@ninkch/dsh-infinite-canvas 0.19.21 → 0.20.1
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.md +14 -0
- package/README.zh-CN.md +14 -0
- package/assets/canvas-layout/SKILL.md +19 -2
- package/assets/canvas-studio/SKILL.md +156 -392
- package/client.js +305 -63
- package/cordis.patch.yml +13 -0
- package/docs/auto-creation-design.md +68 -0
- package/docs/guide.en.md +13 -1
- package/docs/guide.zh-CN.md +14 -0
- package/index.js +867 -14
- package/package.json +2 -1
- package/production.js +427 -0
- package/skill-pack.js +16 -3
package/README.md
CHANGED
|
@@ -63,6 +63,20 @@ Help & feedback:
|
|
|
63
63
|
|
|
64
64
|

|
|
65
65
|
|
|
66
|
+
## Production stage machine & auto creation (experimental)
|
|
67
|
+
|
|
68
|
+
Every canvas (one episode) carries a host-enforced **production stage** — `brief → bible → assets → script → storyboard → shots_ready → generating → qc → post → delivered` — with six code gates (bible-first, storyboard quality, one video per storyboard, no per-shot nodes, destructive-op confirmation, stage order). Gates refuse with the exact missing items and always accept an explicit-user override (audited). `canvas_production_status` is the only basis for claiming "done".
|
|
69
|
+
|
|
70
|
+
With the **Auto creation (experimental)** switch on (Settings → 自动创作), the agent runs a writers room from one sentence — research → two independent pitches → cross-critique → producer decision → bible → script → per-scene storyboards (quality-gated) → **budget sheet** — with zero paid calls; approve once and it generates videos per storyboard and builds a final-cut node. The transcript lands on the canvas as a 编剧室会议 node; questions are forbidden in this mode.
|
|
71
|
+
|
|
72
|
+
## Image models via `dsh-image-gen`
|
|
73
|
+
|
|
74
|
+
If the `dsh-image-gen` plugin is installed and configured, Settings shows 「图片模型(来自 dsh-image-gen)」: pick one of its providers/models and generate directly through its studio bridge — no base URL or API key in the canvas; credentials stay on the image-gen side.
|
|
75
|
+
|
|
76
|
+
## Precise layout
|
|
77
|
+
|
|
78
|
+
`canvas_layout_get` reads the whole board's geometry; `canvas_layout_node` batch-places nodes by coordinate (media nodes can resize, ratio-preserving) and can pin nodes so `canvas_relayout` never moves them. Relayout packs clusters in bands instead of one endless column and never covers pinned nodes.
|
|
79
|
+
|
|
66
80
|
## Work beside the chat
|
|
67
81
|
|
|
68
82
|
The canvas is not a separate window you have to switch to. It opens as a **tab in the Harness right pane**, so you can talk to the agent on the left and storyboard on the right **at the same time** — ask for a change in chat and watch the nodes update in place.
