LoggingCallbackHandler

A LoggingCallbackHandler class that extends StdOutCallbackHandler to provide additional logging capabilities. This handler logs specific events during the chain execution process and captures prompts for further retrieval.

Attributes

• always_verbose: A class attribute set to True to ensure verbose logging.
• logger: Logger instance initialized with the provided name.
• prompt: Stores the prompt formatted during the chain execution.

Methods

init(self, name: str) -> None

Initializes the LoggingCallbackHandler.

• Parameters:
  • name (str): The name of the logger.

on_chat_model_start(self, serialized: dict[str, Any], messages: list[list[BaseMessage]], **kwargs: Any) -> None

Handle the start of a chat model, capturing and logging the first message's content.

• Parameters:
  • serialized (dict[str, Any]): Serialized chat model details.
  • messages (list[list[BaseMessage]]): List of messages in the chat model.
  • kwargs (Any): Additional keyword arguments.

on_chain_start(self, *args: Any, **kwargs: Any) -> None

Placeholder for handling the start of a chain.

on_chain_end(self, *args: Any, **kwargs: Any) -> None

Placeholder for handling the end of a chain.

on_agent_action(self, *args: Any, **kwargs: Any) -> Any

Placeholder for handling agent actions.

on_tool_end(self, *args: Any, **kwargs: Any) -> None

Placeholder for handling the end of a tool.

on_tool_error(self, error: BaseException | KeyboardInterrupt, **kwargs: Any) -> None

Handle tool errors by doing nothing.

• Parameters:
  • error (BaseException | KeyboardInterrupt): The error encountered.
  • kwargs (Any): Additional keyword arguments.

on_text(self, text: str, color: str | None = None, end: str = "", **kwargs: Any) -> None

Placeholder for handling text output when an agent ends.

• Parameters:
  • text (str): The text to output.
  • color (str | None): Optional color for the text.
  • end (str): Optional string to append at the end.
  • kwargs (Any): Additional keyword arguments.

retrieve_prompt(self) -> str

Retrieve the captured prompt.

• Returns:
  • str: The captured prompt.

on_agent_finish(self, *args: Any, **kwargs: Any) -> None

Placeholder for handling when an agent finishes.

Usage Example

import logging
from your_module import BaseMessage, LoggingCallbackHandler
 
# Configuration of logging
logging.basicConfig(level=15)
 
# Initialize the handler
handler = LoggingCallbackHandler(name="langchain")
 
# Example usage with mock data
serialized = {}
messages = [[BaseMessage(content="Hello, how can I assist you today?")]]
 
handler.on_chat_model_start(serialized=serialized, messages=messages)
print(handler.retrieve_prompt())  # Output: "Hello, how can I assist you today?"