Worked on the aws/language-servers repository to deliver a feature optimizing chat history handling and metrics accuracy. Focused on conditional retention of chat context and editor state, the solution ensured that context was preserved only when user input was not a tool result, reducing unnecessary storage and processing. Updated message counting logic to reflect the lengths of editor state and additional context, supporting more accurate analytics and lower storage costs. The work involved performance-focused refactoring in a TypeScript and Node.js codebase, leveraging CodeWhisperer and Language Server Protocol to align with product goals for efficient chat performance and cost optimization.