dsh-plan-steps 0.0.0-stage → 0.1.0
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/LICENSE +21 -0
- package/README.en.md +185 -0
- package/README.md +163 -2
- package/client.js +511 -0
- package/cordis.patch.yml +7 -0
- package/docs/DETAIL.md +161 -0
- package/icon.svg +6 -0
- package/index.js +958 -0
- package/locale/en.json +7 -0
- package/locale/zh.json +7 -0
- package/package.json +48 -4
package/LICENSE
ADDED
|
@@ -0,0 +1,21 @@
|
|
|
1
|
+
MIT License
|
|
2
|
+
|
|
3
|
+
Copyright (c) 2026 GrounzerLiu
|
|
4
|
+
|
|
5
|
+
Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
6
|
+
of this software and associated documentation files (the "Software"), to deal
|
|
7
|
+
in the Software without restriction, including without limitation the rights
|
|
8
|
+
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
9
|
+
copies of the Software, and to permit persons to whom the Software is
|
|
10
|
+
furnished to do so, subject to the following conditions:
|
|
11
|
+
|
|
12
|
+
The above copyright notice and this permission notice shall be included in all
|
|
13
|
+
copies or substantial portions of the Software.
|
|
14
|
+
|
|
15
|
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
16
|
+
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
17
|
+
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
|
18
|
+
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
19
|
+
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
20
|
+
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
|
21
|
+
SOFTWARE.
|
package/README.en.md
ADDED
|
@@ -0,0 +1,185 @@
|
|
|
1
|
+
# dsh-plan-steps
|
|
2
|
+
|
|
3
|
+
[中文](README.md) | English
|
|
4
|
+
|
|
5
|
+
Plan steps for DeepSeek Harness: `plan_write` records a two-level plan,
|
|
6
|
+
`complete_step` signs one item off and the host advances the list, a guard
|
|
7
|
+
refuses an unsigned completion, a turn is pulled back while items are open, and
|
|
8
|
+
the panel shows `计划 · 第 N / M 步` live.
|
|
9
|
+
|
|
10
|
+
The design follows [Reasonix](https://github.com/esengine/DeepSeek-Reasonix) —
|
|
11
|
+
its `todo_write` / `complete_step` / `AdvanceSerialTodo` semantics: two-level
|
|
12
|
+
phases and sub-steps, stable `step_id`, and the phase signing off last. This
|
|
13
|
+
plugin is a rewrite against DSH's event and slot model, not a port.
|
|
14
|
+
|
|
15
|
+
## What it adds
|
|
16
|
+
|
|
17
|
+
| Piece | Effect |
|
|
18
|
+
| --- | --- |
|
|
19
|
+
| `plan_write` | Records a two-level plan: `level` 0 phases, `level` 1 sub-steps, an optional stable `step_id`, exactly one current item. A rewrite is checked for continuity: an existing id is never dropped, the current item is never removed or moved back to pending, and completed history keeps its position. |
|
|
20
|
+
| `complete_step` | Signs the current `in_progress` item off, appends the advanced list as a `todo/write` snapshot (the panel follows), and names the next step. A phase signs off only after its sub-steps. |
|
|
21
|
+
| guard | Refuses a `plan_write` / `todo_write` that turns an item `completed` while the host still has it open. |
|
|
22
|
+
| turn hold | Steers the turn back when it is about to close with open items — once per turn. |
|
|
23
|
+
| `wait_for_user` | Declares a wait, so that turn may end with the item open. |
|
|
24
|
+
| request state | Carries the host list into the request when the visible window has lost it. |
|
|
25
|
+
| dock panel | Replaces the pinned task panel with `计划 · 第 N / M 步` + the current step + a segmented bar, expanded into a nested list. |
|
|
26
|
+
| step row | The same position on every `complete_step` tool row. |
|
|
27
|
+
|
|
28
|
+
The list is the latest `todo/write` event in the session log, so a fold, a fork,
|
|
29
|
+
or a restart derives the same list. Nothing is cached in plugin memory.
|
|
30
|
+
|
|
31
|
+
## Why separate writing a plan from certifying work
|
|
32
|
+
|
|
33
|
+
The shipped `todo_write` is the model's own statement: it both records the plan
|
|
34
|
+
and declares the work done. One action for both cannot tell "I finished it" from
|
|
35
|
+
"I want this off my list" — and after a fold the model no longer even holds the
|
|
36
|
+
table it would be editing.
