From 915eb4b739f5be22058c4fb66a81d0b3a30d262e Mon Sep 17 00:00:00 2001
From: Linghong <ink29535@proton.me>
Date: Fri, 31 Jul 2026 02:43:05 +0000
Subject: [PATCH] =?UTF-8?q?=E4=BF=AE=E5=A4=8D=E5=9B=BE=E7=89=87=E7=94=BB?=
 =?UTF-8?q?=E5=B8=83=E5=86=85=E9=83=A8=E6=A8=A1=E5=9E=8B=E6=90=9C=E7=B4=A2?=
 =?UTF-8?q?=E6=B3=84=E6=BC=8F?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

素材与图层搜索排除内部处理模型和 Provider。
补充模型层与侧栏回归测试，确认审计元数据不被删除。
同步图片画布搜索契约与项目决策记录。
---
 .../shared-memory/decision-log.md             | 10 ++
 ...架构】图片画布编辑器MVP接入方案-2026-06-11.md |  4 +-
 .../ImageCanvasAssetLibraryModel.test.ts      | 75 +++++++++++++++
 .../ImageCanvasAssetLibraryModel.ts           | 14 ++-
 .../ImageCanvasSidebarView.test.tsx           | 96 +++++++++++++++++++
 5 files changed, 193 insertions(+), 6 deletions(-)

diff --git a/docs/project-memory/shared-memory/decision-log.md b/docs/project-memory/shared-memory/decision-log.md
index e823c05c6..c28d9948e 100644
--- a/docs/project-memory/shared-memory/decision-log.md
+++ b/docs/project-memory/shared-memory/decision-log.md
@@ -26,6 +26,16 @@
 
 ---
 
+## 2026-07-30 图片画布搜索不得索引内部模型和 Provider
+
+- 背景：素材和图层详情虽已把内部处理模型显示为 `-`，搜索仍索引原始 `model` 与 `provider`，导致 `BgFilter complex`、`segment-common-image`、`BgFilter` 或 `Aliyun Matting` 等隐藏信息可被查询命中，并出现“命中但无可见匹配字段”的异常体验。
+- 决策：素材与图层搜索只索引用户可见生成模型；统一复用 `isEditorInternalProcessingModel(...)` 排除内部处理模型，并完全排除 `provider`。该规则只约束前端临时搜索值，不删除或改写素材、图层及后端保存的原始审计字段；`gpt-image-2`、`audio1.0` 等正常模型继续支持搜索。
+- 影响范围：`ImageCanvasAssetLibraryModel.ts` 的素材与图层搜索值、图片画布素材 / 图层侧栏搜索和对应前端架构文档；不改变持久化、详情展示、删除、移动或画布保存行为。
+- 验证方式：模型单测覆盖内部模型和 Provider 不命中、正常模型继续命中且原对象元数据不变；侧栏交互测试覆盖素材与图层两类入口。运行对应 Vitest、`npm run typecheck`、`npm run lint:eslint`、`npm run check:encoding` 和 `git diff --check`。
+- 关联文档：`docs/technical/【前端架构】图片画布编辑器MVP接入方案-2026-06-11.md`。
+
+---
+
 ## 2026-07-29 图集切片必须受前置容量和有界 CPU 保护
 
