feat: 同步视觉风格与主体身份母版工作区

对齐后端 dev 059e597，新增项目风格编辑锁定、身份文本生成、身份参考图与母版切换。
整理六个工作区入口，接通人物及场景母版继承说明、形态图片来源追溯。
补充草稿保护、正式 ID 校验、费用确认及部分失败回执，67 项测试和静态检查、构建通过。
未修改后端，未调用真实模型；本环境未完成浏览器视觉验收。

README.md

@@ -1,6 +1,6 @@
 # Short Drama Agent Front
 
-短剧 Agent 的前端工作台。支持 `create-drama`、`breakdown` 与 `storyboard` 工作区，使用真实后端 API，不包含演示数据或浏览器端模型调用。
+短剧 Agent 的前端工作台。支持剧本创作、拆解、视觉风格、主体身份、形态图片与分镜设计，使用真实后端 API，不包含演示数据或浏览器端模型调用。
 
 ## 启动
 
@@ -31,6 +31,9 @@ API_PROXY_TARGET=http://localhost:3412
 | 拆解配置 | 每组集数、人物／场景／道具模块选择、后端分组预览 |
 | 拆解结果 | 主体、别名、视觉形态、Episode → Beat → Shot、主体绑定、抽取任务状态、校验问题、JSON 导出 |
 | 拆解恢复 | 失败抽取重试、缺失剧集镜头补齐、分镜引用与 Form 绑定修复 |
+| 视觉风格 | AI 生成、人工编辑、分类 Prompt、JSON 硬约束、锁定；风格参考图地址登记、启停和删除 |
+| 主体身份 | 按正式主体归并形态，单个／批量生成身份文本，人工编辑及锁定，部分失败诊断 |
+| 身份母版 | Seedream 生图、四种参考视角、指定身份参考图、母版候选与切换、实际 Prompt 和来源追溯 |
 | 分镜检查 | 按剧集、Beat、正式 Shot ID 浏览；导演设计、环境与主体即时状态；单集覆盖率与 JSON 导出 |
 | 分镜生成 | Direction / VisualState 单集生成、项目批量补齐、覆盖重生成；并发和自动修复次数；部分失败诊断 |
 | 镜头工具 | 主体参考图与缺失原因、确定性 GenerationSpec 查询与导出、单镜头／批量视频提示词 |
@@ -67,6 +70,8 @@ Oxfmt 不负责代码质量，Oxlint 不负责 Vue 的完整类型推导；Vue S
 | `src/features/breakdown` | 拆解 graph 的 API、类型、配置页面、主体与分镜组件 |
 | `src/features/storyboard` | 分镜 API、类型、生成操作、依赖判断、设计与状态、镜头资源、生成回执 |
 | `src/features/subject-images` | 正式形态图库、图片生成参数、历史图片、主图选择与批量回执 |
+| `src/features/visual-style` | 项目视觉风格 API、类型、编辑器、锁定和风格参考图管理 |
+| `src/features/subject-identity` | 稳定身份 API、正式主体目录、文本编辑、生图配置、身份图库和母版切换 |
 | `src/features/workflows` | checkpoint 选择、执行记录、长请求互斥与操作确认 |
 | `src/components/ui` | 跨业务使用的 Reka UI 封装与公共展示组件 |
 | `src/composables` | 具备取消和竞态保护的轮询 |
@@ -78,7 +83,7 @@ Oxfmt 不负责代码质量，Oxlint 不负责 Vue 的完整类型推导；Vue S
 ## 接口基线
 
 对齐后端 `qianlanse/short-drama-agent` 的 `dev`：
-`eac8dc3b1b2ce1ff2bdfeb01d7281cd838f07285`，包含本次新增的只读 `GET /projects/:projectId/subject-forms` 接口。请同步更新前后端 `dev` 后重启后端。
+`059e59709359dccd08b08cb35c6aff95c5e03f2b`（2026-08-28），包含 VisualStyle、SubjectIdentity、身份母版、人物／场景形态自动引用母版及道具稳定特征提取。本轮只修改前端，请确保后端依照其开发流程同步了新 Prisma 结构并重启服务。
 
 | HTTP | 路径（以 /api 为前缀） | 用途 |
 | --- | --- | --- |
@@ -102,11 +107,21 @@ Oxfmt 不负责代码质量，Oxlint 不负责 Vue 的完整类型推导；Vue S
 | GET | `/storyboard-shots/:shotId/generation-spec` | 后端确定性编译结果，不调用模型 |
 | POST | `/storyboard-shots/:shotId/video-prompt` | 复用已有 Prompt 或调用模型生成；`force` 控制覆盖 |
 | POST | `/projects/:id/video-prompts/generate` | 项目批量 Prompt；`concurrency`、`force` |
-| GET | `/projects/:id/subject-forms` | 正式 SubjectForm ID、主体名称／引用、主图与图片元数据；本次新增只读接口 |
+| GET | `/projects/:id/subject-forms` | 正式 SubjectForm ID、主体名称／引用、主图与图片元数据；只读查询 |
 | GET | `/subject-forms/:formId/images` | 此形态的全部图片历史及实际 Prompt |
 | POST | `/subject-forms/:formId/images` | 单图生成；显式 `provider: seedream`，可选 Prompt 和成对尺寸，`setPrimary` |
 | PUT | `/subject-forms/:formId/images/:imageId/primary` | 选择已完成图片作为主参考图，不调用模型 |
 | POST | `/projects/:id/subject-images/generate` | 全项目批量生图；`provider`、`concurrency`、`force` |
+| GET / PUT | `/projects/:id/visual-style` | 查询／人工保存视觉风格，包含分类 Prompt、硬约束和锁定 |
+| POST | `/projects/:id/visual-style/generate` | AI 生成风格；已有风格重生成显式 `force: true`，锁定时禁止 |
+| POST | `/projects/:id/visual-style/images` | 登记已有风格图片地址，不上传文件 |
+| PUT | `/projects/:id/visual-style/images/:imageId/enabled` | 风格图片启停 |
+| DELETE | `/projects/:id/visual-style/images/:imageId` | 删除风格图片记录，不删除远程文件 |
+| GET / PUT | `/subjects/:subjectId/identity` | 查询／人工编辑稳定身份；使用正式 Subject ID |
+| POST | `/subjects/:subjectId/identity/generate` | AI 生成身份文本，需要项目视觉风格 |
+| POST | `/projects/:id/subject-identities/generate` | 项目批量身份文本；`force`、`concurrency`，已锁定始终跳过 |
+| GET / POST | `/subjects/:subjectId/identity/images` | 查询带 `isAnchor` 的身份图片／Seedream 生图 |
+| PUT | `/subjects/:subjectId/identity/images/:imageId/anchor` | 将成功的 primary 候选设为身份母版 |
 
 API 层也提供 `state` 与 `breakdown/latest` 方法。页面通过共享 checkpoint 查询取得阶段状态，避免重复拉取同样内容。
 
@@ -146,6 +161,28 @@ API 层也提供 `state` 与 `breakdown/latest` 方法。页面通过共享 chec
 8. 图片使用持久化后的 `/storage/...` 地址，开发代理已配置。生产环境也必须代理 `/storage/`，不能让 SPA 回退返回 index.html。无效 URL、加载失败、无图和后端生成失败有各自提示；不会自动使用可能过期的 Provider 临时地址。
 9. 设为主图后，回到分镜点击“刷新参考图”读取最新绑定。已有视频提示词不会自动重生成，需要按需覆盖。本页面不会调用 `/production/start` 或实际视频生成。
 
+### 视觉风格、身份与形态的衔接
+
+建议操作顺序：**剧本创作 → 剧本拆解 → 视觉风格 → 主体身份 → 形态图片 → 分镜设计**。视觉风格也可以在拆解前准备；这不是前端强制启动的整条生产流程。
+
+| 层级 | 职责 | 本次可操作内容 |
+| --- | --- | --- |
+| VisualStyle | 项目整体美术与视觉语言 | 整体／人物／场景／道具 Prompt、JSON 字符串数组硬约束、锁定、风格图片记录 |
+| SubjectIdentity | 同一主体跨形态不变的特征 | 稳定身份描述、身份 Prompt、锁定、单个及批量 AI 文本生成 |
+| Identity Image | 固定主体身份，人物主要固定脸部与体型 | primary、front、three-quarter、full-body 四类生图、身份锚定来源、母版选择 |
+| SubjectForm Image | 某个具体造型或状态的图片 | 原有单图／批量生图与主图选择；人物／场景图自动引用当前身份母版 |
+
+1. 风格与身份的 `isLocked` 只禁止 AI 自动覆盖文本。已锁定记录仍可人工编辑并保存；身份生图和母版切换也不受文字锁定限制。需要 AI 重生成时，先取消锁定并保存。
+2. 身份 AI 生成需要已创建 VisualStyle；身份图生成还需要已保存的非空 `generationPrompt`。不存在身份时不调用会报错的图片 GET。404 或网络失败不会被当作“尚未创建”。
+3. 后端暂无全项目主体／身份的 GET。前端从正式 `subject-forms` 目录按 `subjectId` 去重，选中主体后按需读取身份和专用图库；无形态主体暂不在目录展示。批量仍由后端处理项目全部主体，可能多于目录可见数，不受当前筛选影响。未使用 checkpoint 或 `DramaCharacter.id` 拼造 Subject ID。
+4. 身份文本批量回执显示 `total`、`targetCount`、`generated`、`skipped`、`skippedLocked`、`failed` 和逐主体失败原因。它不代表图片已生成，也不是实时任务进度。后端没有批量身份生图接口，前端不自行循环触发有费用的图片请求。
+5. 身份图片的 `isAnchor` 来自专用 GET。`enabled: true` 的正面、侧面或全身图不是母版；只有成功且有地址的 `primary` 图可经 PUT 设为母版。第一张 primary 图自动启用；已有母版后，新 primary 图作为停用候选保留，不静默替换。辅助视图不能直接选为母版，但可以作为新 primary 图的参考来源。
+6. 身份生图不传 `referenceImageId` 时，由后端自动选择当前母版；无母版时按文本生成。指定时只能选择当前身份已完成的图片；候选图即使停用也可作为显式参考。后端没有“有母版但忽略它”的参数。尺寸成对留空或提供正整数。自定义 Prompt 会替换后端编译的完整文本，建议保留默认。
+7. 目前**人物与场景形态生图**会自动注入当前身份母版，分别固定人物身份和空间骨架；道具暂不自动引用。没有身份或母版时，原形态生图流程仍可使用。已有母版但缺少 Provider 可访问远程地址时，后端会报错；本地 `/storage/` 能显示并不证明原始 Provider URL 仍可访问。前端显示后端错误，不自动重试消耗额度。
+8. 修改风格、身份文本或母版，不会自动使旧图失效或重生成。已有形态主图仍会被分镜使用；需要保持新身份时，主动生成候选图并设为形态主图，再刷新分镜参考图，按需覆盖视频提示词。图库详情展示当次 `identityAnchorImageId` 或 `referenceImageId`；旧记录没有该字段时明确显示未知，不推断它使用了当前母版。
+9. 风格参考图支持登记已有 HTTP(S)／`/storage/` 地址、分类、排序、启停和删除记录；**没有文件上传接口，也没有风格图 AI 生成接口**。当前身份图／形态图 Provider 调用不会自动消费风格参考图片，页面不暗示登记后已生效。
+10. 编辑草稿与轮询分离。保存成功才清空草稿，失败保留；主体切换会提示放弃未保存修改。离开页面不会自动保存。所有新写操作沿用项目级会话互斥，生成前需要费用与后台任务确认，无法替代后端跨浏览器锁。
+
 ### 后端限制
 
 目前 `checkpoints` 接口返回完整历史与 state，没有分页或轻量状态接口。长篇剧本会增加轮询流量。下一步建议后端增加：
@@ -203,5 +240,11 @@ server {
 8. 给主体形态准备参考图，读取镜头参考图和 GenerationSpec；确认 Prompt 无正文时经确认生成、有正文且不覆盖时复用。不提交实际视频任务。
 9. 打开“形态图片”，检查已有主图自动显示；单个形态新增候选图，查看历史、失败原因和实际 Prompt，再设为主图。
 10. 在小型测试项目中验证批量补齐与“已有主图也新增候选图”；回到分镜刷新参考图，确认使用新选择的主图。
+11. 进入“视觉风格”，保存或生成风格，检查分类 Prompt 和硬约束；锁定后 AI 入口禁用，人工编辑仍可保存。登记一张已有风格图，检查启停和移除确认。
+12. 进入“主体身份”，检查同一主体的多形态归并。生成／保存身份文本并锁定，再生成第一张 primary 图；生成第二张后，确认旧母版未被替换。
+13. 为同一主体选择参考来源并生成 front / three-quarter / full-body 图；确认辅助图不能直接设为母版。将第二张 primary 候选设为母版后，检查旧母版停用且保留。
+14. 对人物和场景形态分别新增图片，在图片详情检查当次母版 ID；旧主图应保持不变，手动切换后再刷新分镜参考图。道具不应显示自动继承母版已生效。
+
+本轮新增测试覆盖风格 JSON／锁定／图片 DELETE、身份归并／前置条件／尺寸／显式参考图／母版切换、部分失败、旧响应、草稿保护与费用确认。模型调用均使用模拟 API。当前执行环境未能通过浏览器访问本地开发服务，**未完成浏览器视觉验收及真实模型联调**，请按上述步骤在本地验收。
 