|
|
37
|
+
|
|
38
|
+
Splitting them makes the distinction structural:
|
|
39
|
+
|
|
40
|
+
- `plan_write` owns the plan: which steps exist, in what order.
|
|
41
|
+
- `complete_step` owns execution: one item, the current one, only once it is done.
|
|
42
|
+
- The host owns the list, so "signed off" and "still open" are facts it can check
|
|
43
|
+
in the guard, at the end of a turn, and in the request.
|
|
44
|
+
|
|
45
|
+
## How advancement works
|
|
46
|
+
|
|
47
|
+
In order, depth first: a phase's sub-steps run in list order; the phase stays
|
|
48
|
+
`pending` while they run and becomes the current item only after all of them are
|
|
49
|
+
completed (the phase signs off last); only then does the next phase's first
|
|
50
|
+
sub-step become current.
|
|
51
|
+
|
|
52
|
+
```
|
|
53
|
+
1 调研与设计 pending
|
|
54
|
+
1.1 读现有实现 in_progress <- current
|
|
55
|
+
1.2 记录要点 pending
|
|
56
|
+
2 实现 pending
|
|
57
|
+
2.1 改代码 pending
|
|
58
|
+
```
|
|
59
|
+
|
|
60
|
+
Sign 1.1 and 1.2 becomes current; sign 1.2 and phase 1 becomes current awaiting
|
|
61
|
+
sign-off; sign 1 and 2.1 becomes current. The rules are enforced, not conventions:
|
|
62
|
+
a phase with unfinished sub-steps can neither be `in_progress` nor be signed off.
|
|
63
|
+
|
|
64
|
+
## Tools
|
|
65
|
+
|
|
66
|
+
**`plan_write`**
|
|
67
|
+
|
|
68
|
+
| Parameter | Meaning |
|
|
69
|
+
| --- | --- |
|
|
70
|
+
| `todos[].content` | Task name; required and unique |
|
|
71
|
+
| `todos[].status` | `pending` / `in_progress` / `completed` |
|
|
72
|
+
| `todos[].level` | `0` phase, `1` sub-step |
|
|
73
|
+
| `todos[].step_id` | Stable identity, copied verbatim on every rewrite |
|
|
74
|
+
| `todos[].activeForm` | Present-continuous label, optional |
|
|
75
|
+
|
|
76
|
+
**`complete_step`**
|
|
77
|
+
|
|
78
|
+
| Parameter | Meaning |
|
|
79
|
+
| --- | --- |
|
|
80
|
+
| `step_id` | Preferred: the stable identity |
|
|
81
|
+
| `step` | Item name, when there is no id |
|
|
82
|
+
| `index` | 1-based position, when neither of the above resolves |
|
|
83
|
+
| `result` | Required: what is now true after finishing the step |
|
|
84
|
+
|
|
85
|
+
**`wait_for_user`**
|
|
86
|
+
|
|
87
|
+
| Parameter | Meaning |
|
|
88
|
+
| --- | --- |
|
|
89
|
+
| `need` | One line naming what you are waiting for; placeholders are refused |
|
|
90
|
+
|
|
91
|
+
## Install
|
|
92
|
+
|
|
93
|
+
Local directory, no build step, no runtime dependencies:
|
|
94
|
+
|
|
95
|
+
```sh
|
|
96
|
+
# in the Harness UI: Plugins > Add plugin > pick this directory
|
|
97
|
+
# or with the manager tool:
|
|
98
|
+
# install_bundle(target: "D:/dsh-plugins/dsh-plan-steps")
|
|
99
|
+
```
|
|
100
|
+
|
|
101
|
+
From GitHub (real files land in the profile, so the client half is served too):
|
|
102
|
+
|
|
103
|
+
```sh
|
|
104
|
+
dsh plugin --profile desktop add github:GrounzerLiu/dsh-plan-steps
|
|
105
|
+
# then add dsh-plan-steps to the profile's dsh.profile.bundles
|
|
106
|
+
```
|
|
107
|
+
|
|
108
|
+
The bundle inserts one row, `include:plan-steps`, and puts `plan_write`,
|
|
109
|
+
`complete_step` and `wait_for_user` into the model's toolset.
