TreeMap
A free, open-source disk-space visualizer. Squarified treemap, duplicate hunting, disk-full forecasting, and cleanup that only ever moves files to the Trash.
翻译准备中
个人 GitHub 账号下公开仓库的所有者可以登录并验证所有权,自行管理介绍,无需管理员批准。
登录并生成验证文件,在 30 分钟内将其提交到此仓库的默认分支。其他分支或分叉仓库中的文件无效。然后在下方检查文件,即可无需管理员批准直接管理项目。请保留文件以维持管理权限。
如需登录,将前往登录页面。登录后请回到此处,再次点击所选的验证按钮。
在项目文件夹中运行 /nomorevibe verify,即可验证所有权并自行管理介绍。
一分钟完成验证Open-source AI companion for your projects, on Mac and iPhone. Talk or type: it remembers, thinks with you, and saves notes as plain files. Works with Claude, OpenAI, Gemini or local models.
翻译准备中
取自仓库 README
正在加载完整 README。
查看完整 README ↗
Broski is a companion you talk to about your projects. Press ⌥⇧B on your Mac and talk or type: Broski remembers what you've told it, thinks with you, and offers to save the decisions, todos and bugs you mention, saving only when you tap Save. Everything is plain files you own, any model can run it (Claude, OpenAI, Gemini and other providers, a model on your Mac, or Apple's on-device model), and an iPhone app shares the same conversation. Claude Code can read your notes through Broski's MCP server.
You need a Mac with Apple silicon on macOS 14 Sonoma or later.
A short welcome sets up the shortcut, your voice, your AI and your project folders. Getting started covers the first run, permissions and voice models.
Build from source with Xcode 26 or later on macOS 15 or later. Clone the repository, then:
scripts/build-app.sh --open
Every feature asks for a role (fast, everyday or deep), and Settings › Models maps each role to a provider. Mix them as you like.
| Provider | What you need | Runs | Reads images |
|---|---|---|---|
| Claude (Anthropic) | An API key | Cloud | Yes, and PDFs |
| OpenAI | An API key | Cloud | Yes |
| Google Gemini | An API key (there's a free tier) | Cloud | Yes |
| OpenRouter | One API key for hundreds of models | Cloud | Yes |
| Groq | An API key | Cloud | No |
| Mistral | An API key | Cloud (Europe) | Yes |
| DeepSeek | An API key | Cloud | No |
| Ollama | Ollama on your Mac, no key | Your Mac | No, unless you mark the model |
| LM Studio | LM Studio's server on your Mac, no key | Your Mac | No, unless you mark the model |
| Apple on-device | macOS 26 or iOS 26 with Apple Intelligence | Your device | No |
| Another server | Any OpenAI-compatible address | Wherever it runs | If you mark it |
Broski tests the provider, reads its own list of models and picks one per role, which you can change. Keys stay on the device. Without any model, quick notes are still filed by Broski's built-in phrase rules. Choosing your AI covers keys, costs, local models and spending.
Your notes, conversations and memory are files in your Broski folder: no account, no Broski server, no analytics. Text reaches a model only through one gateway that masks keys and tokens first. A local-only project's text goes only to models on your device, and with none set up Broski refuses rather than sending anything. Broski never sends audio anywhere itself. Read Privacy for exactly what goes where, private mode, and how the vault is encrypted.
The same conversation, quick notes and your notes in your pocket, synced through iCloud Drive. It isn't in the App Store: you build it with your own Apple ID, and a free one works. See The iPhone app.
Broski --mcp is an MCP server for Claude Code, Claude Desktop and other clients: projects, pages, search, decisions, trackers and what changed. It reads in place, hides local-only projects, and writes only to trackers Broski keeps. Settings › Connections connects Claude Code in one click. See Claude Code and MCP.
Contributions are welcome. Broski is built test-first: write the failing test in Tests/BroskiCoreTests, then the code, and keep the logic in BroskiCore. Install the git hooks once with scripts/install-hooks.sh, and run scripts/ci.sh before you push. Coding agents start with AGENTS.md. See CONTRIBUTING.md, and the code of conduct everyone taking part follows.
Found a way to make Broski leak, act without asking, or read what it shouldn't? Please report it privately: see SECURITY.md. The checklist for outside reviews is in docs/SECURITY-REVIEW.md.
Apache License 2.0: see LICENSE. Third-party credits are in NOTICE.
Speech recognition: Underdog Bark 3.0 by Conway Research (CC BY 4.0), built on NVIDIA's parakeet-tdt-0.6b-v3 (CC BY 4.0). Whisper via whisper.cpp (MIT). The design owes ideas to OpenClaw, VoiceInk and FluidVoice; Architecture says which.