 - 背景：图标与 UI 图集的 alpha 连通域识别会在 async handler 上同步执行；原始连通域合并采用全量两两比较，`64` 个输出限制又晚于排序、裁剪和 PNG 编码。碎块或噪点图会放大 CPU 与内存成本，手动拆分、图标自动拆分和 UI 提取都受影响。另一方面，图标与 UI 的 Alpha 尺寸恢复、provider 原图回读或透明图解码失败此前只记日志，仍会把不可信透明图持久化并拆分。
diff --git a/docs/technical/【前端架构】图片画布编辑器MVP接入方案-2026-06-11.md b/docs/technical/【前端架构】图片画布编辑器MVP接入方案-2026-06-11.md
index 504cdf427..e6e1161da 100644
--- a/docs/technical/【前端架构】图片画布编辑器MVP接入方案-2026-06-11.md
+++ b/docs/technical/【前端架构】图片画布编辑器MVP接入方案-2026-06-11.md
@@ -164,8 +164,8 @@
 - 音频 / 视频素材卡和画布媒体图层必须提供稳定的非文字视觉预览：优先使用 `thumbnailSrc` / 视频 `poster`，没有真实首帧或音频封面时使用由媒体类型、素材名和地址派生的确定性视觉底图。视频图层使用原生 `<video controls preload="metadata" playsInline>` 播放，外层图层仍承接选择和拖拽语义；音频图层播放前继续通过 `/api/assets/read-url` 换签。画布素材导出按 `mediaType` 保留真实媒体格式：图片进入 `images/`，音频 / 视频进入 `media/`，文件扩展名从响应 MIME、`objectKey` 或源 URL 推断，不得把音频 / 视频导出成 PNG。
 - 画布素材 ZIP 的 `metadata.json` 只保存前端信息弹窗和导出文件列表可见的展示快照：项目标题、导出时间、图层标题、文件路径、类型、生成输入、模型显示名、Task 短 ID、Object 显示值、Resolution / 时长和导出错误。导出的生成输入只保留用户实际填写或选择的内容；系统默认兜底提示词、固定工作流提示词、内置图标描述、UI 提取素材固定提示词等内置提示词即使存在于历史 `generationInputs`，也不得写入导出元数据。不得把 `projectId`、`layerId`、`resourceId`、`sourceAssetId`、`sourceResourceId`、原始 `prompt` / `actualPrompt` / `provider` 或画布坐标、锁定、隐藏等布局状态写入导出元数据；Object 字段仅沿用信息弹窗当前可见值。
 - `生成角色动作` 的完成结果按序列帧素材处理：图层主 `src` 使用 `frames[0].imageSrc`，`mediaType` 固定为 `image-sequence`，`assetKind` 固定为 `character-animation`，完整帧列表写入 `imageSequenceFrames`；`previewVideoPath` 只保留为上游预览视频来源，不作为画布主媒体。下载和 ZIP 导出必须因此得到序列帧 ZIP / frames 目录，不能回退为预览视频或首帧 PNG；后端抽帧后逐帧保存带绿幕源图，上传时消费并释放原帧字节，再用该对象的 600 秒 OSS URL 完成透明化并落盘透明帧素材。
-- 素材面板支持按素材名、文件夹名、生成信息、模型、任务和媒体类型搜索，并支持选择模式框选，一次选中多个素材，并可批量移动或删除上传素材。
-- 图层面板支持按图层名、生成信息、模型、任务和媒体类型搜索；支持选择多个图层后创建图层组，组名和 groupId 随画布布局保存。
+- 素材面板支持按素材名、文件夹名、生成信息、用户可见模型、任务和媒体类型搜索，并支持选择模式框选，一次选中多个素材，并可批量移动或删除上传素材。Provider 与 `isEditorInternalProcessingModel(...)` 识别的内部处理模型不得进入搜索索引，避免出现隐藏字段可命中或命中但无可见匹配字段；停止索引不删除素材上供 owner / 后台审计使用的原始元数据。
+- 图层面板支持按图层名、生成信息、用户可见模型、任务和媒体类型搜索；Provider 与内部处理模型同样不得进入搜索索引。支持选择多个图层后创建图层组，组名和 groupId 随画布布局保存。
 - 小地图支持拖拽视口框，拖动时画布 viewport 跟随移动；`pointermove` 更新必须通过 `requestAnimationFrame` 合帧，结束拖拽时 flush 最后一帧，避免高频 pointermove 直接压垮 React 渲染和项目持久化链路。
 - 鼠标滚轮默认按设备原始 `deltaX / deltaY` 二维平移画布视口；当按住 Shift 且设备上报 `deltaX = 0` 时，视口输入适配层把 `deltaY` 映射为横向位移并将纵向位移置零，核心平移模型仍只消费归一化后的 X / Y。按住 Ctrl / Cmd 滚轮才缩放画布，并阻止浏览器页面缩放。缩放比例显示保持现有换算口径，最低可缩小到 `5%`。
 - 工程刷新后能从后端恢复资源、图层布局和 viewport。
diff --git a/src/components/image-editor/ImageCanvasAssetLibraryModel.test.ts b/src/components/image-editor/ImageCanvasAssetLibraryModel.test.ts
index 6675d690c..46253fc37 100644
--- a/src/components/image-editor/ImageCanvasAssetLibraryModel.test.ts
+++ b/src/components/image-editor/ImageCanvasAssetLibraryModel.test.ts
@@ -9,6 +9,8 @@ import {
   filterGroupedAssetsBySearch,
   getSelectableAssets,
   groupAssetsByFolder,
+  layerMatchesSearch,
+  layerStartsWithSearch,
   moveAssetMarqueeToPointer,
   moveAssetToFolderLocally,
   removeAssetById,
@@ -28,6 +30,7 @@ import {
   toggleAssetSelection,
 } from './ImageCanvasAssetLibraryModel';
 import type {
+  CanvasLayer,
   EditorAsset,
   EditorAssetFolder,
 } from './ImageCanvasEditorTypes';
@@ -60,6 +63,24 @@ function createAsset(overrides: Partial<EditorAsset> = {}): EditorAsset {
   };
 }
 