|
|
110
|
+
|
|
111
|
+
**The `client:` half needs a tarball / git install.** With `link:` the Web host
|
|
112
|
+
cannot serve `/plugins/<id>/client.js` (the path resolves outside the profile and
|
|
113
|
+
the browser gets a 404), so the panel stays the shipped one; the host-side tools
|
|
114
|
+
are unaffected. Reload the page once after installing so the client module enters
|
|
115
|
+
the boot graph.
|
|
116
|
+
|
|
117
|
+
## Configure
|
|
118
|
+
|
|
119
|
+
The three switches live on the **plugin's own page**: find the `plan-steps` row in
|
|
120
|
+
the Plugin Manager and press Configure.
|
|
121
|
+
|
|
122
|
+
| Switch | Default | Effect |
|
|
123
|
+
| --- | --- | --- |
|
|
124
|
+
| Refuse unsigned completions | on | off: `plan_write` / `todo_write` may mark an item completed itself |
|
|
125
|
+
| Pull a turn back while items are open | on | off: a turn may end with the list still open |
|
|
126
|
+
| Rebuild the panel snapshot each turn | on | off: after `turn/start` the panel waits for the model's next write |
|
|
127
|
+
|
|
128
|
+
The same fields can be set under the row's `config` in the profile's
|
|
129
|
+
`cordis.patch.yml` as deployment defaults; the plugin page wins. The tools stay
|
|
130
|
+
registered either way.
|
|
131
|
+
|
|
132
|
+
## Uninstall
|
|
133
|
+
|
|
134
|
+
`remove_bundle` in the Plugin Manager, or disable the `plan-steps` row. Nothing is
|
|
135
|
+
written outside the plugin directory, so removal needs no cleanup; lists written
|
|
136
|
+
before removal stay in the session log and remain readable by the shipped tool.
|
|
137
|
+
|
|
138
|
+
## Tests
|
|
139
|
+
|
|
140
|
+
```sh
|
|
141
|
+
node --test test/
|
|
142
|
+
```
|
|
143
|
+
|
|
144
|
+
Nineteen cases: the five two-level advancement positions, the three rewrite
|
|
145
|
+
continuity rules, serial validation, call replay.
|
|
146
|
+
|
|
147
|
+
## Troubleshooting
|
|
148
|
+
|
|
149
|
+
**The panel is blank for a moment at the start of a turn.** The shipped task
|
|
150
|
+
projection resets on `turn/start`. This plugin writes the open list back on the
|
|
151
|
+
next macrotask, and the turn's first step restores it again if that write fails. A
|
|
152
|
+
failed early write logs one warning per session (`plan-steps: could not restore
|
|
153
|
+
the task panel at turn start`), so a host change shows up in the log instead of as
|
|
154
|
+
a silent blank panel.
|
|
155
|
+
|
|
156
|
+
**`complete_step` says the session has no task list.** No `plan_write` has
|
|
157
|
+
succeeded yet, or the host exposes neither a projection nor a session log. Create
|
|
158
|
+
the list first.
|
|
159
|
+
|
|
160
|
+
**A completion is refused.** The guard or the continuity check fired: an item was
|
|
161
|
+
marked `completed` in `plan_write`, or a rewrite dropped a `step_id`, removed or
|
|
162
|
+
regressed the current item, or moved completed history. Sign off with
|
|
163
|
+
`complete_step`, then rewrite.
|
|
164
|
+
|
|
165
|
+
**A turn keeps getting pulled back.** Expected once per turn while items are open.
|
|
166
|
+
To end a turn deliberately, call `wait_for_user` with what you need; to stop the
|
|
167
|
+
behaviour entirely, set `PLAN_STEPS_HOLD=0`.
|
|
168
|
+
|
|
169
|
+
**`step "..." matches no task-list item`.** The list is rebuilt from the log, so
|
|
170
|
+
the reference must match an item in the current snapshot; the error lists every
|
|
171
|
+
item with its status and position.
|
|
172
|
+
|
|
173
|
+
## Licence
|
|
174
|
+
|
|
175
|
+
MIT — see [LICENSE](LICENSE).
