Aller au contenu
Menu
Français

TruLens

Instrumentation and feedback functions for evaluating model applications.

library Évaluation MIT 2 relationships

Aperçu

TruLens instruments a model application so every call is recorded with feedback scores attached, from relevance to groundedness. It solves the problem that an aggregate score cannot tell you which step of a pipeline produced a bad answer. It exists because the useful question is not whether the application is good but where it is going wrong, and answering that needs per-step evaluation rather than a single number.

Graphe des technologies

Comment TruLens se rattache au reste de la pile.

Appliqué à

Outils

Logiciels qui implémentent TruLens ou s'y intègrent.

LLM Frameworks