+function createLayer(overrides: Partial<CanvasLayer> = {}): CanvasLayer {
+  return {
+    id: 'layer-a',
+    resourceId: 'resource-a',
+    title: '图层A',
+    src: 'data:image/png;base64,YQ==',
+    x: 0,
+    y: 0,
+    width: 320,
+    height: 240,
+    originalWidth: 320,
+    originalHeight: 240,
+    zIndex: 1,
+    sourceType: 'generated',
+    ...overrides,
+  };
+}
+
 describe('ImageCanvasAssetLibraryModel', () => {
   it('groups assets by folder and resolves selectable uploaded assets', () => {
     const folders = [
@@ -184,6 +205,60 @@ describe('ImageCanvasAssetLibraryModel', () => {
     ).toEqual([{ id: 'folder-uploaded', assets: ['asset-b', 'asset-c'] }]);
   });
 
+  it('excludes internal processing models and providers from asset and layer search without deleting metadata', () => {
+    const internalAsset = createAsset({
+      id: 'asset-internal',
+      label: '透明角色',
+      sourceType: 'generated',
+      model: 'BgFilter complex',
+      provider: 'BgFilter',
+    });
+    const visibleAsset = createAsset({
+      id: 'asset-visible',
+      label: '生成角色',
+      sourceType: 'generated',
+      model: 'gpt-image-2',
+      provider: 'VectorEngine',
+    });
+    const groupedAssets = groupAssetsByFolder(
+      [createFolder()],
+      [internalAsset, visibleAsset],
+    );
+    const internalLayer = createLayer({
+      model: 'segment-common-image',
+      provider: 'Aliyun Matting',
+    });
+    const visibleLayer = createLayer({
+      id: 'layer-visible',
+      model: 'audio1.0',
+      provider: 'VectorEngine',
+    });
+
+    expect(filterGroupedAssetsBySearch(groupedAssets, 'bgfilter')).toEqual([]);
+    expect(filterGroupedAssetsBySearch(groupedAssets, 'vectorengine')).toEqual(
+      [],
+    );
+    expect(
+      filterGroupedAssetsBySearch(groupedAssets, 'gpt-image-2')[0]?.assets.map(
+        (asset) => asset.id,
+      ),
+    ).toEqual(['asset-visible']);
+    expect(layerMatchesSearch(internalLayer, 'segment-common-image')).toBe(
+      false,
+    );
+    expect(layerStartsWithSearch(internalLayer, 'aliyun')).toBe(false);
+    expect(layerMatchesSearch(visibleLayer, 'audio1.0')).toBe(true);
+    expect(layerMatchesSearch(visibleLayer, 'vectorengine')).toBe(false);
+    expect(internalAsset).toMatchObject({
+      model: 'BgFilter complex',
+      provider: 'BgFilter',
+    });
+    expect(internalLayer).toMatchObject({
+      model: 'segment-common-image',
+      provider: 'Aliyun Matting',
+    });
+  });
+
   it('renames assets, toggles folders and creates local folders', () => {
     expect(renameAssetById([createAsset()], 'asset-a', '新名字')[0]).toMatchObject(
       { label: '新名字' },
diff --git a/src/components/image-editor/ImageCanvasAssetLibraryModel.ts b/src/components/image-editor/ImageCanvasAssetLibraryModel.ts
index 09ba9bd51..c254d1d04 100644
--- a/src/components/image-editor/ImageCanvasAssetLibraryModel.ts
+++ b/src/components/image-editor/ImageCanvasAssetLibraryModel.ts
@@ -1,3 +1,4 @@
+import { isEditorInternalProcessingModel } from './ImageCanvasEditorModel';
 import type {
   AssetMarqueeState,
   CanvasLayer,
@@ -204,13 +205,19 @@ function startsWithSearchText(query: string, values: Array<string | null | undef
   return values.some((value) => value?.trim().toLowerCase().startsWith(query));
 }
 