|
|
176
|
+
|
|
177
|
+
The panel inlines path data from [Material Symbols](https://fonts.google.com/icons)
|
|
178
|
+
(Google, Apache-2.0): the checklist, check-circle and chevron glyphs in
|
|
179
|
+
`client.js`.
|
|
180
|
+
|
|
181
|
+
## Design notes
|
|
182
|
+
|
|
183
|
+
No hard final gate, the panel still riding the shipped projection, and levels
|
|
184
|
+
travelling through the content as a zero-width marker are deliberate choices; the
|
|
185
|
+
reasoning is in [docs/DETAIL.md](docs/DETAIL.md).
|
package/README.md
CHANGED
|
@@ -1,3 +1,164 @@
|
|
|
1
|
-
#
|
|
1
|
+
# dsh-plan-steps
|
|
2
2
|
|
|
3
|
-
|
|
3
|
+
中文 | [English](README.en.md)
|
|
4
|
+
|
|
5
|
+
给 DeepSeek Harness 用的「计划步骤」:`plan_write` 记录两级计划,`complete_step`
|
|
6
|
+
一项一项签核并由宿主推进,守卫拒绝未签核的完成,列表还开着时轮末会被拉回,面板上
|
|
7
|
+
实时显示「计划 · 第 N / M 步」。
|
|
8
|
+
|
|
9
|
+
设计参考了 [Reasonix](https://github.com/esengine/DeepSeek-Reasonix) 的
|
|
10
|
+
`todo_write` / `complete_step` / `AdvanceSerialTodo` 语义(两级阶段与子步骤、稳定
|
|
11
|
+
`step_id`、阶段最后签核)。这个插件是按 DSH 的事件与槽位模型重写的实现,不是它的
|
|
12
|
+
移植。
|
|
13
|
+
|
|
14
|
+
## 带来什么
|
|
15
|
+
|
|
16
|
+
| 部件 | 作用 |
|
|
17
|
+
| --- | --- |
|
|
18
|
+
| `plan_write` | 记录两级计划:`level` 0 是阶段,`level` 1 是它的子步骤,每项可带稳定 `step_id`,全表只有一个当前项。重写时执行连续性校验:已有的 id 不能丢、当前项不能被删或降级、已完成项保持原位。 |
|
|
19
|
+
| `complete_step` | 签掉当前 `in_progress` 那一项,宿主推进列表并写入 `todo/write` 快照(面板随之更新),再把下一步告诉你。阶段必须等子步骤全部完成后才能签。 |
|
|
20
|
+
| 守卫 | 列表里还开着的项被 `plan_write` / `todo_write` 直接标成 `completed` 时拒绝。 |
|
|
21
|
+
| 收尾拉回 | 轮末列表还有未完成项时,用 `agent.steer()` 把这一轮拉回来——每轮最多一次。 |
|
|
22
|
+
| `wait_for_user` | 声明「在等用户」,这一轮就不会被拉回。 |
|
|
23
|
+
| 上下文召回 | 可见窗口里没有当前列表时,把「Host task state」块注入请求。 |
|
|
24
|
+
| 面板 | 接管输入框上方的任务面板,显示「计划 · 第 N / M 步」+ 当前项 + 分段进度条,展开是两级缩进列表。 |
|
|
25
|
+
| 工具行 | 每次 `complete_step` 也会带同样的位置信息。 |
|
|
26
|
+
|
|
27
|
+
列表本身是会话日志里最后一条 `todo/write` 事件,所以折叠、分支、重启之后推导出的
|
|
28
|
+
是同一份列表;插件内存里不缓存任何状态。
|
|
29
|
+
|
|
30
|
+
## 为什么要把「写计划」和「签核」分开
|
|
31
|
+
|
|
32
|
+
内置的 `todo_write` 是模型自己的一句话:它既记录计划,又声明完工。这两件事混在一个
|
|
33
|
+
动作里,就分不出「我做完了」和「我想把它从列表划掉」;而且折叠之后,模型手里连那张
|
|
34
|
+
表都没有了。
|
|
35
|
+
|
|
36
|
+
拆开之后,边界变成结构性的:
|
|
37
|
+
|
|
38
|
+
- `plan_write` 管计划:有哪些步骤、什么顺序
|
|
39
|
+
- `complete_step` 管执行:一次一项、只签当前项、签完才推进
|
|
40
|
+
- 宿主管列表:于是「签过」和「还开着」是它能在守卫、轮末、请求三处校验的事实
|
|
41
|
+
|
|
42
|
+
## 推进规则
|
|
43
|
+
|
|
44
|
+
顺序推进、深度优先:阶段内的子步骤按列表顺序做;子步骤跑的时候阶段保持 `pending`,
|
|
45
|
+
全部完成之后阶段才成为当前项(阶段最后签);签掉阶段才轮到下一个阶段的第一个子步骤。
|
|
46
|
+
|
|
47
|
+
```
|
|
48
|
+
1 调研与设计 pending
|
|
49
|
+
1.1 读现有实现 in_progress ← 当前
|
|
50
|
+
1.2 记录要点 pending
|
|
51
|
+
2 实现 pending
|
|
52
|
+
2.1 改代码 pending
|
|
53
|
+
```
|
|
54
|
+
|
|
55
|
+
签 1.1 → 1.2 成为当前;签 1.2 → 阶段 1 成为当前等待签收;签 1 → 2.1 成为当前。
|
|
56
|
+
这三条在数据上强制:阶段带着未完成子步骤时既不能是 `in_progress`,也不能被签核。
|
|
57
|
+
|
|
58
|
+
## 工具
|
|
59
|
+
|
|
60
|
+
**`plan_write`**
|
|
61
|
+
|
|
62
|
+
| 参数 | 说明 |
|
|
63
|
+
| --- | --- |
|
|
64
|
+
| `todos[].content` | 任务名,必填且不能重复 |
|
|
65
|
+
| `todos[].status` | `pending` / `in_progress` / `completed` |
|
|
66
|
+
| `todos[].level` | `0` 阶段,`1` 子步骤 |
|
|
67
|
+
| `todos[].step_id` | 稳定身份,重写时原样抄回 |
|
|
68
|
+
| `todos[].activeForm` | 进行时描述,可选 |
|
|
69
|
+
|
|
70
|
+
**`complete_step`**
|
|
71
|
+
|
|
72
|
+
| 参数 | 说明 |
|
|
73
|
+
| --- | --- |
|
|
74
|
+
| `step_id` | 首选,稳定身份 |
|
|
75
|
+
| `step` | 项目名,没有 id 时用 |
|
|
76
|
+
| `index` | 1-based 序号,前两者都不可用时用 |
|
|
77
|
+
| `result` | 必填,说明这一步之后什么成立了 |
|
|
78
|
+
|
|
79
|
+
**`wait_for_user`**
|
|
80
|
+
|
|
81
|
+
| 参数 | 说明 |
|
|
82
|
+
| --- | --- |
|
|
83
|
+
| `need` | 一句话说清在等什么;占位词会被拒绝 |
|
|
84
|
+
|
|
85
|
+
## 安装
|
|
86
|
+
|
|
87
|
+
本地目录安装,无构建步骤、无运行时依赖:
|
|
88
|
+
|
|
89
|
+
```sh
|
|
90
|
+
# 界面:插件 > 添加插件 > 选这个目录
|
|
91
|
+
# 或用管理工具:
|
|
92
|
+
# install_bundle(target: "D:/dsh-plugins/dsh-plan-steps")
|
|
93
|
+
```
|
|
94
|
+
|
|
95
|
+
从 GitHub 安装(真实文件落进 profile,客户端半也能被服务):
|
|
96
|
+
|
|
97
|
+
```sh
|
|
98
|
+
dsh plugin --profile desktop add github:GrounzerLiu/dsh-plan-steps
|
|
99
|
+
# 再把 dsh-plan-steps 加进 profile 的 dsh.profile.bundles
|
|
100
|
+
```
|
|
101
|
+
|
|
102
|
+
bundle 会插入一行 `include:plan-steps`,并让 `plan_write`、`complete_step`、
|
|
103
|
+
`wait_for_user` 三个工具进入模型工具集。
|
|
104
|
+
|
|
105
|
+
**客户端半需要 tarball / git 安装。** `link:` 安装时 Web 端取不到
|
|
106
|
+
