Latitude

Latitude is an open-source platform for building and operating LLM features in production. It provides observability, evaluations, prompt management, and an eval-driven reliability loop to continuously improve AI products. Teams adopt Latitude to instrument existing LLM calls for observability and evaluation coverage, then use the reliability loop to turn production failures into repeatable fixes. Key capabilities include a prompt playground, datasets, LLM-as-judge evaluations, experiments, and a prompt optimizer (GEPA) that searches prompt variations against eval suites to reduce recurring failures.