2753 Commits

Author SHA1 Message Date
134
9611fafe0b feat(flow-board): add native XMind markers and bulk editing 2026-10-05 11:49:36 +08:00
134
9f98a68049 fix(flow-board): support image paste into mind map topics
Handle clipboard images while editing or selecting a topic and route Insert image through the same topic attachment behavior. Read image data locally in preview mode and retain the production upload path.

Add four browser regression cases for real clipboard paste, undo/redo, export resources, canvas images, and text input isolation. Record validation in the client comparison notes.
2026-10-05 10:56:37 +08:00
134
822c4cbfb9 fix(flow-board): align Xmind editor UI and restore marquee feedback
Add native-style menus, search and presentation controls, typography resources, and Xmind import/export styling support.

Bundle canvas styles for preview and production, show marquee borders and selection fills, and align editor density with the project UI. Include interaction regression coverage and client comparison notes.
2026-10-04 18:57:02 +08:00
134
a492d082ed feat: recreate XMind editing and file workflows 2026-10-04 18:56:53 +08:00
134
4ada7eaf38 fix: polish canvas icons, context menu and panel layout 2026-10-04 12:13:47 +08:00
134
63e2f48192 fix: restore canvas skills and local uploads, simplify home and canvas headers 2026-10-04 10:31:53 +08:00
134
d376f5959f feat: refresh home and canvas UI with customizable image toolbar 2026-10-04 03:41:03 +08:00
134
2c8f306e74 docs(tvstar): record live production evidence and remaining gaps 2026-10-03 17:58:58 +08:00
134
c4a2753a6e fix(tvstar): enforce structured video analysis and reject truncation 2026-10-03 17:48:41 +08:00
134
0d752d6e38 fix(character-studio): separate identity from view composition 2026-10-03 17:34:01 +08:00
134
96d67955cb fix(canvas): preserve uncertain image submissions without paid failover 2026-10-03 17:29:31 +08:00
134
c45a6a77a0 fix(tvstar): isolate shot inputs and reject observed visual defects 2026-10-03 16:56:57 +08:00
134
88cf96b998 fix(tvstar): validate live video contracts and avoid duplicate planning 2026-10-03 16:18:56 +08:00
134
570375b5f1 fix(video-canvas): enforce production delivery requirements and analyze real footage 2026-10-03 15:33:10 +08:00
134
e7be5af3f2 feat(video-canvas): add TVstar creation and session workflows 2026-10-01 22:58:20 +08:00
134
365bad8539 fix(canvas): distinguish provider balance from image rate limits 2026-10-01 07:45:39 +08:00
134
08eecb0190 docs: record the UUID validation regression that broke detail replacement 2026-09-27 19:19:20 +08:00
134
f75d99baf6 fix(agent): repair the UUID check that made every detail replacement request 400
新增的批次接口把一个 UUID 正则写成了 8-4-4-12(少一组 -{4}),标准 UUID 是
8-4-4-4-12。结果是任何真实项目编号和请求编号都过不了校验:查询接口恒返 400
「项目编号无效」,提交接口恒返 400「任务编号或项目编号无效」,详情替换完全无法使用。

- 校验抽成 isDetailReplacementUuid 并修正为规范的 8-4-4-4-12,路由不再自带正则
- 回归测试覆盖真实线上项目编号、浏览器生成的请求编号,以及导致故障的 8-4-4-12 形状
- 增加源码守卫,防止该路由再次出现自带的错误 UUID 正则
- 前端在项目尚未保存时给出明确提示,而不是把空编号丢给接口
2026-09-27 19:14:56 +08:00
134
db3a038b00 docs: record the detail replacement durable batch incident and rollout 2026-09-26 02:53:01 +08:00
134
0907aaff67 fix(agent): make detail replacement results survive page and server restarts
详情替换原来把整批生图挂在浏览器保持的对话流上,切换标签页、断网或连接中断后
结果就再也回不来,前端还会继续播放按时间推算的 96% 假进度。

