Datasets:
Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -10,11 +10,15 @@ tags:
|
|
| 10 |
- rust
|
| 11 |
- agent
|
| 12 |
- privacy
|
|
|
|
| 13 |
license: mit
|
|
|
|
| 14 |
---
|
| 15 |
|
| 16 |
# 🪵 栖木 Perch — 让 AI 栖息在你的 Mac 上
|
| 17 |
|
|
|
|
|
|
|
| 18 |
**一个真正住在你电脑里、能替你干活的 AI 工作台。**
|
| 19 |
不联网也能用,不用把隐私交出去,算力由你掌控。
|
| 20 |
|
|
@@ -74,13 +78,33 @@ license: mit
|
|
| 74 |
2. **连接池复用**——全局单例 HTTP client,省每条 TCP+TLS 握手
|
| 75 |
3. **锁模型**——求快不偷换模型(速度来自减少浪费,不是降能力)
|
| 76 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 77 |
## 📦 安装
|
| 78 |
|
| 79 |
-
1. 下载
|
| 80 |
2. 拖进「应用程序」
|
| 81 |
3. 打开,配置你的模型 API(设置 → 云端 API)
|
| 82 |
|
| 83 |
> 要求:macOS(Apple Silicon)。所有 AI 服务 API **由你自己配置**,不捆绑、不送额度。
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 84 |
|
| 85 |
## 🛠️ 技术栈
|
| 86 |
|
|
@@ -91,21 +115,32 @@ Tauri 2 + React 19 + rusqlite · **1118+ 测试全绿** · MIT
|
|
| 91 |
|
| 92 |
多端同步 · 记忆回溯 · 插件市场 · 多智能体协作 · 深度分析 · 更强隐私
|
| 93 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 94 |
---
|
| 95 |
|
| 96 |
# 🪵 Perch (English)
|
| 97 |
|
|
|
|
|
|
|
| 98 |
**A local-first AI workbench that lives on your Mac — and actually gets work done.**
|
| 99 |
|
| 100 |
- ⚡ **0.08s first token** — local 0.6B model answers instantly, cloud model seamlessly takes over
|
| 101 |
- 🔒 **Privacy by design** — conversations, knowledge base and memories stay on your machine; optional E2E-encrypted sync
|
| 102 |
- 🤖 **Real agent** — reads/writes files, runs commands, searches the web, batch-processes, schedules tasks
|
| 103 |
- 🧠 **Bring your own models** — local Ollama + any cloud model (DeepSeek / MiMo / Qwen / OpenAI-compatible)
|
|
|
|
| 104 |
|
| 105 |
**Rust 74.7% · TypeScript 18.1% · Tauri 2 + React 19 · 1,118+ tests · MIT**
|
| 106 |
|
| 107 |
Download `Perch_4.5.0_aarch64.dmg` below (Apple Silicon). All AI APIs are self-configured.
|
| 108 |
|
|
|
|
|
|
|
| 109 |
---
|
| 110 |
|
| 111 |
**栖木 Perch** — 让 AI 栖息于你的本机,安静而强大。
|
|
|
|
| 10 |
- rust
|
| 11 |
- agent
|
| 12 |
- privacy
|
| 13 |
+
- llm
|
| 14 |
license: mit
|
| 15 |
+
pretty_name: 栖木 Perch — Local-first AI Workbench for macOS
|
| 16 |
---
|
| 17 |
|
| 18 |
# 🪵 栖木 Perch — 让 AI 栖息在你的 Mac 上
|
| 19 |
|
| 20 |
+

