I build and maintain practical, local-first tools and AI workflows—Windows-first when desktop integration matters. My public work spans desktop utilities, self-hosted assistants, browser tools, and verifiable agent workflows.
我打造並維護實用的本機優先工具與 AI 工作流程；需要桌面整合時以 Windows 優先。公開作品涵蓋桌面工具、自架助理、瀏覽器工具，以及可驗證的 agent 工作流程。
Batch-export books you legally own in Google Play Books: DRM-free titles download as EPUB/PDF, protected ones keep the official .acsm handoff for Adobe Digital Editions.
批次匯出自己合法擁有的 Google Play 圖書：無 DRM 的直接下載 EPUB/PDF，受保護的保留官方 .acsm 交接流程。
Python Playwright httpx Desktop GUI Local-first
Self-hosted personal AI assistant inside LINE — chat, voice, vision, search, schedules, tasks, reminders, weather — on your own OpenAI, LINE, and Supabase credentials. "AI finished" and "LINE delivered" are separate checkpoints, so a delivery retry never re-runs paid AI work. 可自架的 LINE 個人 AI 助理：聊天、語音、圖片、搜尋、行程、任務、提醒、天氣，全跑在自己的帳號上。「AI 已完成」與「LINE 已送達」是兩個 checkpoint，送達重試不會重跑已付費的 AI 工作。
Documentation site / 文件站 · Derived from memochou1993/gpt-ai-assistant, independently maintained.
JavaScript Node.js 24 Vercel Supabase LINE Messaging API Google Calendar / Tasks
Chrome MV3 side panel that embeds the ChatGPT session already signed in on the browser, builds Traditional Chinese prompts from the current page or selection (summary, translate, explain, outline), and inserts them without auto-submit. Chrome MV3 側邊欄擴充功能：嵌入瀏覽器裡已登入的 ChatGPT，依目前頁面或選取文字組成繁中提示詞（摘要／翻譯／解釋／大綱），只寫入輸入框、不自動送出。
Chrome Web Store / Chrome 線上應用程式商店
JavaScript Chrome MV3 Side Panel Local-first
Local-first voice typing for Windows: hotkey recording, on-device Faster-Whisper transcription, optional LLM rewriting and translation, typed straight into the focused app. Windows 本機語音輸入：快捷鍵錄音、本機 Faster-Whisper 辨識、可選 LLM 潤飾與翻譯，直接輸入目前作用中的程式。
Derived from VoiceType4TW, independently maintained.
Python PyQt6 Faster-Whisper CUDA Windows
Portable YouTube channel archiving with GUI and CLI: batch workflows, filters, ffmpeg integration, and content the signed-in user is already authorized to watch. 可攜式 YouTube 頻道保存工具，GUI 與 CLI 並具：批次工作流、篩選、ffmpeg 整合，也支援登入者原本就有權觀看的內容。
Python yt-dlp Tkinter ffmpeg Windows / macOS / Linux
Windows VRM desktop companion that turns an AI assistant's playback into lip sync, motion, states, and message bubbles; compatible agents drive it over a loopback-only MCP server. Windows VRM 桌面角色：把 AI 助理的聲音轉成口型、動作、狀態與訊息氣泡，相容 Agent 可經僅限本機的 MCP 控制。
Derived from xikhar/persona, independently maintained.
TypeScript Electron Three.js VRM / VRMA MCP
Developer-facing tools and engineering references for AI-assisted development, including original projects and clearly attributed maintenance forks. 給 AI 輔助開發使用的工具與工程參考，包含原創專案與清楚標示上游來源的維護型 fork。
Selected AI coding governance stack ｜精選 AI coding 治理堆疊 — four repos constraining agents at four layers, each usable on its own. 四個 repo 分別約束 agent 的四個層面，也可以單獨使用。
| Layer / 層 | Repo | |
|---|---|---|
| Dispatch / 派工決策 | agent-advisor | Risk-gated routing across four agent runtimes ｜四種 agent runtime 的風險分流路由 |
| Execution / 動作攔截 | harness-guard | Runtime hooks blocking dangerous commands, unevidenced completion claims, commits over failing tests ｜攔截危險指令、無證據的完成宣稱、紅燈仍提交 |
| Output / 產出品質 | ai-quality-gates | Gherkin specs, coverage and mutation gates, architecture contracts, bounded agent-loop policy ｜可執行規格、覆蓋率與 mutation gate、架構契約、有界 loop policy |
| Delivery / 交付流程 | paulsha-cortex | Candidate, verification, independent review, completion evidence ｜候選、驗證、獨立審查與完成證據 |
Other agent tooling ｜ 其他 agent 工具
| Repo | ||
|---|---|---|
| github-stars-organizer-playbook | Dry-run-first GitHub Stars Lists workflow with GraphQL batching and verification ｜先預覽、分批更新並重驗的 GitHub Stars Lists 整理流程 | Python original |
| MyR2D2 | Ten Claude Code skills against session amnesia: save-and-verify before shutdown, handoff and pickup between sessions, daily and weekly debriefs, self-check before reporting, and second review by a different model ｜十支對抗 session 失憶的 skills：收工前落地並驗證、跨 session 交接與接手、日結週結、回報前自檢，以及交給另一個模型的二審 | Python fork of tingyulu/MyR2D2 |
| opencodex | Universal provider proxy — run Codex or Claude Code on Claude, Gemini, Grok, DeepSeek, or local Ollama, native model picker intact ｜通用供應商代理，讓 Codex／Claude Code 改用任何 LLM，選擇器仍是原生的 | TypeScript fork of lidge-jun/opencodex |
| agentdeck | Windows tray cockpit: quota monitoring, multi-model roundtable, subagent roles, HTML reports; quota read from local files only ｜系統匣控制台：額度監看、多模型圓桌、subagent 角色、HTML 報告，額度只讀本機檔案 | Python fork of aqua5230/usage |
| book-to-skill | Turn a technical book or docs folder into on-demand Agent Skills — load the relevant chapter, not the whole book ｜把技術書或文件夾轉成按需載入的 Agent 技能，只載入用得到的那一章 | Python fork of virgiliojr94/book-to-skill |
Browse all public repositories →
- Local-first by default ｜預設本機優先 — if it runs on the user's machine, it does not get a hosted backend first.
- Release-oriented ｜以交付為導向 — downloadable, runnable, verifiable beats a demo.
- Privacy-aware ｜重視隱私邊界 — user files, images, and tokens do not go to services that do not need them.
- Explicit boundaries ｜寫明邊界 — licensing, platform rules, unsupported cases, and operational risks are stated, not implied.
- AI-assisted, still verified ｜AI 協作但仍要驗證 — agents speed up prototyping and docs; releases still need tests, packaging, and stated limits.
Recurring themes across the projects: local-first applications, AI in interfaces people already use, and verifiable workflows that turn repetitive work into bounded automation. 專案共通的主題：本機優先的應用程式、把 AI 放進使用者本來就在用的介面，以及把重複工作轉成有界且可驗證的自動化流程。
Private and organization-specific work stays private; this page highlights public personal projects and selected open-source maintenance work. 私人與組織專案維持不公開；本頁只展示公開的個人作品與精選的開源維護工作。
