February 2026 (2026-02) monthly summary for kvcache-ai/sglang. Key feature delivered: CI test for GPT-OSS model evaluation using piecewise CUDA graph execution to verify accuracy thresholds during evaluation. This enhancement increases automated test coverage, reproducibility, and early detection of performance drift for GPT-OSS evaluations, contributing to safer releases and faster feedback loops. No major bugs fixed this month in this repository. Overall impact: improved quality gates for model evaluation and reduced risk of regression in production.

December 2024 focused on hardening the Delta-to-Iceberg conversion path and reducing failure modes in partitioned data migrations. Delivered a targeted bug fix to use a default partition value when entries are missing, improving robustness without impacting user-facing behavior. All tests pass, and the change aligns with ongoing reliability goals for data pipelines.