 自动测试使用模拟 API，覆盖既有流程、分镜接口与竞态，以及形态图片自动读取、正式 ID 生图、主图 PUT、费用确认、尺寸校验、候选图选择、批量部分失败和图片加载错误。不等同于真实模型、MySQL 或浏览器视觉联调。请勿用正式项目随意制造失败以测试恢复。

src/App.vue

@@ -5,6 +5,8 @@ import {
     Clapperboard,
     Camera,
     Images,
+    Palette,
+    Fingerprint,
     FolderOpen,
     FileText,
     Layers,
@@ -56,20 +58,34 @@ const apiBase = import.meta.env.VITE_API_BASE_URL || '/api'
                         title="剧本拆解"
                         ><Layers :size="18" /><span class="sidebar-label">剧本拆解</span
                         ><span class="sidebar-label ml-auto text-[10px] text-stone-500">02</span></RouterLink
+                    ><RouterLink
+                        :to="`/projects/${projectId}/visual-style`"
+                        class="side-link"
+                        active-class="selected"
+                        title="视觉风格"
+                        ><Palette :size="18" /><span class="sidebar-label">视觉风格</span
+                        ><span class="sidebar-label ml-auto text-[10px] text-stone-500">03</span></RouterLink
+                    ><RouterLink
+                        :to="`/projects/${projectId}/subject-identity`"
+                        class="side-link"
+                        active-class="selected"
+                        title="主体身份"
+                        ><Fingerprint :size="18" /><span class="sidebar-label">主体身份</span
+                        ><span class="sidebar-label ml-auto text-[10px] text-stone-500">04</span></RouterLink
                     ><RouterLink
                         :to="`/projects/${projectId}/subject-images`"
                         class="side-link"
                         active-class="selected"
                         title="形态图片"
                         ><Images :size="18" /><span class="sidebar-label">形态图片</span
-                        ><span class="sidebar-label ml-auto text-[10px] text-stone-500">03</span></RouterLink
+                        ><span class="sidebar-label ml-auto text-[10px] text-stone-500">05</span></RouterLink
                     ><RouterLink
                         :to="`/projects/${projectId}/storyboard`"
                         class="side-link"
                         active-class="selected"
                         title="分镜设计"
                         ><Camera :size="18" /><span class="sidebar-label">分镜设计</span
-                        ><span class="sidebar-label ml-auto text-[10px] text-stone-500">04</span></RouterLink
+                        ><span class="sidebar-label ml-auto text-[10px] text-stone-500">06</span></RouterLink
                     >
                 </nav></template
             >

src/features/breakdown/components/SubjectList.vue

@@ -34,6 +34,12 @@ const filtered = computed(() =>
                     >
                 </div>
                 <p class="text-sm leading-7">{{ subject.description }}</p>
+                <RouterLink
+                    v-if="projectId"
+                    :to="{ path: `/projects/${projectId}/subject-identity`, query: { subjectRef: subject.ref } }"
+                    class="text-button mt-3 mr-4"
+                    >管理稳定身份／母版</RouterLink
+                >
                 <RouterLink
                     v-if="projectId"
                     :to="{ path: `/projects/${projectId}/subject-images`, query: { subjectRef: subject.ref } }"

src/features/projects/ProjectLayout.vue

@@ -1,7 +1,7 @@
 <script setup lang="ts">
 import { computed, provide } from 'vue'
 import { useRoute } from 'vue-router'
-import { ArrowLeft, RefreshCw, FileText, Layers, Camera, Images, LoaderCircle } from '@lucide/vue'
+import { ArrowLeft, RefreshCw, FileText, Layers, Camera, Images, Palette, Fingerprint, LoaderCircle } from '@lucide/vue'
 import { StatusBadge } from '../../components/ui'
 import { projectContextKey, useProjectData } from './context'
 import { getOperation } from '../workflows/operations'
@@ -39,18 +39,12 @@ const operation = computed(() => getOperation(id.value))
             </div>
         </div>
         <nav class="workflow-nav" aria-label="项目工作流">
-            <RouterLink :to="`/projects/${id}/create-drama`"
+            <RouterLink :to="`/projects/${id}/create-drama`"><FileText :size="17" />剧本创作</RouterLink>
-                ><FileText :size="17" />剧本创作<span class="nav-code">create-drama</span></RouterLink
+            <RouterLink :to="`/projects/${id}/breakdown`"><Layers :size="17" />剧本拆解</RouterLink>
-            >
+            <RouterLink :to="`/projects/${id}/visual-style`"><Palette :size="17" />视觉风格</RouterLink>
-            <RouterLink :to="`/projects/${id}/breakdown`"
+            <RouterLink :to="`/projects/${id}/subject-identity`"><Fingerprint :size="17" />主体身份</RouterLink>
-                ><Layers :size="17" />剧本拆解<span class="nav-code">breakdown</span></RouterLink
+            <RouterLink :to="`/projects/${id}/subject-images`"><Images :size="17" />形态图片</RouterLink>
-            >
+            <RouterLink :to="`/projects/${id}/storyboard`"><Camera :size="17" />分镜设计</RouterLink>
-            <RouterLink :to="`/projects/${id}/subject-images`"
-                ><Images :size="17" />形态图片<span class="nav-code">subject-image</span></RouterLink
-            >
-            <RouterLink :to="`/projects/${id}/storyboard`"
-                ><Camera :size="17" />分镜设计<span class="nav-code">storyboard</span></RouterLink
-            >
         </nav>
         <p v-if="context.error.value" class="alert alert-error mt-4" role="alert">
             {{ context.error.value }}<span v-if="context.project.value"> 当前保留上次成功读取的数据。</span>

src/features/subject-identity/SubjectIdentityPage.vue

@@ -0,0 +1,317 @@
+<script setup lang="ts">
+import { computed, ref, watch } from 'vue'
+import { useRoute } from 'vue-router'
+import { AssetImage, EmptyState } from '../../components/ui'
+import { downloadText } from '../../lib/format'
+import ConfirmAction from '../workflows/ConfirmAction.vue'
+import { currentAnchor } from './model'
+import { useSubjectIdentity } from './useSubjectIdentity'
+import IdentityEditor from './components/IdentityEditor.vue'
+import IdentityGallery from './components/IdentityGallery.vue'
+import IdentityImageDialog from './components/IdentityImageDialog.vue'
+
+/** 主体身份工作区按正式主体聚合，不把同一主体的多个 Form 当成不同身份。 */
+const route = useRoute()
+const {
+    projectId,
+    operation,
+    selectedId,
+    subject,
+    subjects,
+    catalog,
+    styleQuery,
+    detail,
+    identity,
+    images,
+    session,
+    concurrency,
+    force,
+    editorRevision,
+    dirty,
+    blocked,
+    detailBlocked,
+    hasStyle,
+    canGenerateText,
+    canGenerateImage,
+    batchValid,
+    save,
+    generate,
+    generateProject,
+    generateImage,
+    setAnchor
+} = useSubjectIdentity()
+const search = ref('')
+const module = ref('all')
+const imageOpen = ref(false)
+const pendingSelection = ref('')
+const anchor = computed(() => currentAnchor(images.value))
+const filtered = computed(() =>
+    subjects.value.filter(
+        item =>
+            (module.value === 'all' || item.module === module.value) &&
+            `${item.name} ${item.ref}`.toLowerCase().includes(search.value.trim().toLowerCase())
+    )
+)
+
+/** 深链接仅在目标 ID 变化时选择，不因目录轮询把用户切回原主体。 */
+const linkedSubjectId = computed(
+    () => subjects.value.find(item => item.id === route.query.subjectId || item.ref === route.query.subjectRef)?.id
+)
+watch(
+    linkedSubjectId,
+    id => {
+        if (id) select(id)
+    },
+    { immediate: true }
+)
+watch(selectedId, () => {
+    imageOpen.value = false
+    pendingSelection.value = ''
+})
+
+/** 切换主体前确认放弃当前草稿，避免误丢尚未保存的文本。 */
+function select(id: string) {
+    if (id === selectedId.value) return
+    if (dirty.value) pendingSelection.value = id
+    else selectedId.value = id
+}
+
+/** 用户明确放弃草稿后切换，查询层会取消旧请求。 */
+function discardAndSelect() {
+    if (!pendingSelection.value) return
+    dirty.value = false
+    selectedId.value = pendingSelection.value
+}
+
+/** 保存真实回执便于定位部分失败主体。 */
+function exportReceipt() {
+    downloadText('subject-identities-receipt.json', JSON.stringify(session.value.receipt, null, 2), 'application/json')
+}
+</script>
+
+<template>
+    <section class="mt-7">
+        <div class="flex flex-wrap items-end justify-between gap-4">
+            <div>
+                <h2 class="text-lg font-semibold">主体身份</h2>
+                <p class="mt-2 text-sm text-muted">为同一个人物、场景或道具固定稳定特征，再生成不同造型。</p>
+            </div>
+            <RouterLink :to="`/projects/${projectId}/subject-images`" class="text-button"
+                >下一步：形态图片 →</RouterLink
+            >
+        </div>
+        <p class="alert mt-5 text-xs">
+            人物与场景形态图在有身份母版时自动引用，分别保持人物身份与空间结构；没有母版仍可按原逻辑生图。道具暂不自动引用。旧图不会因母版切换而自动更新。
+        </p>
+        <p v-if="catalog.error.value || styleQuery.error.value" class="alert alert-error mt-4" role="alert">
+            {{ catalog.error.value || styleQuery.error.value }} 当前操作已暂停，请刷新核对后端。
+        </p>
+        <p v-if="styleQuery.data.value && !hasStyle" class="alert mt-4">
+            先创建项目视觉风格，才能 AI 生成身份或身份图片。<RouterLink
+                :to="`/projects/${projectId}/visual-style`"
+                class="text-button ml-2"
+                >设置视觉风格 →</RouterLink
+            >
+        </p>
+        <details class="panel my-5 p-5">
+            <summary class="cursor-pointer text-sm font-medium">项目批量生成身份文本</summary>
+            <div class="mt-4 flex flex-wrap items-end gap-4">
+                <label class="w-28"
+                    ><span class="field-label">并发数</span
+                    ><input
+                        v-model.number="concurrency"
+                        type="number"
+                        min="1"
+                        step="1"
+                        class="input"
+                        :disabled="operation.pending"
+                        aria-label="身份批量并发"
+                /></label>
+                <label class="flex gap-2 pb-2 text-xs"
+                    ><input v-model="force" type="checkbox" :disabled="operation.pending" />覆盖未锁定的已有身份</label
+                >
+                <ConfirmAction
+                    :label="force ? '重新生成未锁定身份' : '补齐项目身份文本'"
+                    :disabled="blocked || !hasStyle || !batchValid || dirty || !subjects.length"
+                    acknowledgement
+                    description="只生成身份描述与提示词，不生成图片。操作面向整个项目，不受筛选影响；已锁定身份始终跳过。覆盖文本不会自动更新旧身份图与形态图。"
+                    @confirm="generateProject"
+                />
+            </div>
+            <p v-if="!batchValid" class="mt-2 text-xs text-danger">并发必须是正整数。</p>
+            <p class="mt-3 text-xs text-muted">
+                后端没有全项目身份查询接口；左侧从正式形态目录归并主体，点击后读取其身份。无形态主体暂不在目录展示，但后端批量会处理项目全部主体。
+            </p>
+        </details>
+        <section v-if="session.receipt" class="panel mb-5 p-5" aria-label="身份文本批量回执">
+            <div class="flex justify-between gap-3">
+                <h3 class="text-sm font-medium">身份文本批量回执</h3>
+                <button class="text-button" @click="exportReceipt">导出诊断</button>
+            </div>
+            <p class="mt-3 text-xs text-muted">
+                共 {{ session.receipt.total }} 个主体 · 目标 {{ session.receipt.targetCount }} · 生成
+                {{ session.receipt.generated }} · 跳过 {{ session.receipt.skipped }}（含锁定
+                {{ session.receipt.skippedLocked }}）· 失败 {{ session.receipt.failed }}
+            </p>
+            <p v-if="session.receipt.failed" class="alert alert-error mt-3" role="alert">
+                部分主体生成失败，已成功结果保留。请检查失败原因，再补齐或单独生成。
+            </p>
+            <ul class="mt-3 space-y-2 text-xs text-danger">
+                <li v-for="failure in session.receipt.failures" :key="failure.subjectId">
+                    {{ failure.subjectRef }} · {{ failure.subjectId }}：{{ failure.error }}
+                </li>
+            </ul>
+            <p class="mt-3 text-[11px] text-muted">回执仅保留于当前浏览器会话，不代表图片已生成。</p>
+        </section>
+        <div v-if="subjects.length" class="identity-workspace">
+            <aside class="panel min-w-0 p-4">
+                <h3 class="text-sm font-medium">
+                    主体目录 <span class="ml-2 text-xs text-muted">{{ subjects.length }}</span>
+                </h3>
+                <input v-model="search" class="input mt-4" placeholder="搜索主体或引用" aria-label="搜索主体身份" />
+                <select v-model="module" class="input mt-3" aria-label="身份主体类型">
+                    <option value="all">全部类型</option>
+                    <option value="character">人物</option>
+                    <option value="scene">场景</option>
+                    <option value="prop">道具</option>
+                </select>
+                <div class="identity-subject-list mt-4">
+                    <button
+                        v-for="item in filtered"
+                        :key="item.id"
+                        class="identity-subject-item"
+                        :class="{ selected: selectedId === item.id }"
+                        :aria-pressed="selectedId === item.id"
+                        @click="select(item.id)"
+                    >
+                        <span class="block font-medium">{{ item.name }}</span
+                        ><span class="mt-1 block text-[11px] text-muted"
+                            >{{ item.ref }} · {{ item.forms.length }} 个形态</span
+                        >
+                    </button>
+                    <p v-if="!filtered.length" class="py-6 text-xs text-muted">没有匹配主体，请调整筛选。</p>
+                </div>
+                <button class="text-button mt-4" :disabled="catalog.loading.value" @click="catalog.refresh">
+                    刷新主体目录
+                </button>
+            </aside>
+            <div class="min-w-0">
+                <div v-if="pendingSelection" class="alert mb-4">
+                    <p class="text-xs">当前主体有未保存修改，切换会丢弃草稿。</p>
+                    <div class="mt-3 flex gap-3">
+                        <button class="button button-secondary" @click="discardAndSelect">放弃草稿并切换</button
+                        ><button class="text-button" @click="pendingSelection = ''">继续编辑</button>
+                    </div>
+                </div>
+                <section v-if="subject" class="panel p-5 sm:p-6">
+                    <div class="flex flex-wrap items-start justify-between gap-3">
+                        <div>
+                            <h3 class="font-semibold">
+                                {{ subject.name }} <code class="subject-ref ml-2">{{ subject.ref }}</code>
+                            </h3>
+                            <p class="mt-2 text-xs text-muted">
+                                {{
+                                    identity
+                                        ? identity.isLocked
+                                            ? '身份文本已锁定'
+                                            : '身份文本未锁定'
+                                        : '尚未创建身份'
+                                }}
+                            </p>
+                        </div>
+                        <button class="text-button" :disabled="detail.loading.value" @click="detail.refresh">
+                            刷新身份
+                        </button>
+                    </div>
+                    <p v-if="detail.error.value" class="alert alert-error mt-4" role="alert">
+                        {{ detail.error.value }} 查询失败时不允许生成或覆盖。
+                    </p>
+                    <p v-if="!detail.data.value && detail.loading.value" class="py-8 text-sm text-muted" role="status">
+                        正在读取主体身份…
+                    </p>
+                    <template v-if="detail.data.value">
+                        <div class="mt-4 flex flex-wrap gap-3">
+                            <ConfirmAction
+                                :label="identity ? 'AI 重新生成身份' : 'AI 生成身份'"
+                                :disabled="!canGenerateText"
+                                acknowledgement
+                                description="调用文本模型生成稳定身份描述和提示词。旧图片不会自动重生成；已锁定身份需先解锁并保存。"
+                                @confirm="generate"
+                            /><button
+                                class="button button-secondary"
+                                :disabled="!canGenerateImage"
+                                @click="imageOpen = true"
+                            >
+                                生成身份参考图
+                            </button>
+                        </div>
+                        <p v-if="dirty" class="mt-3 text-xs text-muted">请先保存身份修改，再使用 AI 或生成图片。</p>
+                        <IdentityEditor
+                            :key="`${selectedId}:${editorRevision}`"
+                            :identity="identity"
+                            :module="subject.module"
+                            :disabled="detailBlocked"
+                            @save="save"
+                            @dirty="dirty = $event"
+                        />
+                        <div v-if="anchor" class="mt-5 flex items-center gap-4 border-t border-line pt-5">
+                            <AssetImage
+                                :src="anchor.imageUrl"
+                                :alt="`${subject.name}当前身份母版`"
+                                class="h-20 w-20 shrink-0 rounded"
+                            />
+                            <div>
+                                <p class="text-sm font-medium">当前身份母版</p>
+                                <p class="mt-2 text-xs text-muted">
+                                    {{
+                                        subject.module === 'prop'
+                                            ? '道具形态生图暂不自动引用此母版。'
+                                            : '后续人物／场景形态图会继承此图的身份或空间结构。'
+                                    }}
+                                </p>
+                            </div>
+                        </div>
+                        <IdentityGallery
+                            :key="selectedId"
+                            :images="images"
+                            :subject-name="subject.name"
+                            :disabled="detailBlocked"
+                            @anchor="setAnchor"
+                        />
+                    </template>
+                    <div class="mt-5 border-t border-line pt-5">
+                        <h4 class="text-xs font-medium">此主体的形态</h4>
+                        <div class="mt-3 flex flex-wrap gap-2">
+                            <span v-for="form in subject.forms" :key="form.id" class="tag"
+                                >{{ form.name }}{{ form.isDefault ? ' · 默认' : '' }}</span
+                            >
+                        </div>
+                        <RouterLink
+                            :to="{ path: `/projects/${projectId}/subject-images`, query: { subjectRef: subject.ref } }"
+                            class="text-button mt-3"
+                            >查看此主体形态图片 →</RouterLink
+                        >
+                    </div>
+                </section>
+            </div>
+        </div>
+        <p v-else-if="catalog.loading.value" class="py-8 text-sm text-muted" role="status">正在读取正式主体目录…</p>
+        <EmptyState
+            v-else-if="catalog.data.value"
+            title="还没有可展示的正式主体"
+            description="先完成剧本拆解与主体形态持久化，再来管理稳定身份。"
+            ><RouterLink :to="`/projects/${projectId}/breakdown`" class="button button-secondary"
+                >前往剧本拆解</RouterLink
+            ></EmptyState
+        >
+        <IdentityImageDialog
+            v-if="subject"
+            v-model:open="imageOpen"
+            :subject-id="subject.id"
+            :subject-name="subject.name"
+            :images="images"
+            :disabled="!canGenerateImage"
+            @generate="generateImage"
+        />
+    </section>
+</template>

