TruLens

TruLens evaluates and traces AI agents and LLM applications, to find where they fail and where costs can be reduced without losing quality. It records the execution of an application with OpenTelemetry-native tracing and scores each step with feedback functions.

Feedback functions measure aspects such as groundedness, context relevance and answer relevance (the “RAG triad”), using LLMs, classical models or custom code. Complete multi-turn conversations can be tracked and evaluated, and a dashboard compares application versions on quality, latency and cost.

TruLens is free and open source under the MIT licence.