Worked on the zed-industries/zed repository to deliver a feature enhancing OpenAI-compatible model configuration, enabling users to tune the reasoning effort for models such as glm-5 and kimi-k2.5. Implemented a new reasoning_effort parameter using Rust, focusing on back end development to support performance tuning for reasoning-intensive models. Reinforced the rollout with comprehensive testing and documentation, including a concrete JSON configuration example and usage snippet to guide adoption. Consolidated repository configurations to streamline future experimentation with model reasoning parameters, ensuring safer deployments and easier upgrades. The work emphasized maintainability and scalability within the Rust-based back end infrastructure.