src/features/subject-identity/api.ts

@@ -0,0 +1,41 @@
+import { request } from '../../lib/http'
+import type {
+    GenerateIdentityImageInput,
+    IdentityBatchResult,
+    IdentityImage,
+    SaveIdentityInput,
+    SubjectIdentity
+} from './types'
+
+/** 身份接口使用正式 Subject ID，与形态 ID、角色表 ID 分开。 */
+function identityPath(subjectId: string) {
+    return `/subjects/${encodeURIComponent(subjectId)}/identity`
+}
+
+/** 身份文本和参考图分开操作；有费用的请求均不自动重试。 */
+export const subjectIdentityApi = {
+    get: (subjectId: string, signal?: AbortSignal) =>
+        request<SubjectIdentity | null>(identityPath(subjectId), { signal }),
+    save: (subjectId: string, input: SaveIdentityInput) =>
+        request<SubjectIdentity>(identityPath(subjectId), { method: 'PUT', body: input }),
+    generate: (subjectId: string, force: boolean) =>
+        request<SubjectIdentity>(`${identityPath(subjectId)}/generate`, {
+            method: 'POST',
+            body: { force },
+            timeoutMs: 0
+        }),
+    generateProject: (projectId: string, input: { force: boolean; concurrency: number }) =>
+        request<IdentityBatchResult>(`/projects/${encodeURIComponent(projectId)}/subject-identities/generate`, {
+            method: 'POST',
+            body: input,
+            timeoutMs: 0
+        }),
+    listImages: (subjectId: string, signal?: AbortSignal) =>
+        request<IdentityImage[]>(`${identityPath(subjectId)}/images`, { signal }),
+    generateImage: (subjectId: string, input: GenerateIdentityImageInput) =>
+        request<IdentityImage>(`${identityPath(subjectId)}/images`, { method: 'POST', body: input, timeoutMs: 0 }),
+    setAnchor: (subjectId: string, imageId: string) =>
+        request<IdentityImage>(`${identityPath(subjectId)}/images/${encodeURIComponent(imageId)}/anchor`, {
+            method: 'PUT'
+        })
+}

src/features/subject-identity/components/IdentityEditor.vue

@@ -0,0 +1,79 @@
+<script setup lang="ts">
+import { computed, reactive, ref, watch } from 'vue'
+import type { SaveIdentityInput, SubjectIdentity } from '../types'
+
+/** 身份文本草稿不随轮询覆盖；父页面切换主体或保存成功后重建编辑器。 */
+const props = defineProps<{ identity: SubjectIdentity | null; disabled: boolean; module: string }>()
+const emit = defineEmits<{ save: [input: SaveIdentityInput]; dirty: [value: boolean] }>()
+const form = reactive({ description: '', generationPrompt: '', isLocked: false })
+const baseline = ref('')
+const dirty = computed(() => JSON.stringify(form) !== baseline.value)
+watch(
+    () => props.identity,
+    value => {
+        if (baseline.value && dirty.value) return
+        Object.assign(form, {
+            description: value?.description ?? '',
+            generationPrompt: value?.generationPrompt ?? '',
+            isLocked: value?.isLocked ?? false
+        })
+        baseline.value = JSON.stringify(form)
+    },
+    { immediate: true }
+)
+watch(dirty, value => emit('dirty', value), { immediate: true })
+
+/** 保存用户输入；空文本允许保存，但生图入口仍会检查必要 Prompt。 */
+function save() {
+    if (!props.disabled) emit('save', { ...form })
+}
+</script>
+
+<template>
+    <form class="mt-5" @submit.prevent="save">
+        <fieldset :disabled="disabled" class="space-y-4">
+            <label class="block"
+                ><span class="field-label">稳定身份描述</span
+                ><textarea
+                    id="identity-description"
+                    v-model="form.description"
+                    class="input min-h-28"
+                    placeholder="只描述跨形态不变的面部、结构、材质等特征"
+                />
+            </label>
+            <label class="block"
+                ><span class="field-label">身份核心提示词（稳定事实）</span
+                ><textarea
+                    id="identity-prompt"
+                    v-model="form.generationPrompt"
+                    class="input min-h-36"
+                    placeholder="不固定某一形态的服装、伤势或临时状态"
+                />
+            </label>
+            <p class="text-xs leading-6 text-muted">
+                {{
+                    module === 'prop'
+                        ? '保留道具固定文字、编号、部件位置、结构与材质组合；不要泛化原有辨识细节，也不要写临时损伤或摆放状态。'
+                        : module === 'scene'
+                          ? '保留建筑骨架、入口、窗户、楼梯与固定布局，不写临时天气、人群或灯光。'
+                          : '保留有依据的五官、骨相、发际线与体型，不固定某套服装、临时表情或伤势。'
+                }}
+                这里保存身份事实；背景、视角和参考图约束由后端另行编译，不是最终整段生图 Prompt。
+            </p>
+            <label class="flex items-start gap-2 text-xs leading-6"
+                ><input
+                    id="identity-lock"
+                    v-model="form.isLocked"
+                    type="checkbox"
+                    class="mt-1.5 accent-accent"
+                />确认并锁定身份文本。锁定后仍可人工保存、生成图片及切换母版。</label
+            >
+            <div class="flex flex-wrap items-center justify-between gap-3">
+                <span class="text-xs text-muted">{{
+                    dirty ? '有未保存修改，自动刷新不会覆盖草稿。' : '图片以已保存的身份提示词为准。'
+                }}</span
+                ><button class="button button-primary" type="submit" :disabled="disabled">保存主体身份</button>
+            </div>
+        </fieldset>
+    </form>
+</template>

src/features/subject-identity/components/IdentityGallery.vue

@@ -0,0 +1,141 @@
+<script setup lang="ts">
+import { computed, ref, watch } from 'vue'
+import { AssetImage, StatusBadge } from '../../../components/ui'
+import { referenceImageUrl } from '../../../lib/assets'
+import { formatDate } from '../../../lib/format'
+import { canBeAnchor, currentAnchor, identityViewLabels, readImageProvenance } from '../model'
+import type { IdentityImage } from '../types'
+
+/** 身份图库区分权威母版、primary 候选和辅助视图，保留后端原始状态。 */
+const props = defineProps<{ images: IdentityImage[]; subjectName: string; disabled: boolean }>()
+const emit = defineEmits<{ anchor: [imageId: string] }>()
+const selectedId = ref('')
+const confirming = ref(false)
+const anchor = computed(() => currentAnchor(props.images))
+const selected = computed(
+    () => props.images.find(image => image.id === selectedId.value) ?? anchor.value ?? props.images[0]
+)
+const canSelect = computed(
+    () => !props.disabled && !!selected.value && canBeAnchor(selected.value) && !selected.value.isAnchor
+)
+const imageUrl = computed(() => referenceImageUrl(selected.value?.imageUrl ?? ''))
+watch(
+    () => selected.value?.id,
+    () => {
+        confirming.value = false
+    }
+)
+
+/** 只发出用户确认过的目标 ID，父页面再核验图片归属。 */
+function chooseAnchor() {
+    if (!canSelect.value || !confirming.value || !selected.value) return
+    emit('anchor', selected.value.id)
+    confirming.value = false
+}
+</script>
+
+<template>
+    <section class="mt-5 border-t border-line pt-5" aria-label="身份参考图库">
+        <div class="flex items-center justify-between gap-3">
+            <h3 class="text-sm font-medium">身份参考图 · {{ images.length }}</h3>
+            <span class="tag">{{ anchor ? '已选母版' : '尚无母版' }}</span>
+        </div>
+        <p class="mt-2 text-xs leading-6 text-muted">
+            身份母版确定“是谁”；形态主图确定“该造型长什么样”。切换母版仅影响之后的生成，不会替换已有形态图、分镜参考图或提示词。
+        </p>
+        <template v-if="selected">
+            <AssetImage
+                :src="selected.status === 'completed' ? selected.imageUrl : null"
+                :alt="`${subjectName} · ${identityViewLabels[selected.viewType]}`"
+                class="asset-image-preview mt-4"
+                :empty-text="selected.status === 'failed' ? '本次生成失败' : '等待后端图片结果'"
+            />
+            <div class="mt-3 flex flex-wrap items-center justify-between gap-3">
+                <div class="flex flex-wrap items-center gap-2">
+                    <StatusBadge :status="selected.status" /><span class="tag">{{
+                        selected.isAnchor
+                            ? '当前身份母版'
+                            : selected.viewType === 'primary'
+                              ? '母版候选'
+                              : identityViewLabels[selected.viewType]
+                    }}</span
+                    ><span class="text-xs text-muted"
+                        >{{ selected.enabled ? '启用' : '停用' }} · {{ selected.width || '—' }} ×
+                        {{ selected.height || '—' }}</span
+                    >
+                </div>
+                <a
+                    v-if="imageUrl && selected.status === 'completed'"
+                    :href="imageUrl"
+                    target="_blank"
+                    rel="noopener noreferrer"
+                    class="text-button"
+                    >打开原图</a
+                >
+            </div>
+            <p v-if="selected.error" class="alert alert-error mt-3" role="alert">{{ selected.error }}</p>
+            <div class="mt-3 flex flex-wrap items-center gap-3">
+                <button
+                    v-if="!selected.isAnchor"
+                    class="button button-secondary"
+                    :disabled="!canSelect"
+                    @click="confirming = true"
+                >
+                    设为身份母版</button
+                ><span class="text-[11px] text-muted"
+                    >{{ selected.provider }} · {{ selected.model }} · {{ formatDate(selected.createdAt) }}</span
+                >
+            </div>
+            <p v-if="selected.viewType !== 'primary'" class="mt-2 text-xs text-muted">
+                辅助视图不能直接设为母版；可以选它作为参考，另生成一张 primary 候选。
+            </p>
+            <div v-if="confirming" class="alert mt-3">
+                <p class="text-xs">
+                    确认切换身份母版？旧 primary
+                    母版将停用并保留。后续人物／场景形态生图会引用新母版，道具暂不自动引用。此操作不调用模型。
+                </p>
+                <div class="mt-3 flex gap-3">
+                    <button class="button button-primary" :disabled="!canSelect" @click="chooseAnchor">
+                        确认切换身份母版</button
+                    ><button class="text-button" @click="confirming = false">取消</button>
+                </div>
+            </div>
+            <details v-if="selected.prompt" class="mt-4 text-xs">
+                <summary class="cursor-pointer text-muted">本次实际提示词</summary>
+                <p class="mt-3 whitespace-pre-wrap leading-6">{{ selected.prompt }}</p>
+            </details>
+            <p
+                v-if="readImageProvenance(selected.rawJson).referenceImageId"
+                class="mt-3 break-all text-[11px] text-muted"
+            >
+                本次锚定图片 ID：{{ readImageProvenance(selected.rawJson).referenceImageId }}
+            </p>
+            <p class="mt-2 break-all font-mono text-[10px] text-muted">Identity image ID · {{ selected.id }}</p>
+            <div class="image-history mt-4" aria-label="身份图片历史">
+                <button
+                    v-for="image in images"
+                    :key="image.id"
+                    class="image-history-item"
+                    :class="{ selected: image.id === selected.id }"
+                    :aria-label="`查看身份图片 ${image.id}`"
+                    :aria-pressed="image.id === selected.id"
+                    @click="selectedId = image.id"
+                >
+                    <AssetImage
+                        :src="image.status === 'completed' ? image.imageUrl : null"
+                        :alt="identityViewLabels[image.viewType]"
+                        :retryable="false"
+                        :empty-text="image.status === 'failed' ? '失败' : '无图'"
+                    /><span class="mt-1 block text-[10px]">{{
+                        image.isAnchor
+                            ? '当前母版'
+                            : image.viewType === 'primary'
+                              ? '母版候选'
+                              : identityViewLabels[image.viewType]
+                    }}</span>
+                </button>
+            </div>
+        </template>
+        <p v-else class="py-8 text-sm text-muted">尚无身份参考图。先保存身份提示词，再生成第一张母版。</p>
+    </section>
+</template>

