LlamaIndex Supervisor is a developer-focused Python framework designed to create, run, and monitor AI agents built on LlamaIndex. It provides tools for defining workflows as nodes—such as retrieval, summarization, and custom processing—and wiring them into directed graphs. The Supervisor oversees each step, validating outputs against schemas, retrying on errors, and logging metrics. This ensures robust, repeatable pipelines for tasks like retrieval-augmented generation, document QA, and data extraction across diverse datasets.

Agent Logging provides a unified logging framework for AI agent frameworks and custom workflows. It intercepts and records each stage of an agent’s execution—prompt generation, tool invocation, LLM response, and final output—along with timestamps and metadata. Logs can be exported in JSON, CSV, or sent to monitoring services. The library supports customizable log levels, hooks for integration with observability platforms, and visualization tools to trace decision paths. With Agent Logging, teams gain insights into agent behavior, spot performance bottlenecks, and maintain transparent records for auditing.