Zhangheng Zeng focused on improving prompt quality and maintainability for the eosphoros-ai/DB-GPT repository during May 2025. He addressed a bug in the Chinese prompt templates, correcting a typographical error that previously caused confusion in data rendering and display types. By refining the natural language used in system prompts, Zhangheng enhanced localization accuracy and reduced ambiguity for Chinese users. His work emphasized traceability, with well-documented commit messages detailing the prompt-related fix. Leveraging skills in Natural Language Processing and prompt engineering, and working primarily in Python, he contributed to the product’s reliability through targeted, quality-driven improvements rather than new feature development.