|
| 21 |
+
|
| 22 |
**一个真正住在你电脑里、能替你干活的 AI 工作台。**
|
| 23 |
不联网也能用,不用把隐私交出去,算力由你掌控。
|
| 24 |
|
|
|
|
| 78 |
2. **连接池复用**——全局单例 HTTP client,省每条 TCP+TLS 握手
|
| 79 |
3. **锁模型**——求快不偷换模型(速度来自减少浪费,不是降能力)
|
| 80 |
|
| 81 |
+
## 📦 版本迭代(看得见的历史)
|
| 82 |
+
|
| 83 |
+
| 版本 | 日期 | 亮点 | 下载 |
|
| 84 |
+
|---|---|---|---|
|
| 85 |
+
| **v4.5.0** | 2026-10-29 | 多对话并发 · 流式节流 · 实时速度条 | [`Perch_4.5.0_aarch64.dmg`](./Perch_4.5.0_aarch64.dmg) |
|
| 86 |
+
| v4.4.1 | 2026-10-28 | 云流修复(90s 空闲超时 + 压缩解码 + 重试) | [`Perch_4.4.1_aarch64.dmg`](./Perch_4.4.1_aarch64.dmg) |
|
| 87 |
+
| v4.1.0 Nirvana | 2026-10-28 | 加密云同步 · 语义起名 · 讨论区 · 待机屏保 | [`Perch_4.1.0-Nirvana_b5_aarch64.dmg`](./Perch_4.1.0-Nirvana_b5_aarch64.dmg) |
|
| 88 |
+
| v3.0.0 LightYear | 2026-10-24 | 搜索增强 · 命名系统 · 留言板 | [`Perch_3.0.0-LightYear-b5.app.zip`](./Perch_3.0.0-LightYear-b5.app.zip) |
|
| 89 |
+
|
| 90 |
+
> **建议直接下 v4.5.0**。旧版本保留是为了让你看见这个项目是一步步迭代出来的。
|
| 91 |
+
|
| 92 |
## 📦 安装
|
| 93 |
|
| 94 |
+
1. 下载 `Perch_4.5.0_aarch64.dmg`
|
| 95 |
2. 拖进「应用程序」
|
| 96 |
3. 打开,配置你的模型 API(设置 → 云端 API)
|
| 97 |
|
| 98 |
> 要求:macOS(Apple Silicon)。所有 AI 服务 API **由你自己配置**,不捆绑、不送额度。
|
| 99 |
+
> 首次打开若提示"未识别开发者",右键 → 打开。
|
| 100 |
+
|
| 101 |
+
## 📸 更多截图
|
| 102 |
+
|
| 103 |
+

|
| 104 |
+
*小栖专属助理:只读工具,盯着会话出主意*
|
| 105 |
+
|
| 106 |
+

|
| 107 |
+
*思考强度滑块:五档连续拖拽(关/低/中/高/炽焰)*
|
| 108 |
|
| 109 |
## 🛠️ 技术栈
|
| 110 |
|
|
|
|
| 115 |
|
| 116 |
多端同步 · 记忆回溯 · 插件市场 · 多智能体协作 · 深度分析 · 更强隐私
|
| 117 |
|
| 118 |
+
## ⭐ 如果你觉得有意思
|
| 119 |
+
|
| 120 |
+
- **在 GitHub 给个 Star**:https://github.com/ghongxv-web/perch
|
| 121 |
+
- **在 HF 点个 Like**(右上角 ❤️)
|
| 122 |
+
- **提 Issue / 提需求**:你的需求就是路线图
|
| 123 |
+
|
| 124 |
---
|
| 125 |
|
| 126 |
# 🪵 Perch (English)
|
| 127 |
|
| 128 |
+

|
| 129 |
+
|
| 130 |
**A local-first AI workbench that lives on your Mac — and actually gets work done.**
|
| 131 |
|
| 132 |
- ⚡ **0.08s first token** — local 0.6B model answers instantly, cloud model seamlessly takes over
|
| 133 |
- 🔒 **Privacy by design** — conversations, knowledge base and memories stay on your machine; optional E2E-encrypted sync
|
| 134 |
- 🤖 **Real agent** — reads/writes files, runs commands, searches the web, batch-processes, schedules tasks
|
| 135 |
- 🧠 **Bring your own models** — local Ollama + any cloud model (DeepSeek / MiMo / Qwen / OpenAI-compatible)
|
| 136 |
+
- 🧵 **Concurrent conversations** — talk to A while B is still generating
|
| 137 |
|
| 138 |
**Rust 74.7% · TypeScript 18.1% · Tauri 2 + React 19 · 1,118+ tests · MIT**
|
| 139 |
|
| 140 |
Download `Perch_4.5.0_aarch64.dmg` below (Apple Silicon). All AI APIs are self-configured.
|
| 141 |
|
| 142 |
+
**Star on GitHub** ⭐ https://github.com/ghongxv-web/perch
|
| 143 |
+
|
| 144 |
---
|
| 145 |
|
| 146 |
**栖木 Perch** — 让 AI 栖息于你的本机,安静而强大。
|