I build tools for AI agents with TypeScript, Python, and Windows automation. My work includes agent integrations, session reliability, image-generation tooling, and reproducible fixes in existing codebases.
个人展示 · Agent 与 AI
Technical notes, personal projects, and an interactive timeline.
Live-site preview captured on October 11, 2026. Click the image to explore.
Four pull requests authored by me have been merged into OpenClaw:
| Contribution | Merged PR |
|---|---|
| Migrate legacy bundled-plugin load paths | #55054 |
| Handle malformed Windows execution paths without a drive letter | #58040 |
| Respect the selected memory slot in dreaming configuration | #62275 |
| Preserve announcement thread IDs during session-list fallback | #63506 |
I also received co-author credit on the merged Ollama thinking-control fix, #53200, which built on my earlier work.
My ongoing MCP TypeScript SDK work concerns exact Accept media-type validation:
main implementation #2481
and v1 backport draft #2738.
These contributions are awaiting maintainer review.
A Node.js/Python CLI and library for image-generation workflows, configurable OpenAI-compatible providers, and installable agent skills. Published on npm, with automated tests and documented upstream attribution.
A Windows utility that snapshots open Claude Code sessions and reopens their conversations in Windows Terminal tabs.
A FastAPI, LangGraph, and Next.js workspace for restaurant-site comparisons. It distinguishes observed data from simulated scenarios and keeps business assumptions visible in its reports.
- Reproduce a reported behavior before changing it.
- Keep implementation scope and regression coverage explicit.
- Use AI-assisted development while checking the resulting code and behavior.
- Document what was tested and what still needs real-environment verification.
For a project question or bug report, please open an issue in the relevant repository. For a tour of my work, visit Agent 与 AI.