src/features/subject-identity/components/IdentityImageDialog.vue

@@ -0,0 +1,132 @@
+<script setup lang="ts">
+import { computed, ref, watch } from 'vue'
+import { AppDialog } from '../../../components/ui'
+import { validImageSize } from '../../subject-images/model'
+import { currentAnchor, identityViewLabels } from '../model'
+import type { GenerateIdentityImageInput, IdentityImage, IdentityViewType } from '../types'
+
+/** 身份图生成配置；默认自动引用母版，不提供后端不存在的“忽略母版”开关。 */
+const props = defineProps<{ subjectId: string; subjectName: string; images: IdentityImage[]; disabled: boolean }>()
+const open = defineModel<boolean>('open', { default: false })
+const emit = defineEmits<{ generate: [input: GenerateIdentityImageInput] }>()
+const viewType = ref<IdentityViewType>('primary')
+const referenceImageId = ref('')
+const prompt = ref('')
+const width = ref<number | ''>('')
+const height = ref<number | ''>('')
+const acknowledged = ref(false)
+const anchor = computed(() => currentAnchor(props.images))
+const references = computed(() => props.images.filter(image => image.status === 'completed' && image.imageUrl))
+const valid = computed(
+    () =>
+        validImageSize(width.value, height.value) &&
+        (!referenceImageId.value || references.value.some(image => image.id === referenceImageId.value))
+)
+watch(
+    () => [open.value, props.subjectId],
+    () => {
+        viewType.value = 'primary'
+        referenceImageId.value = ''
+        prompt.value = ''
+        width.value = ''
+        height.value = ''
+        acknowledged.value = false
+    }
+)
+
+/** 提交显式的 Seedream 参数；省略空参考 ID 以保留后端自动选择逻辑。 */
+function submit() {
+    if (props.disabled || !valid.value || !acknowledged.value) return
+    emit('generate', {
+        provider: 'seedream',
+        viewType: viewType.value,
+        ...(referenceImageId.value ? { referenceImageId: referenceImageId.value } : {}),
+        ...(prompt.value.trim() ? { prompt: prompt.value.trim() } : {}),
+        ...(width.value !== '' && height.value !== '' ? { width: width.value, height: height.value } : {})
+    })
+    open.value = false
+}
+</script>
+
+<template>
+    <AppDialog v-model:open="open" title="生成身份参考图" :description="subjectName">
+        <form class="mt-5 space-y-4" @submit.prevent="submit">
+            <p class="alert text-xs">
+                调用 Seedream，每次新增一张图片，可能产生费用。第一张成功的 primary 图自动成为母版；已有母版时，新
+                primary 图仅作为停用候选保留。
+            </p>
+            <label class="block"
+                ><span class="field-label">参考视角</span
+                ><select id="identity-view" v-model="viewType" class="input">
+                    <option v-for="(label, value) in identityViewLabels" :key="value" :value="value">
+                        {{ label }}（{{ value }}）
+                    </option>
+                </select></label
+            >
+            <label class="block"
+                ><span class="field-label">身份锚定来源</span
+                ><select id="identity-reference" v-model="referenceImageId" class="input">
+                    <option value="">
+                        {{ anchor ? '自动引用当前身份母版' : '自动选择（当前无母版，将按文本生成）' }}
+                    </option>
+                    <option v-for="image in references" :key="image.id" :value="image.id">
+                        {{ identityViewLabels[image.viewType] }} · {{ image.id }}
+                    </option>
+                </select></label
+            >
+            <p class="text-xs leading-6 text-muted">
+                只能选择此主体已完成的图片。后端需取得 Provider 可访问的远程地址；本地预览可见不保证远程地址仍有效。
+            </p>
+            <label class="block"
+                ><span class="field-label">覆盖本次完整提示词（可选）</span
+                ><textarea
+                    id="identity-image-prompt"
+                    v-model="prompt"
+                    class="input min-h-24"
+                    placeholder="建议留空，使用后端编译的身份约束和视角要求"
+                />
+            </label>
+            <p v-if="prompt.trim()" class="text-xs text-danger">
+                自定义文本会替换后端编译的完整提示词，请自行包含身份约束和视角要求。
+            </p>
+            <div class="grid grid-cols-2 gap-4">
+                <label
+                    ><span class="field-label">宽度（px）</span
+                    ><input
+                        id="identity-width"
+                        v-model.number="width"
+                        type="number"
+                        min="1"
+                        step="1"
+                        class="input"
+                        placeholder="后端默认" /></label
+                ><label
+                    ><span class="field-label">高度（px）</span
+                    ><input
+                        id="identity-height"
+                        v-model.number="height"
+                        type="number"
+                        min="1"
+                        step="1"
+                        class="input"
+                        placeholder="后端默认"
+                /></label>
+            </div>
+            <p v-if="!valid" class="text-xs text-danger" role="alert">尺寸须成对填写正整数；指定的参考图必须仍可用。</p>
+            <label class="flex items-start gap-2 text-xs leading-6"
+                ><input
+                    id="identity-image-cost"
+                    v-model="acknowledged"
+                    type="checkbox"
+                    class="mt-1.5 accent-accent"
+                />我已确认后台没有同项目任务，了解本次会调用模型并可能产生费用。</label
+            >
+            <div class="dialog-footer">
+                <button type="button" class="button button-secondary" @click="open = false">取消</button
+                ><button class="button button-primary" type="submit" :disabled="disabled || !valid || !acknowledged">
+                    确认生成身份图
+                </button>
+            </div>
+        </form>
+    </AppDialog>
+</template>

src/features/subject-identity/identity.test.ts

@@ -0,0 +1,124 @@
+import { flushPromises, mount, type VueWrapper } from '@vue/test-utils'
+import { afterEach, describe, expect, it } from 'vitest'
+import IdentityImageDialog from './components/IdentityImageDialog.vue'
+import IdentityGallery from './components/IdentityGallery.vue'
+import { canBeAnchor, currentAnchor, groupIdentitySubjects, readImageProvenance } from './model'
+import { identityImageFixture } from './testing/fixtures'
+import { formFixture } from '../subject-images/testing/fixtures'
+
+let wrapper: VueWrapper | undefined
+afterEach(() => {
+    wrapper?.unmount()
+    wrapper = undefined
+    document.body.innerHTML = ''
+})
+
+/** 从真实 Reka Portal 内获取确认按钮。 */
+function button(label: string) {
+    const item = [...document.querySelectorAll('button')].find(element => element.textContent?.trim() === label)
+    if (!item) throw new Error(`缺少按钮 ${label}`)
+    return item
+}
+
+/** 修改 Portal 表单控件并触发 Vue 绑定。 */
+function input(selector: string, value: string) {
+    const item = document.querySelector<HTMLInputElement | HTMLSelectElement>(selector)!
+    item.value = value
+    item.dispatchEvent(new Event(item.tagName === 'SELECT' ? 'change' : 'input', { bubbles: true }))
+}
+
+describe('身份图与母版契约', () => {
+    it('启用的辅助视角不是母版，primary 候选停用时仍可选为母版', () => {
+        const front = identityImageFixture({ id: 'front', viewType: 'front', isAnchor: false })
+        const candidate = identityImageFixture({ id: 'candidate', enabled: false, isAnchor: false })
+        expect(canBeAnchor(front)).toBe(false)
+        expect(canBeAnchor(candidate)).toBe(true)
+        expect(currentAnchor([front, candidate])).toBeUndefined()
+        expect(canBeAnchor(identityImageFixture({ status: 'failed' }))).toBe(false)
+        expect(canBeAnchor(identityImageFixture({ imageUrl: null }))).toBe(false)
+    })
+
+    it('正式主体关联校验不接受不同主体的 form，追溯 JSON 兼容旧数据', () => {
+        expect(() => groupIdentitySubjects([{ ...formFixture(), subjectId: 'wrong' }])).toThrow('不匹配')
+        expect(readImageProvenance('{bad')).toEqual({})
+        expect(readImageProvenance(null)).toEqual({})
+        expect(readImageProvenance('{"identityAnchorImageId":"anchor","referenceImageId":7}')).toMatchObject({
+            identityAnchorImageId: 'anchor',
+            referenceImageId: undefined
+        })
+    })
+
+    it('辅助视角传递明确参考图、成对尺寸和本次 Prompt，不传形态生图字段', async () => {
+        wrapper = mount(IdentityImageDialog, {
+            attachTo: document.body,
+            props: {
+                open: true,
+                subjectId: 's1',
+                subjectName: '林夏',
+                images: [identityImageFixture()],
+                disabled: false
+            }
+        })
+        await flushPromises()
+        input('#identity-view', 'three-quarter')
+        input('#identity-reference', 'identity-image-1')
+        input('#identity-width', '2048')
+        document.querySelector<HTMLInputElement>('#identity-image-cost')!.click()
+        await flushPromises()
+        expect(button('确认生成身份图').disabled).toBe(true)
+        input('#identity-height', '2048')
+        input('#identity-image-prompt', ' 自定义身份提示词 ')
+        await flushPromises()
+        button('确认生成身份图').click()
+        await flushPromises()
+        expect(wrapper.emitted('generate')).toEqual([
+            [
+                {
+                    provider: 'seedream',
+                    viewType: 'three-quarter',
+                    referenceImageId: 'identity-image-1',
+                    width: 2048,
+                    height: 2048,
+                    prompt: '自定义身份提示词'
+                }
+            ]
+        ])
+    })
+
+    it('切换主体清空生图配置和费用确认，失效参考图不能提交', async () => {
+        wrapper = mount(IdentityImageDialog, {
+            attachTo: document.body,
+            props: {
+                open: true,
+                subjectId: 's1',
+                subjectName: '林夏',
+                images: [identityImageFixture()],
+                disabled: false
+            }
+        })
+        await flushPromises()
+        input('#identity-reference', 'identity-image-1')
+        document.querySelector<HTMLInputElement>('#identity-image-cost')!.click()
+        await flushPromises()
+        await wrapper.setProps({ images: [] })
+        expect(button('确认生成身份图').disabled).toBe(true)
+        await wrapper.setProps({ subjectId: 's2', subjectName: '陆川' })
+        expect(document.querySelector<HTMLSelectElement>('#identity-reference')!.value).toBe('')
+        expect(document.querySelector<HTMLInputElement>('#identity-image-cost')!.checked).toBe(false)
+    })
+
+    it('辅助图不能切换母版，提示词按纯文本展示', async () => {
+        wrapper = mount(IdentityGallery, {
+            attachTo: document.body,
+            props: {
+                subjectName: '林夏',
+                disabled: false,
+                images: [identityImageFixture({ viewType: 'front', isAnchor: false })]
+            }
+        })
+        expect(button('设为身份母版').disabled).toBe(true)
+        expect(wrapper.text()).toContain('<script>不执行</script>')
+        expect(wrapper.find('script').exists()).toBe(false)
+        expect(wrapper.emitted('anchor')).toBeUndefined()
+    })
+})

src/features/subject-identity/index.ts

@@ -0,0 +1,4 @@
+/** 主体身份模块公共入口。 */
+export { subjectIdentityApi } from './api'
+export { currentAnchor, readImageProvenance } from './model'
+export type { SubjectIdentity, IdentityImage, GenerateIdentityImageInput } from './types'

src/features/subject-identity/model.ts

@@ -0,0 +1,60 @@
+import { reactive } from 'vue'
+import type { SubjectFormAsset } from '../subject-images/types'
+import type { IdentityBatchResult, IdentityImage, IdentitySubject } from './types'
+
+/** 视角名称保持与后端四种固定值对应。 */
+export const identityViewLabels = {
+    primary: '身份母版',
+    front: '正面',
+    'three-quarter': '三分之四侧面',
+    'full-body': '全身'
+} as const
+
+/** 将同一主体的多个形态归并成一条，避免为每个形态重复创建 Identity。 */
+export function groupIdentitySubjects(forms: SubjectFormAsset[]): IdentitySubject[] {
+    const subjects = new Map<string, IdentitySubject>()
+    for (const form of forms) {
+        if (form.subjectId !== form.subject.id) throw new Error('形态与主体 ID 不匹配，请刷新后重试。')
+        const subject = subjects.get(form.subjectId) ?? { ...form.subject, forms: [] }
+        subject.forms.push(form)
+        subjects.set(form.subjectId, subject)
+    }
+    return [...subjects.values()]
+}
+
+/** 只有成功且有图片的 primary 候选可成为母版；停用候选也允许重新启用。 */
+export function canBeAnchor(image: IdentityImage): boolean {
+    return image.viewType === 'primary' && image.status === 'completed' && !!image.imageUrl
+}
+
+/** 使用后端 isAnchor，不把 front/full-body 的 enabled 错当成母版。 */
+export function currentAnchor(images: IdentityImage[]): IdentityImage | undefined {
+    return images.find(image => image.isAnchor && image.enabled && canBeAnchor(image))
+}
+
+/** 仅读取追溯所需 ID，不把原始 Provider 数据渲染为 HTML。 */
+export function readImageProvenance(rawJson?: string | null): {
+    referenceImageId?: string
+    identityAnchorImageId?: string
+} {
+    try {
+        const parsed: unknown = JSON.parse(rawJson || '{}')
+        if (!parsed || typeof parsed !== 'object') return {}
+        const value = parsed as Record<string, unknown>
+        return {
+            referenceImageId: typeof value.referenceImageId === 'string' ? value.referenceImageId : undefined,
+            identityAnchorImageId:
+                typeof value.identityAnchorImageId === 'string' ? value.identityAnchorImageId : undefined
+        }
+    } catch {
+        return {}
+    }
+}
+
+/** 批量回执按项目保存在当前会话，切页不丢失，刷新浏览器后不伪造恢复。 */
+const sessions = reactive<Record<string, { receipt: IdentityBatchResult | null }>>({})
+
+/** 取得指定项目的身份文本批量回执。 */
+export function getIdentitySession(projectId: string) {
+    return (sessions[projectId] ??= { receipt: null })
+}

src/features/subject-identity/testing/fixtures.ts

@@ -0,0 +1,39 @@
+import type { IdentityImage, SubjectIdentity } from '../types'
+
+/** 身份文本测试数据；subjectId 使用正式数据库主体 ID。 */
+export function identityFixture(overrides: Partial<SubjectIdentity> = {}): SubjectIdentity {
+    return {
+        id: 'identity-db-1',
+        subjectId: 'subject-db-1',
+        description: '稳定面部特征',
+        generationPrompt: '保持相同五官与骨相',
+        isLocked: false,
+        images: [],
+        createdAt: '2026-08-28T00:00:00Z',
+        updatedAt: '2026-08-28T00:00:00Z',
+        ...overrides
+    }
+}
+
+/** 专用身份图库查询包含 isAnchor，辅助视角即使启用也不是母版。 */
+export function identityImageFixture(overrides: Partial<IdentityImage> = {}): IdentityImage {
+    return {
+        id: 'identity-image-1',
+        identityId: 'identity-db-1',
+        source: 'generated',
+        viewType: 'primary',
+        provider: 'seedream',
+        model: 'configured-model',
+        prompt: '<script>不执行</script>',
+        imageUrl: '/storage/identity.png',
+        width: 2048,
+        height: 2048,
+        status: 'completed',
+        enabled: true,
+        isAnchor: true,
+        error: null,
+        createdAt: '2026-08-28T00:00:00Z',
+        updatedAt: '2026-08-28T00:00:00Z',
+        ...overrides
+    }
+}

