TruLens

TruLens is an open-source evaluation and tracing framework designed to help developers move agentic workflows to production faster. It enables objective measurement of AI agent quality and effectiveness using feedback functions. These functions programmatically evaluate critical components of an app's execution flow, such as retrieved context, tool calls, and plans, expediting and scaling up experiment evaluation across various use cases including agents, retrieval-augmented generation (RAG), and summarization. The platform allows users to evaluate performance across multiple feedback functions like Groundedness, Context Relevance, and Coherence. It provides an extensible library of built-in feedback functions and supports custom ones, helping identify weaknesses in apps to inform iteration on prompts and hyperparameters. TruLens also facilitates testing by comparing different LLM applications on a metrics leaderboard to identify the best-performing versions, offering interoperable tracing through OpenTelemetry compatibility for integration with existing observability stacks. It is shepherded by Snowflake and is a community-driven project.