I build agent applications and developer tools, with a current focus on MCP integrations and reliability around data systems.
关注 Agent 应用与开发工具、MCP 集成,以及数据系统的可靠性。当前的公开实践包括邮轮决策工具和 Doris HTTP 行为实验。
- Doris HTTP Reliability Lab: an independent, executable case study using pinned Doris MCP HTTP client code. It separates a reproduced timeout defect from expected retry behavior and an open timeout-scope question. Experiments use loopback networking; this is not an Apache project or a production Doris benchmark.
- Cruise AI Skills: reusable agent skills for cruise package value checks, cruise line comparisons, and shore excursion decisions.
Proposed fixes and regression tests cover session cleanup in mcp-use, URL elicitation errors in mcp-go, and finite HTTP timeouts in Apache Doris MCP Server.
Browse these upstream pull requests for current review and merge status.
相关实验、补丁和测试包含 AI 辅助工作;验证方法、结果与限制见各项目和 PR。个人实验与上游提案分别列出。