src/features/subject-identity/types.ts

@@ -0,0 +1,66 @@
+import type { SubjectImageStatus, SubjectFormAsset } from '../subject-images/types'
+
+/** 只有 primary 视图可以选为身份母版，其余视图是辅助参考。 */
+export type IdentityViewType = 'primary' | 'front' | 'three-quarter' | 'full-body'
+
+/** 身份参考图；isAnchor 仅由专用图片查询接口返回，不能用 enabled 代替。 */
+export interface IdentityImage {
+    id: string
+    identityId: string
+    source: 'upload' | 'generated'
+    viewType: IdentityViewType
+    provider: string | null
+    model: string | null
+    prompt: string | null
+    imageUrl: string | null
+    width: number | null
+    height: number | null
+    status: SubjectImageStatus
+    enabled: boolean
+    isAnchor?: boolean
+    error: string | null
+    rawJson?: string | null
+    createdAt: string
+    updatedAt: string
+}
+
+/** 稳定身份描述及生图提示词，不混入具体形态的临时状态。 */
+export interface SubjectIdentity {
+    id: string
+    subjectId: string
+    description: string | null
+    generationPrompt: string | null
+    isLocked: boolean
+    images: IdentityImage[]
+    createdAt: string
+    updatedAt: string
+}
+
+/** 从正式形态目录归并主体，不使用 checkpoint 的候选 ID。 */
+export type IdentitySubject = SubjectFormAsset['subject'] & {
+    forms: SubjectFormAsset[]
+}
+
+/** 人工保存身份；锁定只保护 AI 文本重生成，不限制人工保存或生图。 */
+export type SaveIdentityInput = Partial<Pick<SubjectIdentity, 'description' | 'generationPrompt' | 'isLocked'>>
+
+/** 单张身份生图；空 referenceImageId 应省略，由后端自动引用当前母版。 */
+export interface GenerateIdentityImageInput {
+    provider: 'seedream'
+    viewType: IdentityViewType
+    referenceImageId?: string
+    prompt?: string
+    width?: number
+    height?: number
+}
+
+/** 身份文本批量生成回执，不代表已生成参考图。 */
+export interface IdentityBatchResult {
+    total: number
+    targetCount: number
+    generated: number
+    skipped: number
+    skippedLocked: number
+    failed: number
+    failures: { subjectId: string; subjectRef: string; error: string }[]
+}

src/features/subject-identity/useSubjectIdentity.ts

@@ -0,0 +1,201 @@
+import { computed, ref, watch } from 'vue'
+import { usePolling } from '../../composables/usePolling'
+import { visualStyleApi } from '../visual-style'
+import { subjectImagesApi } from '../subject-images/api'
+import { hasRunningImages } from '../subject-images/model'
+import { runOperation } from '../workflows/operations'
+import { useProjectMutationGuard } from '../workflows/useProjectMutationGuard'
+import { subjectIdentityApi } from './api'
+import { canBeAnchor, getIdentitySession, groupIdentitySubjects } from './model'
+import type { GenerateIdentityImageInput, IdentityImage, SaveIdentityInput, SubjectIdentity } from './types'
+
+/** 单主体保存响应必须与提交时固定的正式 ID 一致。 */
+function assertIdentity(value: SubjectIdentity, id: string) {
+    if (!value || value.subjectId !== id || value.images.some(image => image.identityId !== value.id))
+        throw new Error('后端未返回匹配的主体身份，请刷新核对。')
+}
+
+/** 主体目录、项目风格和当前身份分开查询，未创建 Identity 不误报为空图库错误。 */
+export function useSubjectIdentity() {
+    const { projectId, operation, blocked: projectBlocked } = useProjectMutationGuard()
+    const selectedId = ref('')
+    const editorRevision = ref(0)
+    const dirty = ref(false)
+    const concurrency = ref(3)
+    const force = ref(false)
+    const catalog = usePolling(projectId, async (id, signal) => {
+        const forms = await subjectImagesApi.listForms(id, signal)
+        if (
+            forms.some(
+                form => form.subject.projectId !== id || form.images.some(image => image.subjectFormId !== form.id)
+            )
+        )
+            throw new Error('主体目录与当前项目不匹配，请刷新后重试。')
+        return { subjects: groupIdentitySubjects(forms), running: forms.some(form => hasRunningImages(form.images)) }
+    })
+    const styleQuery = usePolling(projectId, async (id, signal) => {
+        const style = await visualStyleApi.get(id, signal)
+        if (style && style.projectId !== id) throw new Error('视觉风格与当前项目不匹配。')
+        return { style }
+    })
+    const subjects = computed(() => catalog.data.value?.subjects ?? [])
+    const subject = computed(() => subjects.value.find(item => item.id === selectedId.value))
+    const selectionKey = computed(() => subject.value?.id ?? '')
+    const detail = usePolling(selectionKey, async (id, signal) => {
+        if (!id) return null
+        const identity = await subjectIdentityApi.get(id, signal)
+        if (!identity) return { identity: null, images: [] as IdentityImage[] }
+        assertIdentity(identity, id)
+        const images = await subjectIdentityApi.listImages(id, signal)
+        if (images.some(image => image.identityId !== identity.id)) throw new Error('身份图片与当前主体不匹配。')
+        return { identity, images }
+    })
+    const identity = computed(() => detail.data.value?.identity ?? null)
+    const images = computed(() => detail.data.value?.images ?? [])
+    const session = computed(() => getIdentitySession(projectId.value))
+    const blocked = computed(
+        () =>
+            projectBlocked.value ||
+            !!catalog.error.value ||
+            !catalog.data.value ||
+            !!catalog.data.value.running ||
+            !!styleQuery.error.value ||
+            !styleQuery.data.value
+    )
+    const detailBlocked = computed(
+        () =>
+            blocked.value ||
+            !subject.value ||
+            !detail.data.value ||
+            !!detail.error.value ||
+            hasRunningImages(images.value)
+    )
+    const hasStyle = computed(() => !!styleQuery.data.value?.style)
+    const canGenerateText = computed(
+        () => !detailBlocked.value && hasStyle.value && !identity.value?.isLocked && !dirty.value
+    )
+    const canGenerateImage = computed(
+        () => !detailBlocked.value && hasStyle.value && !!identity.value?.generationPrompt?.trim() && !dirty.value
+    )
+    const batchValid = computed(() => Number.isSafeInteger(concurrency.value) && concurrency.value > 0)
+
+    watch(subjects, rows => {
+        if (!rows.some(item => item.id === selectedId.value)) selectedId.value = rows[0]?.id ?? ''
+    })
+    watch(selectionKey, () => {
+        dirty.value = false
+        editorRevision.value++
+    })
+
+    /** 长请求跨主体切换后只更新原项目的操作回执，不用旧响应覆盖新主体。 */
+    async function writeIdentity(label: string, action: (id: string) => Promise<SubjectIdentity>) {
+        if (detailBlocked.value || !subject.value) return
+        const id = subject.value.id
+        const project = projectId.value
+        const ok = await runOperation(project, label, async () => {
+            const result = await action(id)
+            assertIdentity(result, id)
+            if (selectionKey.value === id && projectId.value === project) {
+                // 图片的 isAnchor 只能从专用查询获取，保存文本时暂时保留已有图片列表。
+                detail.data.value = { identity: result, images: images.value }
+                editorRevision.value++
+                dirty.value = false
+            }
+        })
+        if (ok) await detail.refresh()
+    }
+
+    /** 人工保存可修改已锁定身份；锁定只阻止自动重生成。 */
+    function save(input: SaveIdentityInput) {
+        return writeIdentity('保存主体身份', id => subjectIdentityApi.save(id, input))
+    }
+
+    /** 单主体 AI 生成与已存在文本的覆盖均需要页面确认。 */
+    function generate() {
+        if (!canGenerateText.value) return
+        const overwrite = !!identity.value
+        return writeIdentity('AI 生成主体身份', id => subjectIdentityApi.generate(id, overwrite))
+    }
+
+    /** 批量仅生成身份文本，已锁定项由后端跳过，回执保留部分失败。 */
+    async function generateProject() {
+        if (blocked.value || !hasStyle.value || !batchValid.value || dirty.value || !subjects.value.length) return
+        const id = projectId.value
+        const target = getIdentitySession(id)
+        const input = { concurrency: concurrency.value, force: force.value }
+        target.receipt = null
+        await runOperation(id, '批量生成主体身份文本', async () => {
+            target.receipt = await subjectIdentityApi.generateProject(id, input)
+        })
+        await detail.refresh()
+    }
+
+    /** 生图不修改文字锁定；即使有自定义 Prompt，也需已有身份 Prompt 与项目风格。 */
+    async function generateImage(input: GenerateIdentityImageInput) {
+        if (!canGenerateImage.value || !subject.value || !identity.value) return
+        if (
+            input.referenceImageId &&
+            !images.value.some(
+                image => image.id === input.referenceImageId && image.status === 'completed' && image.imageUrl
+            )
+        )
+            return
+        const id = subject.value.id
+        const identityId = identity.value.id
+        await runOperation(projectId.value, `生成 ${subject.value.name} 身份参考图`, async () => {
+            const image = await subjectIdentityApi.generateImage(id, input)
+            if (!image || image.identityId !== identityId || image.status !== 'completed' || !image.imageUrl)
+                throw new Error(image?.error || '后端未返回已完成的身份图片，请先刷新核对，不要立即重复生图。')
+        })
+        await detail.refresh()
+    }
+
+    /** 候选切换母版只允许成功的 primary 图，不能将辅助视角直接升级为母版。 */
+    async function setAnchor(imageId: string) {
+        const image = images.value.find(item => item.id === imageId)
+        if (detailBlocked.value || !subject.value || !image || !canBeAnchor(image) || image.isAnchor) return
+        const id = subject.value.id
+        const identityId = image.identityId
+        await runOperation(projectId.value, '切换主体身份母版', async () => {
+            const result = await subjectIdentityApi.setAnchor(id, imageId)
+            if (
+                !result ||
+                result.id !== imageId ||
+                result.identityId !== identityId ||
+                !result.enabled ||
+                !canBeAnchor(result)
+            )
+                throw new Error('接口未确认身份母版切换，请刷新核对。')
+        })
+        await detail.refresh()
+    }
+
+    return {
+        projectId,
+        operation,
+        selectedId,
+        subject,
+        subjects,
+        catalog,
+        styleQuery,
+        detail,
+        identity,
+        images,
+        session,
+        concurrency,
+        force,
+        editorRevision,
+        dirty,
+        blocked,
+        detailBlocked,
+        hasStyle,
+        canGenerateText,
+        canGenerateImage,
+        batchValid,
+        save,
+        generate,
+        generateProject,
+        generateImage,
+        setAnchor
+    }
+}

src/features/subject-images/SubjectImagesPage.vue

@@ -86,12 +86,17 @@ function resetFilters() {
         <div class="flex flex-wrap items-end justify-between gap-4">
             <div>
                 <h2 class="text-lg font-semibold">形态图片</h2>
-                <p class="mt-2 text-sm text-muted">为人物、场景和道具准备参考图。已有图片直接从数据库读取。</p>
+                <p class="mt-2 text-sm text-muted">为人物、场景和道具准备造型参考图。已有图片直接从数据库读取。</p>
             </div>
             <RouterLink :to="`/projects/${id}/storyboard`" class="text-button"
                 >去检查分镜参考图<ArrowRight :size="14"
             /></RouterLink>
         </div>
+        <p class="alert mt-5 text-xs">
+            人物与场景形态在有身份母版时自动引用，以保持人物身份或空间结构一致；无母版仍可独立生成。道具暂不自动引用。
+            <RouterLink :to="`/projects/${id}/subject-identity`" class="text-button ml-2">管理主体身份 →</RouterLink>
+            更换母版后，已有形态主图仍保留，需主动生图并选择新主图。
+        </p>
         <div class="panel mt-5 p-5">
             <div class="flex flex-wrap items-center justify-between gap-4">
                 <p class="text-sm">
@@ -208,6 +213,11 @@ function resetFilters() {
                     <p v-if="form.description" class="mt-3 line-clamp-2 text-xs leading-6 text-muted">
                         {{ form.description }}
                     </p>
+                    <RouterLink
+                        :to="{ path: `/projects/${id}/subject-identity`, query: { subjectId: form.subjectId } }"
+                        class="text-button mt-3"
+                        >查看主体身份与母版 →</RouterLink
+                    >
                     <div class="mt-3 flex flex-wrap items-center gap-2">
                         <StatusBadge v-if="hasRunningImages(form.images)" status="generating" /><span
                             v-if="primaryImage(form.images)"

src/features/subject-images/components/GenerateImageDialog.vue

@@ -52,6 +52,14 @@ function submit() {
     >
         <form class="mt-6 space-y-5" @submit.prevent="submit">
             <p class="alert text-xs">使用后端配置的 Seedream 模型，可能产生费用。每次新增一张图片，不删除历史结果。</p>
+            <p class="text-xs leading-6 text-muted">
+                {{
+                    form?.subject.module === 'character' || form?.subject.module === 'scene'
+                        ? '后端会自动引用此人物／场景当前的身份母版，保持身份或空间结构；没有母版时仍可按文本生成。更换母版不会自动更新已有形态图。'
+                        : '当前道具形态生图暂不自动引用身份母版。'
+                }}
+                身份母版与此处的“形态主参考图”是两种不同用途的图片。
+            </p>
             <label class="block"
                 ><span class="field-label">自定义提示词（可选，仅用于本次）</span
                 ><textarea

src/features/subject-images/components/ImageGalleryDialog.vue

@@ -5,6 +5,7 @@ import { AppDialog, AssetImage, StatusBadge } from '../../../components/ui'
 import { usePolling } from '../../../composables/usePolling'
 import { referenceImageUrl } from '../../../lib/assets'
 import { formatDate } from '../../../lib/format'
+import { readImageProvenance } from '../../subject-identity'
 import { getOperation, runOperation } from '../../workflows/operations'
 import { subjectImagesApi } from '../api'
 import { hasRunningImages, primaryImage } from '../model'
@@ -144,6 +145,13 @@ async function choosePrimary() {
                 </p>
             </details>
             <p class="mt-3 break-all font-mono text-[10px] text-muted">Image ID · {{ selected.id }}</p>
+            <p class="mt-2 break-all text-[11px] text-muted">
+                {{
+                    readImageProvenance(selected.rawJson).identityAnchorImageId
+                        ? `本次引用身份母版 ID：${readImageProvenance(selected.rawJson).identityAnchorImageId}`
+                        : '此记录未提供身份母版来源，不能据此判断是否使用了当前母版。'
+                }}
+            </p>
             <div class="image-history mt-5" aria-label="图片历史">
                 <button
                     v-for="image in images"

src/features/subject-images/images.test.ts

@@ -23,6 +23,15 @@ afterEach(() => {
 })
 
 describe('形态图片选择与操作', () => {
+    it.each(['character', 'scene', 'prop'])('按后端最新 %s 模块展示母版继承范围', async module => {
+        const form = formFixture()
+        form.subject.module = module
+        wrapper = mount(GenerateImageDialog, { attachTo: document.body, props: { open: true, form, disabled: false } })
+        await flushPromises()
+        expect(document.body.textContent).toContain(
+            module === 'prop' ? '道具形态生图暂不自动引用' : '后端会自动引用此人物／场景当前的身份母版'
+        )
+    })
     it('已有主图优先于新候选图和失败记录，无主图才回退到成功候选图', () => {
         const form = formFixture()
         form.images.unshift(

src/features/subject-images/model.ts

@@ -20,7 +20,7 @@ export function coverImage(form: SubjectFormAsset): SubjectImage | undefined {
 }
 
