SDK Overview
Reference documentation for the Judgeval SDK
The Judgeval SDK provides a complete toolkit for evaluating, monitoring, and optimizing AI agents. Trace agent execution in production, run evaluations with custom scoring rubrics, manage test datasets, and version prompts with integrated evaluation workflows.
Core SDK Components
JudgmentClient
The main client for running evaluations and managing projects
Tracer
Capture and monitor agent execution traces for debugging and analysis
Dataset
Manage collections of examples and traces for batch evaluation
PromptScorer
Create custom LLM-as-a-judge scorers with natural language rubrics
Prompt
Version and manage prompts with integrated evaluation