Repository navigation
fix(video-agent-kit): retain videos without scene cuts - #91
Open
Ricardo-M-L wants to merge 1 commit into
Open
Ricardo-M-L wants to merge 1 commit into
Ricardo-M-L wants to merge 1 commit into
Conversation
Collaborator
|
✅ Internal validation passed. A ZCode maintainer will review the change next; after acceptance, release checks run separately and this PR is closed automatically. |
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Plugin type (required) / 插件类型(必填)
Affected plugin name(s) / 涉及的插件名称:
video-agent-kit(Standard plugin)What / 改动内容
detect_shotsreturns zero shots for a video with no scene cuts because PySceneDetect excludes an uncut scene by default. Both normal and frame-skipping detection are affected, and the empty result can be reused from cache. Downstream recap/editing workflows then lose the entire source video.Use
SceneManager.get_scene_list(start_in_scene=True)in both modes. A two-second uncut video now produces one shot, including when its duration is shorter thanmin_len. Using the SceneManager API also preserves compatibility with early SceneDetect 0.6 versions whose conveniencedetect()wrapper lacks the option. Real cut boundaries remain contiguous.Upgrade the cache contract to
cover-all-v3so old empty caches are rebuilt, synchronize the MCP tool description, add plugin-local real-video regressions and equivalent English/Chinese README notes, and bump both manifests plus the marketplace from0.4.3to0.4.4.No dependency, permission, network access, or third-party material is added. The tool continues to write the existing shots timeline, rebuilding caches from the older contract.
This is independent of #89 (concat path escaping) and #38 (Python launcher resolution). If another
video-agent-kitPR publishes first, this branch will need to rebase and use the next available plugin version.Validation
min_len, cut continuity, old-cache invalidation, and new-cache reuse. Early SceneDetect end timestamps are allowed their existing one-frame precision difference.pnpm install --frozen-lockfileandpnpm buildpassed with Node.js 24.20.0.python3 scripts/validate.py: 34 plugins validated.python3 scripts/build_dist.py: 34 artifacts / 292 assets built.git diff --checkpassed.The tool function was exercised with real FFmpeg and SceneDetect. Installation and invocation inside the ZCode desktop host were not performed.
Self-check / 自查清单
categoryis set independently of publisher ownership / 用途分类已填写,与维护归属分开versionbumped in BOTHplugin.jsonandmarketplace.json/ 插件清单与市场版本已同步升级README.md(EN) andREADME_CN.md(中文) both updated and equivalent — this legacy official plugin has only a ChineseREADME.md; the new explanation is provided in both languages in that existing filedescription_i18nhas bothenandzh-CN/ 包含两种语言python3 scripts/validate.pyandpython3 scripts/build_dist.pypass locally / 本地校验与构建均通过