|
package/README.zh-CN.md
CHANGED
|
@@ -63,6 +63,20 @@
|
|
|
63
63
|
|
|
64
64
|

|
|
65
65
|
|
|
66
|
+
## 制作阶段机与自动创作(实验)
|
|
67
|
+
|
|
68
|
+
每个画布(一集)都有宿主强制的**制作阶段**:`brief → bible → assets → script → storyboard → shots_ready → generating → qc → post → delivered`,配合六道代码闸门(设定集先行 / 分镜质量 / 一表一视频 / 禁逐镜节点 / 破坏性确认 / 阶段越序)。闸门拒绝时会列出确切缺失项,并始终接受用户显式 override(进审计);宣称"完成"必须以 `canvas_production_status` 的 `ready:true` 为据。
|
|
69
|
+
|
|
70
|
+
打开「**自动创作(实验)**」开关(设置页)后,对话智能体从一句话创意跑通编剧室:调研 → 编剧A/B 独立提案 → 互评 → 监制决策 → 设定集 → 剧本 → 逐场分镜表(过质量闸门)→ **预算单**——全程零付费;批准一次后逐分镜表生成视频并建成片节点。讨论记录落在画布「编剧室会议」节点;该模式下**禁止向用户提问**。
|
|
71
|
+
|
|
72
|
+
## 图片模型:直接用 `dsh-image-gen`
|
|
73
|
+
|
|
74
|
+
装了 `dsh-image-gen` 并配置好模型后,设置页出现「图片模型(来自 dsh-image-gen)」:直接选它的服务商/模型生图——画布侧**无需**再填 baseURL/API Key,凭据留在 image-gen 侧。
|
|
75
|
+
|
|
76
|
+
## 精准布局
|
|
77
|
+
|
|
78
|
+
`canvas_layout_get` 读取全图几何;`canvas_layout_node` 按坐标批量摆放(媒体节点可改尺寸且保持画面比例),并可 `pinned` 钉住——`canvas_relayout` 永远不动钉住的节点,簇按带高打包、不再拍成一整列。
|
|
79
|
+
|
|
66
80
|
## 在对话旁边工作
|
|
67
81
|
|
|
68
82
|
画布不是必须切过去的另一个窗口——它可以作为 **Harness 右侧面板里的一个页签**打开,左边对话、右边分镜,**同时进行**:在对话里提要求,右边节点实时跟着变。
|
|
@@ -21,8 +21,8 @@ description: 画布节点布局纪律:一类型一列、竖向排列、绝不
|
|
|
21
21
|
2. **连线不改列**:连线只表达数据流(剧本→分镜、分镜→视频、资产→参考图)。
|
|
22
22
|
分镜表不会因为连了"某场文本"就移动到那一列旁边——它永远待在「分镜表」列里。
|
|
23
23
|
3. **绝不重叠**:任意两个节点框不得相交,最小间距 24px。
|
|
24
|
-
|
|
25
|
-
|
|
24
|
+
插件已内建自动挤开(节点变高、被拖动、被拉大时,相交的节点会被自动推到其下方)。
|
|
25
|
+
**批量/自动建点不要手动挪别人的节点**;用户要求"精准放到某处 / 对齐 / 分列"时,用下方「精确摆放」两个工具(明确意图允许重叠)。
|
|
26
26
|
4. **建点顺序就是叙事顺序**:同类型节点按 场次/序号 顺序创建,
|
|
27
27
|
后建的自动追加在列尾,列的竖向顺序天然等于叙事顺序。
|
|
28
28
|
**不要回填到列中间**(会造成列内乱序,后续只能靠 `canvas_relayout` 修正)。
|
|
@@ -36,6 +36,23 @@ description: 画布节点布局纪律:一类型一列、竖向排列、绝不
|
|
|
36
36
|
「铺开到场景 / 存为资产」放在生成按钮下方、正文之前,用户要找很久)。
|
|
37
37
|
输入类操作(上传、选模型、清空)不受此限,留在输入区。
|
|
38
38
|
|
|
39
|
+
## 一·五、精确摆放(0.20.0)
|
|
40
|
+
|
|
41
|
+
自动排版的列纪律只解决"批量建点不打架"。**用户或对话要求"把某个节点放到某处 / 对齐 / 按坐标调"时,必须用精确工具**:
|
|
42
|
+
|
|
43
|
+
- `canvas_layout_get`:先读全图几何(每个节点的 `nodeId/name/type/x/y/w/h/pinned` + 画布包围盒)。**不许靠猜坐标**。
|
|
44
|
+
- `canvas_layout_node`:一次批量摆放 `{ nodes: [{ nodeId, x, y, w?, h?, pinned? }] }`。
|
|
45
|
+
- `x/y` 任意有限数值(可为负);**允许重叠**——精确摆放是明确意图,不做碰撞拒绝;
|
|
46
|
+
- `w/h` 只有 `image` / `video` 媒体节点可改(w≥160、h≥120);
|
|
47
|
+
- `pinned: true` 钉住:`canvas_relayout` 重排时**跳过它**(保留你摆好的位置)。
|
|
48
|
+
- 何时用哪个:
|
|
49
|
+
| 需求 | 工具 |
|
|
50
|
+
|---|---|
|
|
51
|
+
| "把这三张参考图竖排在 x=100、间距 480" / "把视频列对齐到 x=2600" | `canvas_layout_get` + `canvas_layout_node` |
|
|
52
|
+
| "整块太乱了,大体理一下" | `canvas_relayout`(粗排兜底,尊重 pinned) |
|
|
53
|
+
|
|
54
|
+
精摆完把关键节点 `pinned:true`,后续任何 `canvas_relayout` 都不会冲掉你的布局。
|
|
55
|
+
|
|
39
56
|
## 二、标准建点节奏(保持列干净的写法)
|
|
40
57
|
|
|
41
58
|
```
|