 /** 查询发现后台仍在运行时也阻止重复提交，不只依赖本地 loading。 */
-export function hasRunningImages(images: SubjectImage[]): boolean {
+export function hasRunningImages(images: Pick<SubjectImage, 'status'>[]): boolean {
     return images.some(image => image.status === 'pending' || image.status === 'generating')
 }
 

src/features/subject-images/types.ts

@@ -18,6 +18,8 @@ export interface SubjectImage {
     /** 详情查询才返回实际使用的 Prompt；列表不包含长文本。 */
     prompt?: string
     negativePrompt?: string | null
+    /** 详情保留生成追溯信息，可读取当时引用的身份母版 ID。 */
+    rawJson?: string | null
 }
 
 /** 正式 SubjectForm 及其所属主体、图片，不使用 checkpoint 的领域 formId。 */

src/features/visual-style/VisualStylePage.vue

@@ -0,0 +1,138 @@
+<script setup lang="ts">
+import { computed, ref } from 'vue'
+import { usePolling } from '../../composables/usePolling'
+import { runOperation } from '../workflows/operations'
+import { useProjectMutationGuard } from '../workflows/useProjectMutationGuard'
+import ConfirmAction from '../workflows/ConfirmAction.vue'
+import { visualStyleApi } from './api'
+import type { AddStyleImageInput, SaveVisualStyleInput, VisualStyle, VisualStyleImage } from './types'
+import StyleEditor from './components/StyleEditor.vue'
+import StyleImages from './components/StyleImages.vue'
+
+/** 项目视觉风格工作区：文本编辑、锁定和风格图记录分开管理。 */
+const { projectId, blocked } = useProjectMutationGuard()
+const query = usePolling(projectId, async (id, signal) => {
+    const style = await visualStyleApi.get(id, signal)
+    assertProject(style, id)
+    return { style }
+})
+const style = computed(() => query.data.value?.style ?? null)
+const disabled = computed(() => blocked.value || !!query.error.value || !query.data.value)
+const editorRevision = ref(0)
+const imageRevision = ref(0)
+const dirty = ref(false)
+
+/** 防止错误项目的数据进入编辑器；空风格是正常的首次使用状态。 */
+function assertProject(value: VisualStyle | null, id: string) {
+    if (value && (value.projectId !== id || value.images.some(image => image.visualStyleId !== value.id)))
+        throw new Error('视觉风格与当前项目不匹配，请刷新后重试。')
+}
+
+/** 操作成功后重建草稿，失败则保留用户输入；不因查询失败重发写请求。 */
+async function writeStyle(label: string, action: (id: string) => Promise<VisualStyle>) {
+    if (disabled.value) return
+    const id = projectId.value
+    const ok = await runOperation(id, label, async () => {
+        const result = await action(id)
+        if (!result) throw new Error('后端未返回已保存的视觉风格，请刷新核对。')
+        assertProject(result, id)
+        if (projectId.value === id) {
+            query.data.value = { style: result }
+            editorRevision.value++
+            dirty.value = false
+        }
+    })
+    if (ok) await query.refresh()
+}
+
+/** 人工保存允许修改锁定风格。 */
+function save(input: SaveVisualStyleInput) {
+    void writeStyle('保存项目视觉风格', id => visualStyleApi.save(id, input))
+}
+
+/** AI 覆盖必须先保存或放弃草稿，并在已锁定时禁止提交。 */
+function generate() {
+    if (style.value?.isLocked || dirty.value) return
+    const force = !!style.value
+    void writeStyle('AI 生成项目视觉风格', id => visualStyleApi.generate(id, force))
+}
+
+/** 图片写操作固定项目与风格归属，保存后再读取列表。 */
+async function writeImage(label: string, action: (id: string) => Promise<VisualStyleImage>, reset = false) {
+    if (disabled.value || !style.value) return
+    const id = projectId.value
+    const styleId = style.value.id
+    const ok = await runOperation(id, label, async () => {
+        const result = await action(id)
+        if (!result || result.visualStyleId !== styleId) throw new Error('后端未确认风格图片变更，请刷新核对。')
+    })
+    if (ok && reset && projectId.value === id) imageRevision.value++
+    await query.refresh()
+}
+
+/** 登记已存在的图片地址，不上传文件或调用模型。 */
+function addImage(input: AddStyleImageInput) {
+    void writeImage('登记风格参考图', id => visualStyleApi.addImage(id, input), true)
+}
+
+/** 风格图启停不受文字锁定影响。 */
+function toggleImage(image: VisualStyleImage) {
+    void writeImage('修改风格参考图启用状态', id => visualStyleApi.setImageEnabled(id, image.id, !image.enabled))
+}
+
+/** 删除已由子组件二次确认，不删除实际存储文件。 */
+function removeImage(image: VisualStyleImage) {
+    void writeImage('移除风格参考图记录', id => visualStyleApi.removeImage(id, image.id))
+}
+</script>
+
+<template>
+    <section class="mt-7">
+        <div class="flex flex-wrap items-end justify-between gap-4">
+            <div>
+                <h2 class="text-lg font-semibold">项目视觉风格</h2>
+                <p class="mt-2 text-sm text-muted">确定项目统一的视觉语言，再为人物、场景和道具建立稳定身份。</p>
+            </div>
+            <RouterLink :to="`/projects/${projectId}/subject-identity`" class="text-button"
+                >下一步：主体身份 →</RouterLink
+            >
+        </div>
+        <div class="my-5 flex flex-wrap items-center gap-3">
+            <span class="tag">{{ style ? (style.isLocked ? '已锁定' : '未锁定') : '尚未创建' }}</span>
+            <ConfirmAction
+                :label="style ? 'AI 重新生成风格' : 'AI 生成风格'"
+                :disabled="disabled || !!style?.isLocked || dirty"
+                acknowledgement
+                description="调用文本模型生成整体及分类风格。重新生成会覆盖已保存的文本与硬约束；旧身份、旧图片和已有提示词不会自动更新。"
+                @confirm="generate"
+            />
+            <button class="text-button" :disabled="query.loading.value" @click="query.refresh">刷新风格</button>
+        </div>
+        <p v-if="query.error.value" class="alert alert-error mb-4" role="alert">
+            {{ query.error.value }} 请确认后端 dev 已更新并重启；查询失败不等同于尚未创建。
+        </p>
+        <p v-if="dirty" class="mb-4 text-xs text-muted">
+            有未保存修改，AI 重生成暂不可用。自动刷新不会覆盖编辑内容；离开页面不会自动保存。
+        </p>
+        <p class="alert mb-5 text-xs">
+            修改风格只影响之后的相关生成。已有主体身份、形态图片和视频提示词不会自动更新，请按需要逐步重生成。
+        </p>
+        <StyleEditor
+            v-if="query.data.value"
+            :key="editorRevision"
+            :style="style"
+            :disabled="disabled"
+            @save="save"
+            @dirty="dirty = $event"
+        />
+        <p v-else-if="query.loading.value" class="py-8 text-sm text-muted" role="status">正在读取视觉风格…</p>
+        <StyleImages
+            :key="imageRevision"
+            :images="style?.images ?? []"
+            :disabled="disabled || !style"
+            @add="addImage"
+            @toggle="toggleImage"
+            @remove="removeImage"
+        />
+    </section>
+</template>

src/features/visual-style/api.ts

@@ -0,0 +1,25 @@
+import { request } from '../../lib/http'
+import type { AddStyleImageInput, SaveVisualStyleInput, VisualStyle, VisualStyleImage } from './types'
+
+/** 固定项目范围并编码正式 ID。 */
+function stylePath(projectId: string) {
+    return `/projects/${encodeURIComponent(projectId)}/visual-style`
+}
+
+/** 风格查询、编辑与图片登记；只有 generate 会调用文本模型。 */
+export const visualStyleApi = {
+    get: (projectId: string, signal?: AbortSignal) => request<VisualStyle | null>(stylePath(projectId), { signal }),
+    save: (projectId: string, input: SaveVisualStyleInput) =>
+        request<VisualStyle>(stylePath(projectId), { method: 'PUT', body: input }),
+    generate: (projectId: string, force: boolean) =>
+        request<VisualStyle>(`${stylePath(projectId)}/generate`, { method: 'POST', body: { force }, timeoutMs: 0 }),
+    addImage: (projectId: string, input: AddStyleImageInput) =>
+        request<VisualStyleImage>(`${stylePath(projectId)}/images`, { method: 'POST', body: input }),
+    setImageEnabled: (projectId: string, imageId: string, enabled: boolean) =>
+        request<VisualStyleImage>(`${stylePath(projectId)}/images/${encodeURIComponent(imageId)}/enabled`, {
+            method: 'PUT',
+            body: { enabled }
+        }),
+    removeImage: (projectId: string, imageId: string) =>
+        request<VisualStyleImage>(`${stylePath(projectId)}/images/${encodeURIComponent(imageId)}`, { method: 'DELETE' })
+}

src/features/visual-style/components/StyleEditor.vue

@@ -0,0 +1,124 @@
+<script setup lang="ts">
+import { computed, reactive, ref, watch } from 'vue'
+import type { SaveVisualStyleInput, VisualStyle } from '../types'
+
+/** 编辑草稿独立于轮询结果，刷新不会覆盖尚未保存的输入。 */
+const props = defineProps<{ style: VisualStyle | null; disabled: boolean }>()
+const emit = defineEmits<{ save: [input: SaveVisualStyleInput]; dirty: [value: boolean] }>()
+const form = reactive({
+    name: '',
+    prompt: '',
+    characterPrompt: '',
+    scenePrompt: '',
+    propPrompt: '',
+    constraints: '[]',
+    isLocked: false
+})
+const baseline = ref('')
+const dirty = computed(() => JSON.stringify(form) !== baseline.value)
+const constraintsError = computed(() => {
+    try {
+        const value: unknown = JSON.parse(form.constraints)
+        return Array.isArray(value) && value.every(item => typeof item === 'string')
+            ? ''
+            : '硬约束须为字符串数组，例如 ["真人写实", "禁止二次元"]。'
+    } catch {
+        return '硬约束 JSON 格式不正确，请修正后保存。'
+    }
+})
+watch(
+    () => props.style,
+    value => {
+        if (baseline.value && dirty.value) return
+        Object.assign(form, {
+            name: value?.name ?? '默认视觉风格',
+            prompt: value?.prompt ?? '',
+            characterPrompt: value?.characterPrompt ?? '',
+            scenePrompt: value?.scenePrompt ?? '',
+            propPrompt: value?.propPrompt ?? '',
+            constraints: JSON.stringify(value?.hardConstraints ?? [], null, 2),
+            isLocked: value?.isLocked ?? false
+        })
+        baseline.value = JSON.stringify(form)
+    },
+    { immediate: true }
+)
+watch(dirty, value => emit('dirty', value), { immediate: true })
+
+/** 显式校验 JSON 后保存结构化约束；不对错误结构做静默转换。 */
+function save() {
+    if (props.disabled || constraintsError.value) return
+    emit('save', {
+        name: form.name.trim(),
+        prompt: form.prompt,
+        characterPrompt: form.characterPrompt,
+        scenePrompt: form.scenePrompt,
+        propPrompt: form.propPrompt,
+        hardConstraints: JSON.parse(form.constraints) as string[],
+        isLocked: form.isLocked
+    })
+}
+</script>
+
+<template>
+    <form class="panel p-5 sm:p-6" @submit.prevent="save">
+        <fieldset :disabled="disabled" class="space-y-5">
+            <div class="flex flex-wrap items-center justify-between gap-3">
+                <h3 class="font-medium">风格说明</h3>
+                <span v-if="dirty" class="tag">有未保存修改</span>
+            </div>
+            <label class="block"
+                ><span class="field-label">风格名称</span><input id="style-name" v-model="form.name" class="input"
+            /></label>
+            <label class="block"
+                ><span class="field-label">整体视觉语言</span
+                ><textarea
+                    id="style-prompt"
+                    v-model="form.prompt"
+                    class="input min-h-32"
+                    placeholder="媒介、真实感、色彩和整体美术方向"
+                />
+            </label>
+            <div class="grid gap-5 lg:grid-cols-3">
+                <label
+                    ><span class="field-label">人物风格补充</span
+                    ><textarea v-model="form.characterPrompt" class="input min-h-32" />
+                </label>
+                <label
+                    ><span class="field-label">场景风格补充</span
+                    ><textarea v-model="form.scenePrompt" class="input min-h-32" />
+                </label>
+                <label
+                    ><span class="field-label">道具风格补充</span
+                    ><textarea v-model="form.propPrompt" class="input min-h-32" />
+                </label>
+            </div>
+            <label class="block"
+                ><span class="field-label">硬约束 · JSON 字符串数组</span
+                ><textarea
+                    id="style-constraints"
+                    v-model="form.constraints"
+                    class="input min-h-24 font-mono text-xs"
+                    spellcheck="false"
+                />
+            </label>
+            <p v-if="constraintsError" class="text-xs text-danger" role="alert">{{ constraintsError }}</p>
+            <label class="flex items-start gap-2 text-sm"
+                ><input
+                    id="style-lock"
+                    v-model="form.isLocked"
+                    type="checkbox"
+                    class="mt-1 accent-accent"
+                />确认并锁定风格，阻止后续 AI 自动覆盖</label
+            >
+            <p class="text-xs leading-6 text-muted">
+                锁定后仍可人工编辑并保存。取消勾选并保存后，才能再次使用 AI 重生成。
+            </p>
+            <div class="flex justify-end">
+                <button class="button button-primary" type="submit" :disabled="disabled || !!constraintsError">
+                    保存视觉风格
+                </button>
+            </div>
+        </fieldset>
+    </form>
+</template>

