dsh-builtin-browser 0.1.0 → 0.1.1
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.en.md +8 -0
- package/README.md +8 -0
- package/package.json +1 -1
package/README.en.md
CHANGED
|
@@ -60,6 +60,14 @@ The plugin mounts through `cordis.patch.yml`; per-row config:
|
|
|
60
60
|
| `browser_screenshot` | PNG capture, optional `fullPage` |
|
|
61
61
|
| `browser_list_tabs` / `browser_switch_tab` / `browser_close_tab` / `browser_reset` | Multi-tab session management |
|
|
62
62
|
|
|
63
|
+
### Interaction discipline (clicking / filling forms)
|
|
64
|
+
|
|
65
|
+
- **Prefer DOM semantics over coordinates**: submit forms with `form.requestSubmit()`, click with `element.click()`; coordinate clicks are the last resort.
|
|
66
|
+
- **Target the right element**: pages often carry hidden copies (e.g. mobile buttons). Use `browser_execute` to filter visible elements (`getBoundingClientRect()` with width/height > 0, computed style not `display:none`) before reading coordinates.
|
|
67
|
+
- **Click immediately after reading coordinates**: do not interleave other actions (filling, scrolling moves elements; stale coordinates fail instantly).
|
|
68
|
+
- **Verify the hit before clicking**: `document.elementFromPoint(x, y)` confirms the coordinate actually lands on the target button/link before issuing a real click.
|
|
69
|
+
- **Mind DPR**: CDP input uses CSS pixels; on high-DPI screens, calibrate with `elementFromPoint` instead of blind coordinate tries.
|
|
70
|
+
|
|
63
71
|
## How it works
|
|
64
72
|
|
|
65
73
|
```
|
package/README.md
CHANGED
|
@@ -60,6 +60,14 @@ provider 与工具以 `ctx.get('electronViewHost')` 是否存在为门控,因此
|
|
|
60
60
|
| `browser_screenshot` | PNG 截图,可选 `fullPage` |
|
|
61
61
|
| `browser_list_tabs` / `browser_switch_tab` / `browser_close_tab` / `browser_reset` | 多标签会话管理 |
|
|
62
62
|
|
|
63
|
+
### 操作纪律(点击/填表)
|
|
64
|
+
|
|
65
|
+
- **优先用 DOM 语义而非坐标**:表单提交优先 `form.requestSubmit()`;点击优先 `element.click()`;坐标点击是最后手段。
|
|
66
|
+
- **选中正确的元素**:页面常有隐藏副本(如移动端按钮),用 `browser_execute` 过滤可见元素(`getBoundingClientRect()` 宽高 > 0、`getComputedStyle` 非 `display:none`),再取坐标。
|
|
67
|
+
- **取坐标后立即点击**:中间不要插入其他操作(填表、滚动会移动元素,旧坐标立即失效)。
|
|
68
|
+
- **点击前验证命中**:`document.elementFromPoint(x, y)` 确认该坐标确实是目标元素(按钮/链接),再执行真实点击。
|
|
69
|
+
- **DPR 注意**:CDP 输入使用 CSS 像素;高 DPI 屏上若点击落空,用 `elementFromPoint` 校准,不要盲试坐标。
|
|
70
|
+
|
|
63
71
|
## 工作原理
|
|
64
72
|
|
|
65
73
|
```
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "dsh-builtin-browser",
|
|
3
|
-
"version": "0.1.
|
|
3
|
+
"version": "0.1.1",
|
|
4
4
|
"description": "Shared real browser capability for DeepSeek Harness: a visible browser the human watches and can take over, driven by the agent over CDP (snapshot/execute/content/tab management); host-side seam, provider, and browser_* tools",
|
|
5
5
|
"type": "module",
|
|
6
6
|
"main": "lib/index.js",
|