org card: remove README.md (re-add next, to refresh the card)
Browse files
README.md
DELETED
|
@@ -1,46 +0,0 @@
|
|
| 1 |
-
---
|
| 2 |
-
title: README
|
| 3 |
-
emoji: 🟧
|
| 4 |
-
colorFrom: gray
|
| 5 |
-
colorTo: red
|
| 6 |
-
sdk: static
|
| 7 |
-
pinned: false
|
| 8 |
-
short_description: Open-source computer use for your Mac
|
| 9 |
-
---
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
**Small enough to run on your Mac. Smart enough to ask.**
|
| 14 |
-
小到能在你的 Mac 上跑，聪明到知道该问你。
|
| 15 |
-
|
| 16 |
-
[Website · 官网](https://deskmind.dev) · [Download for Mac (0.3.0)](https://github.com/deskmind-ai/app/releases/latest) ·
|
| 17 |
-
[GitHub](https://github.com/deskmind-ai) · [Bench](https://github.com/deskmind-ai/bench)
|
| 18 |
-
|
| 19 |
-
DeskMind is open-source computer use for your Mac. It reads the screen, works out the next step and acts, with small
|
| 20 |
-
models running locally with MLX. When a task could mean two things, it asks instead of guessing.
|
| 21 |
-
|
| 22 |
-
| Model | What it does |
|
| 23 |
-
|---|---|
|
| 24 |
-
| [brain-0.8b](https://huggingface.co/deskmind/brain-0.8b) | Fast tier: decides each step, with a probability for every option (MLX, 8-bit) |
|
| 25 |
-
| [brain-4b](https://huggingface.co/deskmind/brain-4b) | Strong tier: checks the steps the 0.8B is unsure of (MLX, 8-bit) |
|
| 26 |
-
| [eyes-4b](https://huggingface.co/deskmind/eyes-4b) | Screen grounding: finds the target on a screenshot, for apps without an accessibility tree |
|
| 27 |
-
|
| 28 |
-
**Current release: G18b** (revision `g18b-q8`, also `main`). On the real macOS desktop (bench v25, 13 tasks × 3 runs
|
| 29 |
-
through the DeskMind app on an M4 Pro), the 0.8B → 4B router at threshold 0.96 passed **39 of 39** runs and never said
|
| 30 |
-
"done" early. A decision takes about 0.5 s when the 0.8B answers and about 3.6 s when the 4B checks. Small sample;
|
| 31 |
-
what it still gets wrong is in [brain results](https://github.com/deskmind-ai/brain/blob/main/docs/results.md). Any
|
| 32 |
-
agent can use the planner through `/v1/systemone`.
|
| 33 |
-
|
| 34 |
-
---
|
| 35 |
-
|
| 36 |
-
**DeskMind 得心**是在你 Mac 上运行的开源电脑操作 AI：看屏幕、想下一步、动手操作，靠的是本机的小模型（MLX）；任务有两种理解时，
|
| 37 |
-
它先问你，不瞎猜。
|
| 38 |
-
|
| 39 |
-
- **brain-0.8b**：判断每一步，每个选项都给出概率；**brain-4b**：复核 0.8B 没把握的步骤；**eyes-4b**：在截图上找到要点的位置，
|
| 40 |
-
用于没有辅助功能结构的应用。
|
| 41 |
-
- **当前版本 G18b**：真实 macOS 桌面上（bench v25，13 个任务各跑 3 次，经 App 运行，M4 Pro），0.8B → 4B 路由（门槛 0.96）
|
| 42 |
-
39 次全部通过，没有一次没做完就说完成。样本不大，还做不好的地方见
|
| 43 |
-
[brain results](https://github.com/deskmind-ai/brain/blob/main/docs/results.zh-CN.md)。
|
| 44 |
-
- 国内下载：ModelScope 上有同样的文件（[gxcsoccer/brain-0.8b](https://www.modelscope.cn/models/gxcsoccer/brain-0.8b)、
|
| 45 |
-
[gxcsoccer/brain-4b](https://www.modelscope.cn/models/gxcsoccer/brain-4b)、
|
| 46 |
-
[gxcsoccer/eyes-4b](https://www.modelscope.cn/models/gxcsoccer/eyes-4b)）。
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|