+function resolveSearchableEditorModel(model: string | null | undefined) {
+  const normalizedModel = model?.trim();
+  return normalizedModel && !isEditorInternalProcessingModel(normalizedModel)
+    ? normalizedModel
+    : undefined;
+}
+
 function buildAssetSearchValues(asset: EditorAsset) {
   return [
     asset.label,
     asset.prompt,
     asset.actualPrompt,
-    asset.model,
-    asset.provider,
+    resolveSearchableEditorModel(asset.model),
     asset.taskId,
     asset.assetKind,
     asset.mediaType,
@@ -222,8 +229,7 @@ function buildLayerSearchValues(layer: CanvasLayer) {
     layer.title,
     layer.prompt,
     layer.actualPrompt,
-    layer.model,
-    layer.provider,
+    resolveSearchableEditorModel(layer.model),
     layer.taskId,
     layer.assetKind,
     layer.mediaType,
diff --git a/src/components/image-editor/ImageCanvasSidebarView.test.tsx b/src/components/image-editor/ImageCanvasSidebarView.test.tsx
index 3062bf96e..0cc8e360f 100644
--- a/src/components/image-editor/ImageCanvasSidebarView.test.tsx
+++ b/src/components/image-editor/ImageCanvasSidebarView.test.tsx
@@ -269,6 +269,52 @@ describe('ImageCanvasSidebarView', () => {
     ).toBeNull();
   });
 
+  it('does not expose internal models or providers through asset search', () => {
+    const folder = createFolder();
+    const internalAsset = createAsset({
+      id: 'asset-internal',
+      label: '透明角色',
+      sourceType: 'generated',
+      model: 'BgFilter complex',
+      provider: 'BgFilter',
+    });
+    const visibleAsset = createAsset({
+      id: 'asset-visible',
+      label: '生成角色',
+      sourceType: 'generated',
+      model: 'gpt-image-2',
+      provider: 'VectorEngine',
+    });
+    render(
+      <ImageCanvasSidebarView
+        {...createSidebarProps({
+          assets: [internalAsset, visibleAsset],
+          groupedAssets: [
+            { ...folder, assets: [internalAsset, visibleAsset] },
+          ],
+        })}
+      />,
+    );
+
+    const sidebar = screen.getByRole('complementary', { name: '图片资源栏' });
+    const search = within(sidebar).getByLabelText('搜索素材');
+
+    fireEvent.change(search, { target: { value: 'BgFilter' } });
+    expect(
+      within(sidebar).queryByRole('button', { name: '添加透明角色' }),
+    ).toBeNull();
+
+    fireEvent.change(search, { target: { value: 'VectorEngine' } });
+    expect(
+      within(sidebar).queryByRole('button', { name: '添加生成角色' }),
+    ).toBeNull();
+
+    fireEvent.change(search, { target: { value: 'gpt-image-2' } });
+    expect(
+      within(sidebar).getByRole('button', { name: '添加生成角色' }),
+    ).toBeTruthy();
+  });
+
   it('renders the layer panel with grouped and locked layer details', () => {
     render(
       <ImageCanvasSidebarView
@@ -346,6 +392,56 @@ describe('ImageCanvasSidebarView', () => {
     ).toBeNull();
   });
 
+  it('does not expose internal models or providers through layer search', () => {
+    render(
+      <ImageCanvasSidebarView
+        {...createSidebarProps({
+          activeSidebarPanel: 'layers',
+          layers: [
+            createLayer({
+              id: 'layer-internal',
+              title: '透明角色图层',
+              model: 'segment-common-image',
+              provider: 'Aliyun Matting',
+              zIndex: 2,
+            }),
+            createLayer({
+              id: 'layer-visible',
+              title: '生成角色图层',
+              model: 'gpt-image-2',
+              provider: 'VectorEngine',
+              zIndex: 1,
+            }),
+          ],
+        })}
+      />,
+    );
+
+    const sidebar = screen.getByRole('complementary', { name: '图片资源栏' });
+    const search = within(sidebar).getByLabelText('搜索画布图层');
+
+    fireEvent.change(search, { target: { value: 'segment-common-image' } });
+    expect(
+      within(sidebar).queryByRole('button', {
+        name: '选择图层透明角色图层',
+      }),
+    ).toBeNull();
+
+    fireEvent.change(search, { target: { value: 'VectorEngine' } });
+    expect(
+      within(sidebar).queryByRole('button', {
+        name: '选择图层生成角色图层',
+      }),
+    ).toBeNull();
+
+    fireEvent.change(search, { target: { value: 'gpt-image-2' } });
+    expect(
+      within(sidebar).getByRole('button', {
+        name: '选择图层生成角色图层',
+      }),
+    ).toBeTruthy();
+  });
+
   it('keeps sidebar tab switching and repeated-click closing behavior', () => {
     render(<SidebarTabsHarness />);