现在改为服务器端后台批次:提交时先落任务,再按参考图逐张生成并把每张结果写进
项目,页面刷新、切页或服务重启后都能按任务状态恢复,不再依赖前端连接。

- 新增 /api/agents/detail-replacement/batches(提交/查询、幂等去重、重试上限)
- 前端只在服务端确认批次后显示处理中,卡片显示排队/生成/完成/失败真实状态
- 去掉假进度动画;没有产品输入图时提交前直接拦截
- 服务重启后自动恢复未完成批次,长批次不再被通用 30 分钟清理误判
- 单张失败只落到该卡片并继续后续图片,已完成的图片不重复生成和扣费
- 详情替换默认模型由持续 404 的 gpt-image-2.5-flare 改回 gpt-image-2.5
2026-09-26 02:39:11 +08:00
134
1c2acee547 Preserve uploaded media aspect ratios in video canvas 2026-09-24 14:36:22 +08:00
134
439af60c2a Update video canvas home and Aliyun analysis flow 2026-09-24 13:36:31 +08:00
134
e97296a9e8 Add share entry to the canvas top bar
The share popover (publish to site, copy link) and publish modal had
no reachable entry; add a share button next to the save status,
matching the reference product's top bar.
2026-09-19 17:09:39 +08:00
134
94b49b7fad Document shortcut system rollout and server-side verification 2026-09-19 14:57:29 +08:00
134
ea4f34bf89 Re-enable the video toolbar HD entry
The P40 upscale flow (popover, HD node creation, task submission) is
fully wired; the toolbar entry had been left disabled and mislabeled
as not connected.
2026-09-19 14:42:52 +08:00
134
1141ecc43a Bound separation input WAV duration to the probed length
The unbounded apad filter made ffmpeg pad the 3-second test clip
endlessly (a 22.9GB WAV) until the disk filled, which was the real
cause of the voice separation failures. Trim to the probed duration
and drop apad.
2026-09-19 14:33:06 +08:00
134
8b2955eb0b Add canvas undo/redo and full shortcut system
Snapshot-based undo/redo (max 50 steps) with Cmd+Z / Cmd+Shift+Z,
ignoring volatile generation progress fields. Bind the remaining
reference shortcuts: Cmd+D duplicate, Cmd+Enter generate selection,
Tab new-node palette, Cmd+F canvas node search panel, Cmd+0 fit view,
Cmd+=/- zoom, V pan-tool toggle, H hold-to-pan, plain G to group.
Expand the shortcut help panel to the full four-group layout.
2026-09-19 12:53:07 +08:00
134
94b3ac9cb3 Restyle story script nodes to dark theme
Align the story script generator table and script text node with the
reference product: dark cards with subtle white borders, muted header
row, dark status pills, white primary button and dark textarea card,
replacing the previous light table chrome.
2026-09-18 22:57:07 +08:00
134
8a1876e016 Render canvas image nodes with object-contain
Show the full picture inside image resource nodes instead of cropping
with object-cover; matches the reference product where node frames
adapt to the image and no content is clipped, and fixes existing
nodes created before aspect fitting was added.
2026-09-18 18:45:39 +08:00
134
b4b6e1a738 Size dropped image nodes to intrinsic aspect ratio
Dropped images landed in fixed 420x280 nodes and rendered with
object-cover, cropping portraits and wide photos. Read the image's
natural size during ingestion and fit the node frame to the aspect
ratio (landscape caps height, portrait caps width) so the full
picture stays visible like the reference product.
2026-09-18 18:34:23 +08:00
134
885cab9fca Document voice separation rollout 2026-09-18 18:21:47 +08:00
134
6ed8879b28 Wire canvas voice/background separation to remote HTDemucs service
The private FunASR/HTDemucs audio service was already live and used by
the smart edit pipeline, but the canvas video toolbar still claimed
voice/background separation was unavailable. Add a dedicated
/api/canvas/audio-separate route (download, ffmpeg stem extraction,
remote separation, storage upload) and route the toolbar's voice and
background modes through it, replacing the stale not-connected notice.
2026-09-18 18:16:44 +08:00
134
c1975942ac Document video canvas shell cleanup and media upload fix rollout 2026-09-18 18:04:51 +08:00
134
a4354dc535 Remove video canvas demo shells and fix media upload timeouts
- Starter/toolbox presets no longer seed fake generated results; demo
  fallback media URLs in tool actions now fail with honest guidance
- Delete dead demo scene components, fake director console, demo
  history constants and unreachable submit toast fallbacks
- Smart edit dialog accepts desktop drag-and-drop video files
- Raise media upload client timeout to 10 minutes and surface
  friendly timeout errors instead of raw 'signal timed out'
2026-09-18 17:40:49 +08:00
134
3002704d77 Document video canvas production rollout 2026-09-18 15:24:10 +08:00
134
fe1c9b60f3 Fix video canvas media imports and node layout 2026-09-18 15:09:34 +08:00
134
4f9453b8b2 Add durable remote audio inference and verified offline speech models 2026-09-18 14:12:28 +08:00
134
229f4d72cb Record verified production rollout of canvas groups and video upscaling 2026-09-18 13:37:54 +08:00
134
12576c565f Complete durable canvas groups and add isolated P40 video upscaling 2026-09-18 13:18:48 +08:00
134
b17b8b7689 Make canvas connections resolve live inputs and respect disconnects 2026-09-18 13:18:48 +08:00
134
0d5558c50c Fix handwritten storyboard entry and generic grid presets; record GPU allocation audit 2026-09-18 13:18:48 +08:00
134
c6533e92f3 Complete smart edit audio processing and font catalogue with main app validation 2026-09-18 13:18:48 +08:00
134
17e17b9ad5 Fix smart edit captions, audio gain and timeline parity 2026-09-18 13:18:48 +08:00
134
689fa615e8 Audit smart edit parity and fix caption timing and outline previews 2026-09-18 13:18:47 +08:00
134
725c1ec3d2 Match smart edit transition previews and centered effects to reference 2026-09-18 13:18:47 +08:00
134
84bfd6b9f4 Support local preview transforms and detached smart edit audio tracks 2026-09-18 13:18:47 +08:00
134
6869bd8c18 Align smart edit canvas and timeline UI with reference workflow 2026-09-17 18:05:29 +08:00
134
a5f4ad0385 Bound per-input decoder and scene detection threads 2026-09-17 16:42:06 +08:00
134
6eff0a7332 Use static import in remote rendering helper test 2026-09-17 16:39:51 +08:00
134
a00a910227 Verify remote render recovery identity and endpoint binding 2026-09-17 16:39:51 +08:00