During September 2025, Donal Evans developed a new feature for the elastic/elasticsearch repository, focusing on the Gemini 2.5 Inference API. He implemented thinking budget configuration, allowing users to fine-tune resource allocation for Gemini 2.5 models during chat completions. This work involved designing and parameterizing the inference API to expose budget controls, using Java and applying principles from API development and machine learning. By enabling more granular control over model behavior, Donal’s contribution addressed the need for flexibility and cost management in production deployments. The feature was delivered through a version-controlled commit, reflecting a focused and technically sound engineering effort.