DeepSeek Harness Remote
DeepSeek Harness Remote 可將手機或另一台電腦與你運行 DeepSeek Harness 的電腦安全連接,隨時隨地繼續你的工作,方便隨時查看進度並繼續工作。
AI 翻譯
尚未偵測到更新。GitHub 發布版本後會顯示在這裡。
個人 GitHub 帳號下公開儲存庫的擁有者可以登入並驗證所有權,自行管理介紹,無需管理員核准。
登入並產生驗證檔案,在 30 分鐘內將其提交至此儲存庫的預設分支。其他分支或分叉儲存庫中的檔案無效。然後在下方檢查檔案,即可無需管理員核准直接管理專案。請保留檔案以維持管理權限。
如需登入,將前往登入頁面。登入後請回到此處,再次點擊所選的驗證按鈕。
在專案資料夾中執行 /nomorevibe verify,即可驗證所有權並自行管理介紹。
一分鐘完成驗證Local, open-source "workday diary"; a menu-bar/tray app that logs your own screen activity to plain JSONL files so your AI assistant has context on your day. macOS, Windows, and Linux. Fully local, secrets redacted before writing.
翻譯準備中
取自儲存庫 README
正在載入完整 README。
檢視完整 README ↗A local, open-source "workday diary" for your own computer. Squidpal is a small menu-bar / tray app that records what you see and do on your own machine — the apps you use, the windows and pages you open, the text on screen and what you type — and writes it to plain text files in a folder you choose. The point is to give your own AI assistant durable context about your day: follow-ups, what you were working on, how things connect.
It runs on macOS, Windows and Linux, entirely locally. It makes no network requests. Secrets are stripped out before anything is written to disk.
Status: early prototype, in alpha testing. Expect rough edges. (User-visible names, the macOS bundle id
io.squidler.squidpal, and the app/ binary names are all Squidpal.dayloglingers only as an internal code symbol — the Swift module, the C# namespace, the Python package and the repo's source folders — which can be renamed in a later pass without any user-facing effect.)
Squidpal is a screen-activity recorder. Please understand what that means:
See docs/PRIVACY.md for the full picture.
Each platform has its own reader for the OS accessibility layer — the same technology screen readers use — plus a shared design for redaction and the output format:
flowchart LR
subgraph R1["Per-platform reader — reads the active window: text, focus, URL"]
direction TB
M["macOS · Accessibility API"]
W["Windows · UI Automation"]
L["Linux · AT-SPI"]
end
M --> RD
W --> RD
L --> RD
RD["Redactor<br/>strip secrets<br/>identical rules on every platform"] --> J["JSONL day files<br/>one line per event · append-only"]
The redaction rules are identical across all three platforms and are tested
against one shared set of fixtures (shared/redaction-fixtures.json),
so masking cannot silently drift between macOS, Windows and Linux.
| Platform | Reads via | UI | Language | Notes |
|---|---|---|---|---|
| macOS 14+ | Accessibility API (AXUIElement) | Menu-bar (SwiftUI) | Swift | Needs Accessibility permission. |
| Windows 10/11 | UI Automation (FlaUI) | System tray (WinForms) | C# / .NET 8 | No permission prompt; Chrome/Edge read best. |
| Linux | AT-SPI (pyatspi) | AppIndicator tray | Python / GTK 3 | Works under X11 and Wayland. GNOME needs the AppIndicator extension. |
Each has its own build and run instructions:
The fastest way to see something real is the redaction self-test — it needs no GUI and no accessibility permissions:
# Linux / macOS (Python 3):
python3 linux/selftest.py
# Windows / anywhere with .NET 8:
dotnet run --project windows/src/DayLog.SelfTest -c Release -- shared/redaction-fixtures.json
# macOS, with the Swift toolchain:
cd macos && ./build.sh test
All three run the same cases from shared/redaction-fixtures.json and should
report the same result.
Then build the app for your platform from its README above.
squidpal/
├── macos/ Swift package — menu-bar app + redactor core + tests
├── windows/ .NET 8 solution — tray app + core + self-test
├── linux/ Python package — tray app + core + self-test
├── shared/ redaction-fixtures.json — the one source of truth for masking
├── docs/ file format, redaction, privacy
└── .github/ CI (runs all three test suites) and templates
Contributions are welcome — see CONTRIBUTING.md. The golden rule: any change to redaction must keep all platforms passing the shared fixtures, and new masking behavior needs a new fixture.
To report a vulnerability or a case where Squidpal writes a secret it should have masked, see SECURITY.md. Please don't open a public issue with a real secret in it.
MIT © 2026 Squidler AB.