Worked on performance optimizations for the GrepChunksTool in the Tencent/WeKnora repository, focusing on backend development using Go. Introduced a caching layer for token sets to eliminate redundant maximum marginal relevance computations, which reduced CPU usage and improved memory efficiency during token selection. Additionally, implemented swap deletion to optimize slice deletions, further enhancing processing speed on larger datasets. The work emphasized performance engineering and maintainability, with clear, incremental commits and concise documentation. No major bugs were reported during this period, as the primary focus remained on scalability, code quality, and optimizing the token selection workflow for better responsiveness and throughput.