`/plugins/<id>/client.js`(路径落在 profile 之外,浏览器 404),面板不会变样;
|
|
107
|
+
宿主侧的工具不受影响。装完后刷新一次页面,客户端模块才会进入启动图。
|
|
108
|
+
|
|
109
|
+
## 配置
|
|
110
|
+
|
|
111
|
+
三个开关在**插件自己的页面**上:插件管理器里找到 `plan-steps` 那一行,点右侧的「配置」进入,页面上直接勾选/取消。
|
|
112
|
+
|
|
113
|
+
| 开关 | 默认 | 作用 |
|
|
114
|
+
| --- | --- | --- |
|
|
115
|
+
| 拒绝未签核的完成 | 开 | 关掉后 `plan_write` / `todo_write` 可以自己把项标成完成 |
|
|
116
|
+
| 列表开着时拉回一轮 | 开 | 关掉后轮末不再 steer,列表开着也能结束 |
|
|
117
|
+
| 每轮重建面板快照 | 开 | 关掉后 `turn/start` 之后面板要等模型自己写一次才恢复 |
|
|
118
|
+
|
|
119
|
+
同样的字段也可以写在 profile 的 `cordis.patch.yml` 里该行的 `config` 下,作为部署默认;插件页上的值优先。关掉开关不影响工具注册。
|
|
120
|
+
|
|
121
|
+
## 卸载
|
|
122
|
+
|
|
123
|
+
插件管理里 `remove_bundle`,或关掉 `plan-steps` 那一行。插件不往目录外写东西,删了
|
|
124
|
+
不需要清理;之前写下的任务列表仍留在会话日志里,内置面板照样能读。
|
|
125
|
+
|
|
126
|
+
## 测试
|
|
127
|
+
|
|
128
|
+
```sh
|
|
129
|
+
node --test test/
|
|
130
|
+
```
|
|
131
|
+
|
|
132
|
+
19 个用例,覆盖两级推进的五种走位、重写的三条连续性规则、序列校验、调用重放等。
|
|
133
|
+
|
|
134
|
+
## 排错
|
|
135
|
+
|
|
136
|
+
**每轮开始面板会空一下。** 内置任务投影在 `turn/start` 重置。插件在下一个宏任务把
|
|
137
|
+
未完成的列表写回去,第一步还会再兜底一次;早写失败时每个会话记一条警告
|
|
138
|
+
(`plan-steps: could not restore the task panel at turn start`),所以宿主行为变了
|
|
139
|
+
能在日志里看出来,而不是悄悄空白。
|
|
140
|
+
|
|
141
|
+
**`complete_step` 说没有任务列表。** 本会话还没有成功的 `plan_write`,或宿主既没有
|
|
142
|
+
投影也没有会话日志可读。先建列表。
|
|
143
|
+
|
|
144
|
+
**完成被拒绝。** 守卫或连续性校验拦下了:可能在 `plan_write` 里直接标了
|
|
145
|
+
`completed`,也可能重写时丢掉了 id、删掉或降级了当前项、移动了已完成项。先
|
|
146
|
+
`complete_step`,再改列表。
|
|
147
|
+
|
|
148
|
+
**一轮总被拉回。** 列表还开着时每轮最多拉一次,这是设计。要正常结束就调
|
|
149
|
+
`wait_for_user` 说明在等什么;要彻底关掉就设 `PLAN_STEPS_HOLD=0`。
|
|
150
|
+
|
|
151
|
+
**`step "..." matches no task-list item`。** 列表是从日志重建的,引用必须命中当前
|
|
152
|
+
快照里的项;报错会列出每一项的状态与位置。
|
|
153
|
+
|
|
154
|
+
## 许可
|
|
155
|
+
|
|
156
|
+
MIT,见 [LICENSE](LICENSE)。
|
|
157
|
+
|
|
158
|
+
面板里的图标路径数据来自 [Material Symbols](https://fonts.google.com/icons)
|
|
159
|
+
(Google,Apache-2.0):`client.js` 里的 checklist / check-circle / chevron。
|
|
160
|
+
|
|
161
|
+
## 设计取舍
|
|
162
|
+
|
|
163
|
+
没有硬性收尾闸门、面板仍归内置那份投影、层级靠内容里的零宽标记穿过投影——这些是
|
|
164
|
+
刻意的取舍,理由写在 [docs/DETAIL.md](docs/DETAIL.md)。
|