Loop v2026.02.12 - Beta Release#
Released: February 12, 2026
This beta introduces AI-assisted evaluator generation, scoring improvements, and expanded SQL querying capabilities alongside timeline visualization enhancements and better trace metadata.
Features & Enhancements#
- Align Judge — Create LLM evaluators through a guided AI-assisted workflow. Align Judge analyzes your dataset and generates tailored evaluator prompts, making it faster to set up evaluation criteria. AI-dependent are disabled when Lens Prism is not configured.
- SQL Skill: Evaluation & Dataset Tables — Query evaluation and dataset tables directly using the SQL skill, enabling custom analysis and reporting across your evaluation data.
- Timeline Visualization — Improved timeline bar rendering with larger font size and better title positioning for easier trace exploration.
- Source Environment in Traces — New source environment ID column in the traces view, making it easier to distinguish spans from different deployment environments.
Fixes & Improvements#
- Scoring — Better numeric score validation, accurate sorted score indexing, and improved edit/delete guards for manual scoring workflows.
- Server Stability — Improved handling of malformed spans and OTLP int64 values to prevent ingestion errors.
- OpenAI Compatibility — Added support for max_completion_tokens parameter used by newer OpenAI models.
- UI Fixes — Fixed drawer closing behavior, span deletion refresh, auth URL browser link handling, and time range persistence across views.