src/features/visual-style/components/StyleImages.vue

@@ -0,0 +1,120 @@
+<script setup lang="ts">
+import { computed, reactive, ref } from 'vue'
+import { AssetImage } from '../../../components/ui'
+import { referenceImageUrl } from '../../../lib/assets'
+import type { AddStyleImageInput, StyleCategory, VisualStyleImage } from '../types'
+
+/** 风格图只登记已有地址；不伪造文件上传或风格生图功能。 */
+const props = defineProps<{ images: VisualStyleImage[]; disabled: boolean }>()
+const emit = defineEmits<{
+    add: [input: AddStyleImageInput]
+    toggle: [image: VisualStyleImage]
+    remove: [image: VisualStyleImage]
+}>()
+const form = reactive({ imageUrl: '', category: 'overall' as StyleCategory, sortOrder: 0, enabled: true })
+const confirmingId = ref('')
+const categories = { overall: '整体', character: '人物', scene: '场景', prop: '道具' } as const
+const valid = computed(() => !!referenceImageUrl(form.imageUrl) && Number.isSafeInteger(form.sortOrder))
+
+/** 仅通过校验后发出新增请求，输入框在失败时保留供修正。 */
+function add() {
+    if (props.disabled || !valid.value) return
+    emit('add', { ...form, imageUrl: form.imageUrl.trim(), source: 'upload' })
+}
+
+/** 删除只移除记录，需确认且不声称删除远程文件。 */
+function remove(image: VisualStyleImage) {
+    if (props.disabled || confirmingId.value !== image.id) return
+    confirmingId.value = ''
+    emit('remove', image)
+}
+</script>
+
+<template>
+    <section class="panel mt-5 p-5 sm:p-6" aria-label="风格参考图">
+        <h3 class="font-medium">
+            风格参考图 <span class="ml-2 text-xs text-muted">{{ images.length }} 张</span>
+        </h3>
+        <p class="mt-2 text-xs leading-6 text-muted">
+            登记已有 HTTP(S) 或 /storage/
+            图片地址。后端尚无文件上传接口；这些图片目前仅管理记录，现有身份／形态生图不会自动将风格图片传给模型。
+        </p>
+        <form class="mt-4" @submit.prevent="add">
+            <fieldset :disabled="disabled" class="flex flex-wrap items-end gap-3">
+                <label class="min-w-48 flex-1"
+                    ><span class="field-label">图片地址</span
+                    ><input
+                        v-model="form.imageUrl"
+                        class="input"
+                        placeholder="https://… 或 /storage/…"
+                        aria-label="风格图片地址"
+                /></label>
+                <label
+                    ><span class="field-label">分类</span
+                    ><select v-model="form.category" class="input">
+                        <option v-for="(label, value) in categories" :key="value" :value="value">{{ label }}</option>
+                    </select></label
+                >
+                <label class="w-24"
+                    ><span class="field-label">排序</span
+                    ><input
+                        v-model.number="form.sortOrder"
+                        class="input"
+                        type="number"
+                        step="1"
+                        aria-label="风格图片排序"
+                /></label>
+                <label class="flex gap-2 pb-2 text-xs"><input v-model="form.enabled" type="checkbox" />启用</label>
+                <button class="button button-secondary" type="submit" :disabled="disabled || !valid">登记参考图</button>
+            </fieldset>
+        </form>
+        <p v-if="form.imageUrl && !valid" class="mt-2 text-xs text-danger" role="alert">
+            请填写有效图片地址和整数排序。
+        </p>
+        <div v-if="images.length" class="mt-5 grid gap-4 sm:grid-cols-2 xl:grid-cols-4">
+            <article v-for="image in images" :key="image.id" class="min-w-0 rounded border border-line p-3">
+                <AssetImage
+                    :src="image.imageUrl"
+                    :alt="`${categories[image.category]}风格参考图`"
+                    class="aspect-square"
+                />
+                <p class="mt-3 text-xs">
+                    {{ categories[image.category] }} · {{ image.enabled ? '已启用' : '已停用' }} · 排序
+                    {{ image.sortOrder }}
+                </p>
+                <p v-if="image.provider || image.model" class="mt-2 break-words text-[11px] text-muted">
+                    {{ image.provider }} {{ image.model }}
+                </p>
+                <details v-if="image.prompt" class="mt-2 text-xs">
+                    <summary>实际提示词</summary>
+                    <p class="mt-2 whitespace-pre-wrap">{{ image.prompt }}</p>
+                </details>
+                <div class="mt-3 flex flex-wrap gap-3">
+                    <a
+                        v-if="referenceImageUrl(image.imageUrl)"
+                        :href="referenceImageUrl(image.imageUrl)!"
+                        class="text-button"
+                        target="_blank"
+                        rel="noopener noreferrer"
+                        >原图</a
+                    >
+                    <button class="text-button" :disabled="disabled" @click="emit('toggle', image)">
+                        {{ image.enabled ? '停用' : '启用' }}
+                    </button>
+                    <button class="text-button text-danger" :disabled="disabled" @click="confirmingId = image.id">
+                        移除
+                    </button>
+                </div>
+                <div v-if="confirmingId === image.id" class="mt-3 text-xs">
+                    <p>仅移除参考图记录，不删除远程文件。</p>
+                    <div class="mt-2 flex gap-3">
+                        <button class="text-button text-danger" :disabled="disabled" @click="remove(image)">
+                            确认移除</button
+                        ><button class="text-button" @click="confirmingId = ''">取消</button>
+                    </div>
+                </div>
+            </article>
+        </div>
+        <p v-else class="mt-5 text-sm text-muted">尚无风格参考图。先保存视觉风格，再登记图片。</p>
+    </section>
+</template>

src/features/visual-style/index.ts

@@ -0,0 +1,3 @@
+/** 视觉风格模块公共入口。 */
+export { visualStyleApi } from './api'
+export type { VisualStyle, VisualStyleImage, StyleCategory, SaveVisualStyleInput } from './types'

src/features/visual-style/testing/fixtures.ts

@@ -0,0 +1,39 @@
+import type { VisualStyle, VisualStyleImage } from '../types'
+
+/** 仅供测试使用的项目视觉风格，与后端持久化返回字段一致。 */
+export function styleFixture(overrides: Partial<VisualStyle> = {}): VisualStyle {
+    return {
+        id: 'style-db-1',
+        projectId: 'page-test-project',
+        name: '都市悬疑真人短剧',
+        prompt: '真人写实，冷暖对比',
+        characterPrompt: '真实皮肤质感',
+        scenePrompt: '现代城市',
+        propPrompt: '真实材质',
+        hardConstraints: ['禁止二次元'],
+        isLocked: false,
+        images: [],
+        createdAt: '2026-08-28T00:00:00Z',
+        updatedAt: '2026-08-28T00:00:00Z',
+        ...overrides
+    }
+}
+
+/** 风格图记录不包含文件上传或模型调用。 */
+export function styleImageFixture(overrides: Partial<VisualStyleImage> = {}): VisualStyleImage {
+    return {
+        id: 'style-image-1',
+        visualStyleId: 'style-db-1',
+        category: 'overall',
+        source: 'upload',
+        imageUrl: '/storage/style.png',
+        enabled: true,
+        sortOrder: 0,
+        provider: null,
+        model: null,
+        prompt: null,
+        createdAt: '2026-08-28T00:00:00Z',
+        updatedAt: '2026-08-28T00:00:00Z',
+        ...overrides
+    }
+}

src/features/visual-style/types.ts

@@ -0,0 +1,51 @@
+/** 风格参考图分类，与后端枚举保持一致。 */
+export type StyleCategory = 'overall' | 'character' | 'scene' | 'prop'
+
+/** 已登记的风格参考图；这里只登记地址，不代表前端上传了文件。 */
+export interface VisualStyleImage {
+    id: string
+    visualStyleId: string
+    category: StyleCategory
+    source: 'upload' | 'generated'
+    imageUrl: string
+    enabled: boolean
+    sortOrder: number
+    provider: string | null
+    model: string | null
+    prompt: string | null
+    createdAt: string
+    updatedAt: string
+}
+
+/** 项目统一视觉风格；hardConstraints 保留后端 JSON，避免静默丢弃旧结构。 */
+export interface VisualStyle {
+    id: string
+    projectId: string
+    name: string
+    prompt: string | null
+    characterPrompt: string | null
+    scenePrompt: string | null
+    propPrompt: string | null
+    hardConstraints: unknown
+    isLocked: boolean
+    images: VisualStyleImage[]
+    createdAt: string
+    updatedAt: string
+}
+
+/** 人工保存只发送本次可编辑字段，锁定仍允许用户主动保存。 */
+export type SaveVisualStyleInput = Partial<
+    Pick<
+        VisualStyle,
+        'name' | 'prompt' | 'characterPrompt' | 'scenePrompt' | 'propPrompt' | 'hardConstraints' | 'isLocked'
+    >
+>
+
+/** 新增参考图使用已有地址，允许指定分类、启用状态及展示顺序。 */
+export interface AddStyleImageInput {
+    category: StyleCategory
+    source: 'upload'
+    imageUrl: string
+    enabled: boolean
+    sortOrder: number
+}

src/features/workflows/ConfirmAction.vue

@@ -1,5 +1,5 @@
 <script setup lang="ts">
-import { ref } from 'vue'
+import { ref, watch } from 'vue'
 import { DialogTrigger } from 'reka-ui'
 import { AppDialog } from '../../components/ui'
 
@@ -15,6 +15,11 @@ const emit = defineEmits<{ confirm: [] }>()
 const open = ref(false)
 const acknowledged = ref(false)
 
+/** 取消后重新打开，或确认目标改变，都必须重新确认任务与费用。 */
+watch([open, () => props.label, () => props.description], () => {
+    acknowledged.value = false
+})
+
 /** 发出事件后关闭弹窗，操作状态由项目级锁负责。 */
 function confirm() {
     if (props.disabled || (props.acknowledgement && !acknowledged.value)) return

src/features/workflows/pages.test.ts

@@ -22,6 +22,30 @@ import { getOperation } from './operations'
 import SubjectImagesPage from '../subject-images/SubjectImagesPage.vue'
 import { formFixture, imageFixture } from '../subject-images/testing/fixtures'
 import { getImageSession } from '../subject-images/model'
+import VisualStylePage from '../visual-style/VisualStylePage.vue'
+import SubjectIdentityPage from '../subject-identity/SubjectIdentityPage.vue'
+import { styleFixture, styleImageFixture } from '../visual-style/testing/fixtures'
+import { identityFixture, identityImageFixture } from '../subject-identity/testing/fixtures'
+import { getIdentitySession } from '../subject-identity/model'
+
+/** 新资产页面复用已有项目上下文和内存路由，不接入真实模型。 */
+async function mountAssets(page: 'style' | 'identity', query = '') {
+    const component = page === 'style' ? VisualStylePage : SubjectIdentityPage
+    const router = createRouter({ history: createMemoryHistory(), routes: [{ path: '/assets', component }] })
+    await router.push('/assets' + query)
+    const provided = context()
+    wrapper = mount(component, {
+        attachTo: document.body,
+        global: { plugins: [router], provide: { [projectContextKey as symbol]: provided } }
+    })
+    await flushPromises()
+    return provided
+}
+
+/** 统一构造独立响应，避免复用已消费的 Response body。 */
+function jsonResponse(data: unknown) {
+    return new Response(JSON.stringify({ data }))
+}
 
 /** 模拟 API 响应仅用于测试，不会打包进生产页面。 */
 const fixture: ProjectDetail = {
@@ -99,6 +123,7 @@ async function mountSubjectImages() {
 }
 
 afterEach(() => {
+    getIdentitySession(fixture.id).receipt = null
     wrapper?.unmount()
     wrapper = undefined
     document.body.innerHTML = ''
@@ -108,6 +133,362 @@ afterEach(() => {
     Object.assign(getOperation(fixture.id), { pending: false, label: '', error: '', notice: '' })
 })
 
