A Comprehensive Guide to AGENT-DOSSIER: The Structured Logging Library for AI Agents

AGENT-DOSSIER is a structured logging library designed for AI agents, offering session management and object unpacking capabilities. With Dossier, logs are well-organized by agent session, making it easy to search for logs related to a specific session. The library enables users to log a variety of objects, including dataclasses, Pydantic models, and dicts, with automatic unpacking for structured logging. Dossier also provides features like session registry for reusing sessions, custom processors for advanced log processing, and integrations with popular libraries like LangChain and Pydantic. Users can enhance their logging by adding custom fields or transforming data with Dossier’s processors. Moreover, they can implement stateful class processors for tracking cumulative data across log calls. A real-world example of a cost tracking processor showcases Dossier’s ability to track API costs. Dossier offers seamless integration with various object types, making it a versatile logging solution for AI agents and other applications.