm_end(self, response: LLMResult, **kwargs) -> None:
print(f"Chat model ended, response: {response}")
def on_chain_start( self, serialized: Dict[str, Any], inputs: Dict[str, Any], **kwargs
) -> None:
print(f"Chain {serialized.get('name')} started")
def on_chain_end(self, outputs: Dict[str, Any], **kwargs) -> None:
print(f"Chain ended, outputs: {outputs}")
anthropic_model = ChatAnthropic(model=“claude-sonnet-4-5”)
def get_weather(city: str) → str:
“”“Get weather for a given city.”“”
return f"It’s always sunny in {city}!"
agent = create_react_agent(
model=anthropic_model,
tools=[get_weather],
prompt=“You are a helpful assistant”
)
callbacks = [LoggingHandler()]
Run the agent
print(agent.invoke(
{“messages”: [{“role”: “user”, “content”: “what is the weather in sf”}]},config={“callbacks”: callbacks}
))
something is wrong with your recent message format. Could you fix it please? It’s hard to read, not to mention difficult to copy. Thanks in advance
from langgraph.prebuilt import create_react_agent
from langchain_anthropic import ChatAnthropic
import os
from langchain_core.callbacks import BaseCallbackHandler
from typing import Any, Dict, List,Optional
from uuid import UUID
from langchain_core.agents import AgentAction, AgentFinish
from langchain_anthropic import ChatAnthropic
from langchain_core.callbacks import BaseCallbackHandler
from langchain_core.messages import BaseMessage
from langchain_core.outputs import LLMResult
from langchain_core.prompts import ChatPromptTemplate
class LoggingHandler(BaseCallbackHandler):
def on_agent_finish(
self,
finish: AgentFinish,
*,
run_id: UUID,
parent_run_id: Optional[UUID] = None,
**kwargs: Any,
) -> Any:
print("agent finish")
def on_agent_action(
action: AgentAction,
*,
run_id: UUID,
parent_run_id: UUID | None = None,
**kwargs: Any,
)-> Any:
print("agent started")
def on_chat_model_start(
self, serialized: Dict[str, Any], messages: List[List[BaseMessage]], **kwargs
) -> None:
print("Chat model started")
def on_llm_end(self, response: LLMResult, **kwargs) -> None:
print(f"Chat model ended, response: {response}")
def on_chain_start(
self, serialized: Dict[str, Any], inputs: Dict[str, Any], **kwargs
) -> None:
print(f"Chain {serialized.get('name')} started")
def on_chain_end(self, outputs: Dict[str, Any], **kwargs) -> None:
print(f"Chain ended, outputs: {outputs}")
anthropic_model = ChatAnthropic(model="claude-sonnet-4-5")
def get_weather(city: str) -> str:
"""Get weather for a given city."""
return f"It's always sunny in {city}!"
agent = create_react_agent(
model=anthropic_model,
tools=[get_weather],
prompt="You are a helpful assistant"
)
callbacks = [LoggingHandler()]
# Run the agent
print(agent.invoke(
{"messages": [{"role": "user", "content": "what is the weather in sf"}]},config={"callbacks": callbacks}
))
thanks for the snippet
create_react_agent returns a LangGraph runnable compiled from a graph. It is not the LangChain AgentExecutor loop. The agent-specific callbacks on_agent_action and on_agent_finish are emitted by the AgentExecutor implementation, not by LangGraph’s runnable/graph. With LangGraph prebuilt agents you’ll typically see chain/LLM/tool callbacks, but not the agent-level ones.
Thanks @pawel-twardziak .. understood.. additional question. How to pass extra arguments to onchain or onllmend ? Don’t see any documentation
Have you trued sth like this?
from langchain_core.callbacks import BaseCallbackHandler
class MyHandler(BaseCallbackHandler):
def __init__(self):
self._by_run = {}
def on_chain_start(self, serialized, inputs, *, run_id, tags=None, metadata=None, **_):
self._by_run[run_id] = {"tags": tags or [], "metadata": metadata or {}}
def on_chain_end(self, outputs, *, run_id, **_):
ctx = self._by_run.pop(run_id, {})
# use ctx["metadata"], ctx["tags"] here
And then
chain.invoke(
user_input,
config={
"callbacks": [MyHandler()],
"tags": ["job:42"],
"metadata": {"jobId": 42, "user": "alice"},
},
)
See there which events metadata are being passed to langchain/libs/core/langchain_core/callbacks/manager.py at 5f9e3e33cd51c3fffd0111302476412f49a06e01 · langchain-ai/langchain · GitHub
@pawel-twardziak this we tried . Onllmend kwargs is not working.. it is coming empty
what kwargs do you expect from on_llm_end?
Metadata are being passed only to on_..._start events, not ..._end.
Say I have internal project name which I need to log on every on_llm_end… I can get this from kwargs ..but how i can pass the kwargs so I can get in llmend
You can’t pass arbitrary extra args directly into on_llm_start/on_llm_end from invoke. Handlers accept kwargs, but built‑ins only forward specific fields.
What you CAN pass:
- on_llm_start: tags and metadata are forwarded. Use config to supply them, then stash by run_id
- on_llm_end: tags are forwarded; metadata is not. Read what you stashed at start using run_id
on_llm_start and on_llm_end work together
class MyHandler(BaseCallbackHandler):
def __init__(self):
self._ctx = {}
def on_llm_start(self, serialized, prompts, *, run_id, metadata=None, **_):
self._ctx[run_id] = metadata or {}
def on_llm_end(self, response, *, run_id, **_):
meta = self._ctx.pop(run_id, {})
projectName = meta.get("projectName")
if projectName is None:
return # or handle/log as needed
llm.invoke(
"prompt",
config={"callbacks": [MyHandler()], "tags": ["job:42"], "metadata": {"projectName": "MyProject"}},
)