adk-bot
1.4 KiB
title
| title |
|---|
| Debug Logging |
The DebugLoggingPlugin captures detailed interaction data to a YAML file for debugging.
Overview
The DebugLoggingPlugin is a tool for developers to capture detailed data from an agent's interactions. This data includes LLM (Large Language Model) requests and responses, tool calls, events, and the agent's session state. The plugin saves this information in a human-readable YAML file, which can be used for debugging and analyzing the agent's behavior.
Usage
To use the DebugLoggingPlugin, you need to import it and add it to the list of plugins when you create a Runner.
from google.adk.plugins import DebugLoggingPlugin
debug_plugin = DebugLoggingPlugin(output_path="debug_logs.yaml")
runner = Runner(agent=my_agent, plugins=[debug_plugin])
Output Format
The output is a YAML file where each invocation is a separate document, separated by ---. This format makes it easy to read and share for debugging.
Configuration
You can configure the DebugLoggingPlugin with the following parameters:
| Parameter | Description |
|---|---|
output_path |
The path to the output file. The default is adk_debug.yaml. |
include_session_state |
A boolean that determines whether to include the session state in the output. |
include_system_instruction |
A boolean that determines whether to include system instructions in the output. |