+describe('视觉风格与主体身份工作区', () => {
+    it('从形态深链接进入后，手动切换主体不会被目录刷新切回', async () => {
+        const form = formFixture()
+        const second = {
+            ...form,
+            id: 'form-2',
+            subjectId: 'subject-2',
+            images: [],
+            subject: { ...form.subject, id: 'subject-2', name: '陆川', ref: '@CH0002' }
+        }
+        vi.stubGlobal(
+            'fetch',
+            vi.fn<typeof fetch>().mockImplementation(async url => {
+                if (String(url).endsWith('/subject-forms')) return jsonResponse([form, second])
+                if (String(url).endsWith('/visual-style')) return jsonResponse(styleFixture())
+                return jsonResponse(null)
+            })
+        )
+        await mountAssets('identity', '?subjectId=subject-db-1')
+        await wrapper!.findAll('.identity-subject-item')[1]!.trigger('click')
+        await flushPromises()
+        button('刷新主体目录').click()
+        await flushPromises()
+        expect(wrapper!.findAll('.identity-subject-item')[1]!.attributes('aria-pressed')).toBe('true')
+    })
+
+    it('身份文本保存发送正式主体 ID 与锁定状态，草稿切换需确认', async () => {
+        let identity = identityFixture()
+        const form = formFixture()
+        const second = {
+            ...form,
+            id: 'form-2',
+            subjectId: 'subject-2',
+            images: [],
+            subject: { ...form.subject, id: 'subject-2', name: '陆川', ref: '@CH0002' }
+        }
+        const fetcher = vi.fn<typeof fetch>().mockImplementation(async (url, init) => {
+            if (String(url).endsWith('/subject-forms')) return jsonResponse([form, second])
+            if (String(url).endsWith('/visual-style')) return jsonResponse(styleFixture())
+            if (String(url).endsWith('/identity/images')) return jsonResponse([])
+            if (String(url).includes('/subject-2/')) return jsonResponse(null)
+            if (init?.method === 'PUT') identity = { ...identity, ...JSON.parse(init.body as string) }
+            return jsonResponse(identity)
+        })
+        vi.stubGlobal('fetch', fetcher)
+        await mountAssets('identity')
+        await wrapper!.get('#identity-description').setValue('人工稳定身份')
+        await wrapper!.get('#identity-lock').setValue(true)
+        button('保存主体身份').click()
+        await flushPromises()
+        const put = fetcher.mock.calls.find(([, init]) => init?.method === 'PUT')!
+        expect(put[0]).toBe('/api/subjects/subject-db-1/identity')
+        expect(JSON.parse(put[1]!.body as string)).toMatchObject({ description: '人工稳定身份', isLocked: true })
+        expect(button('AI 重新生成身份').disabled).toBe(true)
+        await wrapper!.get('#identity-description').setValue('新的未保存草稿')
+        await wrapper!.findAll('.identity-subject-item')[1]!.trigger('click')
+        expect(wrapper!.text()).toContain('切换会丢弃草稿')
+        expect((wrapper!.get('#identity-description').element as HTMLTextAreaElement).value).toBe('新的未保存草稿')
+        button('放弃草稿并切换').click()
+        await flushPromises()
+        expect((wrapper!.get('#identity-description').element as HTMLTextAreaElement).value).toBe('')
+    })
+
+    it('风格图拒绝危险 URL，登记只传已有地址及分类排序', async () => {
+        const image = styleImageFixture()
+        const fetcher = vi
+            .fn<typeof fetch>()
+            .mockImplementation(async (_url, init) => jsonResponse(init?.method === 'POST' ? image : styleFixture()))
+        vi.stubGlobal('fetch', fetcher)
+        await mountAssets('style')
+        await wrapper!.get('[aria-label="风格图片地址"]').setValue('javascript:alert(1)')
+        expect(button('登记参考图').disabled).toBe(true)
+        await wrapper!.get('[aria-label="风格图片地址"]').setValue('/storage/style.png')
+        button('登记参考图').click()
+        await flushPromises()
+        const post = fetcher.mock.calls.find(([, init]) => init?.method === 'POST')!
+        expect(post[0]).toBe('/api/projects/page-test-project/visual-style/images')
+        expect(JSON.parse(post[1]!.body as string)).toEqual({
+            imageUrl: '/storage/style.png',
+            category: 'overall',
+            source: 'upload',
+            enabled: true,
+            sortOrder: 0
+        })
+        expect((wrapper!.get('[aria-label="风格图片地址"]').element as HTMLInputElement).value).toBe('')
+    })
+
+    it('AI 确认取消后重新打开必须再次确认，不能复用上次勾选', async () => {
+        const fetcher = vi.fn<typeof fetch>().mockImplementation(async () => jsonResponse(styleFixture()))
+        vi.stubGlobal('fetch', fetcher)
+        await mountAssets('style')
+        button('AI 重新生成风格').click()
+        await flushPromises()
+        document.querySelector<HTMLInputElement>('[role="dialog"] input[type="checkbox"]')!.click()
+        await flushPromises()
+        expect(button('确认AI 重新生成风格').disabled).toBe(false)
+        button('取消').click()
+        await flushPromises()
+        button('AI 重新生成风格').click()
+        await flushPromises()
+        expect(button('确认AI 重新生成风格').disabled).toBe(true)
+        expect(fetcher.mock.calls.some(([, init]) => init?.method === 'POST')).toBe(false)
+    })
+    it('视觉风格首次查询为空时可人工保存，JSON 校验和锁定字段准确传递', async () => {
+        let style: ReturnType<typeof styleFixture> | null = null
+        const fetcher = vi.fn<typeof fetch>().mockImplementation(async (_url, init) => {
+            if (init?.method === 'PUT') style = styleFixture(JSON.parse(init.body as string))
+            return jsonResponse(style)
+        })
+        vi.stubGlobal('fetch', fetcher)
+        await mountAssets('style')
+        expect(wrapper!.text()).toContain('尚未创建')
+        await wrapper!.get('#style-constraints').setValue('{"不应静默丢弃":true}')
+        expect(button('保存视觉风格').disabled).toBe(true)
+        await wrapper!.get('#style-constraints').setValue('["真人写实"]')
+        await wrapper!.get('#style-lock').setValue(true)
+        await wrapper!.get('#style-name').setValue('雨夜风格')
+        button('保存视觉风格').click()
+        await flushPromises()
+        const put = fetcher.mock.calls.find(([, init]) => init?.method === 'PUT')!
+        expect(put[0]).toBe('/api/projects/page-test-project/visual-style')
+        expect(JSON.parse(put[1]!.body as string)).toMatchObject({
+            name: '雨夜风格',
+            isLocked: true,
+            hardConstraints: ['真人写实']
+        })
+        expect(button('AI 重新生成风格').disabled).toBe(true)
+        expect(button('保存视觉风格').disabled).toBe(false)
+        expect(fetcher.mock.calls.some(([, init]) => init?.method === 'POST')).toBe(false)
+    })
+
+    it('刷新风格不覆盖未保存草稿，草稿存在时禁止 AI 重生成', async () => {
+        let style = styleFixture()
+        vi.stubGlobal(
+            'fetch',
+            vi.fn<typeof fetch>().mockImplementation(async () => jsonResponse(style))
+        )
+        await mountAssets('style')
+        await wrapper!.get('#style-prompt').setValue('人工草稿')
+        style = styleFixture({ prompt: '后台新内容' })
+        button('刷新风格').click()
+        await flushPromises()
+        expect((wrapper!.get('#style-prompt').element as HTMLTextAreaElement).value).toBe('人工草稿')
+        expect(button('AI 重新生成风格').disabled).toBe(true)
+    })
+
+    it('风格接口 404 显示错误而不是当作空风格，禁止写入', async () => {
+        vi.stubGlobal(
+            'fetch',
+            vi.fn<typeof fetch>().mockResolvedValue(new Response('{"message":"路由不存在"}', { status: 404 }))
+        )
+        await mountAssets('style')
+        expect(wrapper!.text()).toContain('路由不存在')
+        expect(button('AI 生成风格').disabled).toBe(true)
+        expect(wrapper!.find('#style-prompt').exists()).toBe(false)
+    })
+
+    it('风格图启停用 PUT，移除经确认用 DELETE，均不调用生图接口', async () => {
+        let image = styleImageFixture()
+        let removed = false
+        const fetcher = vi.fn<typeof fetch>().mockImplementation(async (url, init) => {
+            if (init?.method === 'PUT') {
+                image = { ...image, enabled: false }
+                return jsonResponse(image)
+            }
+            if (init?.method === 'DELETE') {
+                removed = true
+                return jsonResponse(image)
+            }
+            expect(String(url)).toContain('/visual-style')
+            return jsonResponse(styleFixture({ images: removed ? [] : [image] }))
+        })
+        vi.stubGlobal('fetch', fetcher)
+        await mountAssets('style')
+        button('停用').click()
+        await flushPromises()
+        const put = fetcher.mock.calls.find(([, init]) => init?.method === 'PUT')!
+        expect(put[0]).toBe('/api/projects/page-test-project/visual-style/images/style-image-1/enabled')
+        expect(JSON.parse(put[1]!.body as string)).toEqual({ enabled: false })
+        button('移除').click()
+        await flushPromises()
+        expect(fetcher.mock.calls.some(([, init]) => init?.method === 'DELETE')).toBe(false)
+        button('确认移除').click()
+        await flushPromises()
+        expect(fetcher.mock.calls.find(([, init]) => init?.method === 'DELETE')?.[0]).toBe(
+            '/api/projects/page-test-project/visual-style/images/style-image-1'
+        )
+        expect(wrapper!.text()).toContain('尚无风格参考图')
+    })
+
+    it('主体目录按正式 Subject ID 去重，无 Identity 时不请求会报错的图片接口', async () => {
+        const form = formFixture()
+        const fetcher = vi.fn<typeof fetch>().mockImplementation(async url => {
+            if (String(url).endsWith('/subject-forms'))
+                return jsonResponse([form, { ...form, id: 'form-db-2', images: [], name: '雨夜造型' }])
+            if (String(url).endsWith('/visual-style')) return jsonResponse(styleFixture())
+            return jsonResponse(null)
+        })
+        vi.stubGlobal('fetch', fetcher)
+        await mountAssets('identity')
+        expect(wrapper!.findAll('.identity-subject-item')).toHaveLength(1)
+        expect(wrapper!.text()).toContain('2 个形态')
+        expect(fetcher.mock.calls.some(([url]) => String(url).endsWith('/subjects/subject-db-1/identity'))).toBe(true)
+        expect(fetcher.mock.calls.some(([url]) => String(url).endsWith('/identity/images'))).toBe(false)
+        expect(button('生成身份参考图').disabled).toBe(true)
+        expect(button('AI 生成身份').disabled).toBe(false)
+    })
+
+    it('锁定阻止身份 AI 覆盖但不阻止人工编辑和生图，生图默认省略 referenceImageId', async () => {
+        const fetcher = vi.fn<typeof fetch>().mockImplementation(async (url, init) => {
+            if (String(url).endsWith('/subject-forms')) return jsonResponse([formFixture()])
+            if (String(url).endsWith('/visual-style')) return jsonResponse(styleFixture({ isLocked: true }))
+            if (init?.method === 'POST') return jsonResponse(identityImageFixture())
+            if (String(url).endsWith('/identity/images')) return jsonResponse([identityImageFixture()])
+            return jsonResponse(identityFixture({ isLocked: true }))
+        })
+        vi.stubGlobal('fetch', fetcher)
+        await mountAssets('identity')
+        expect(button('AI 重新生成身份').disabled).toBe(true)
+        expect(button('保存主体身份').disabled).toBe(false)
+        button('生成身份参考图').click()
+        await flushPromises()
+        expect(button('确认生成身份图').disabled).toBe(true)
+        document.querySelector<HTMLInputElement>('#identity-image-cost')!.click()
+        await flushPromises()
+        button('确认生成身份图').click()
+        await flushPromises()
+        const post = fetcher.mock.calls.find(([, init]) => init?.method === 'POST')!
+        expect(post[0]).toBe('/api/subjects/subject-db-1/identity/images')
+        expect(JSON.parse(post[1]!.body as string)).toEqual({ provider: 'seedream', viewType: 'primary' })
+    })
+
+    it('无项目风格时禁止生成身份和图片，但仍可人工保存身份', async () => {
+        vi.stubGlobal(
+            'fetch',
+            vi.fn<typeof fetch>().mockImplementation(async url => {
+                if (String(url).endsWith('/subject-forms')) return jsonResponse([formFixture()])
+                if (String(url).endsWith('/visual-style')) return jsonResponse(null)
+                if (String(url).endsWith('/identity/images')) return jsonResponse([])
+                return jsonResponse(identityFixture())
+            })
+        )
+        await mountAssets('identity')
+        expect(button('AI 重新生成身份').disabled).toBe(true)
+        expect(button('生成身份参考图').disabled).toBe(true)
+        expect(button('保存主体身份').disabled).toBe(false)
+    })
+
+    it('身份批量生成区分锁定跳过和部分失败，不将 HTTP 200 当作全部成功', async () => {
+        const fetcher = vi.fn<typeof fetch>().mockImplementation(async (url, init) => {
+            if (init?.method === 'POST')
+                return jsonResponse({
+                    total: 3,
+                    targetCount: 2,
+                    generated: 1,
+                    skipped: 1,
+                    skippedLocked: 1,
+                    failed: 1,
+                    failures: [{ subjectId: 'failed-db', subjectRef: '@CH0003', error: '文本模型超时' }]
+                })
+            if (String(url).endsWith('/subject-forms')) return jsonResponse([formFixture()])
+            if (String(url).endsWith('/visual-style')) return jsonResponse(styleFixture())
+            return jsonResponse(null)
+        })
+        vi.stubGlobal('fetch', fetcher)
+        await mountAssets('identity')
+        await confirmGeneration('补齐项目身份文本')
+        const post = fetcher.mock.calls.find(([, init]) => init?.method === 'POST')!
+        expect(post[0]).toBe('/api/projects/page-test-project/subject-identities/generate')
+        expect(JSON.parse(post[1]!.body as string)).toEqual({ force: false, concurrency: 3 })
+        expect(wrapper!.text()).toContain('部分主体生成失败')
+        expect(wrapper!.text()).toContain('含锁定 1')
+        expect(wrapper!.text()).toContain('文本模型超时')
+    })
+
+    it('身份候选母版切换需确认，只发送正确 Subject ID 的 anchor PUT', async () => {
+        let selected = false
+        const fetcher = vi.fn<typeof fetch>().mockImplementation(async (url, init) => {
+            if (String(url).endsWith('/subject-forms')) return jsonResponse([formFixture()])
+            if (String(url).endsWith('/visual-style')) return jsonResponse(styleFixture())
+            if (init?.method === 'PUT') {
+                selected = true
+                return jsonResponse(identityImageFixture({ id: 'candidate' }))
+            }
+            if (String(url).endsWith('/identity/images'))
+                return jsonResponse([identityImageFixture({ id: 'candidate', enabled: selected, isAnchor: selected })])
+            return jsonResponse(identityFixture())
+        })
+        vi.stubGlobal('fetch', fetcher)
+        await mountAssets('identity')
+        button('设为身份母版').click()
+        await flushPromises()
+        expect(fetcher.mock.calls.some(([, init]) => init?.method === 'PUT')).toBe(false)
+        button('确认切换身份母版').click()
+        await flushPromises()
+        expect(fetcher.mock.calls.find(([, init]) => init?.method === 'PUT')?.[0]).toBe(
+            '/api/subjects/subject-db-1/identity/images/candidate/anchor'
+        )
+        expect(wrapper!.text()).toContain('当前身份母版')
+    })
+
+    it('切换主体会取消旧查询，迟到的身份结果不覆盖当前主体', async () => {
+        let finish!: (value: Response) => void
+        const form = formFixture()
+        const second = {
+            ...form,
+            id: 'form-2',
+            subjectId: 'subject-2',
+            images: [],
+            subject: { ...form.subject, id: 'subject-2', name: '陆川', ref: '@CH0002' }
+        }
+        const fetcher = vi.fn<typeof fetch>().mockImplementation(async url => {
+            const path = String(url)
+            if (path.endsWith('/subject-forms')) return jsonResponse([form, second])
+            if (path.endsWith('/visual-style')) return jsonResponse(styleFixture())
+            if (path.endsWith('/subject-db-1/identity'))
+                return new Promise(resolve => {
+                    finish = resolve
+                })
+            if (path.endsWith('/identity/images')) return jsonResponse([])
+            return jsonResponse(
+                identityFixture({ id: 'identity-2', subjectId: 'subject-2', description: '陆川的稳定身份' })
+            )
+        })
+        vi.stubGlobal('fetch', fetcher)
+        await mountAssets('identity')
+        await wrapper!.findAll('.identity-subject-item')[1]!.trigger('click')
+        await flushPromises()
+        expect(
+            fetcher.mock.calls.find(([url]) => String(url).endsWith('/subject-db-1/identity'))?.[1]?.signal?.aborted
+        ).toBe(true)
+        finish(jsonResponse(identityFixture()))
+        await flushPromises()
+        expect((wrapper!.get('#identity-description').element as HTMLTextAreaElement).value).toBe('陆川的稳定身份')
+    })
+
+    it('生成失败不自动重发，后端正在拆解时禁止身份写操作', async () => {
+        const fetcher = vi.fn<typeof fetch>().mockImplementation(async (url, init) => {
+            if (String(url).endsWith('/subject-forms')) return jsonResponse([formFixture()])
+            if (String(url).endsWith('/visual-style')) return jsonResponse(styleFixture())
+            if (String(url).endsWith('/identity/images')) return jsonResponse([])
+            if (init?.method === 'POST') throw new Error('connection lost')
+            return jsonResponse(identityFixture())
+        })
+        vi.stubGlobal('fetch', fetcher)
+        const provided = await mountAssets('identity')
+        await confirmGeneration('AI 重新生成身份')
+        expect(fetcher.mock.calls.filter(([, init]) => init?.method === 'POST')).toHaveLength(1)
+        expect(getOperation(fixture.id).error).toContain('无法连接后端')
+        provided.data.value!.project = { ...fixture, status: 'generating' }
+        await flushPromises()
+        expect(button('保存主体身份').disabled).toBe(true)
+        expect(button('生成身份参考图').disabled).toBe(true)
+    })
+})
+
 describe('工作台页面交互', () => {
     it('形态图库自动显示后端已有图片，不需要再次生图；筛选只改变显示', async () => {
         const fetcher = vi

src/features/workflows/useProjectMutationGuard.ts

@@ -0,0 +1,21 @@
+import { computed } from 'vue'
+import { useProjectContext } from '../projects/context'
+import { getOperation } from './operations'
+import { workflowCheckpoints } from './selectors'
+
+/** 资产编辑沿用项目级互斥；剧本生成或拆解期间不修改其下游资产。 */
+export function useProjectMutationGuard() {
+    const context = useProjectContext()
+    const projectId = computed(() => context.project.value?.id ?? '')
+    const operation = computed(() => getOperation(projectId.value))
+    const blocked = computed(
+        () =>
+            !projectId.value ||
+            !!context.error.value ||
+            operation.value.pending ||
+            context.project.value?.status === 'generating' ||
+            workflowCheckpoints(context.checkpoints.value, 'breakdown').at(-1)?.state.workflowExecution?.status ===
+                'running'
+    )
+    return { projectId, operation, blocked }
+}

src/lib/http.ts

@@ -12,7 +12,7 @@ export class ApiError extends Error {
 
 /** 请求选项；长工作流显式关闭超时，不自动重试任何 POST。 */
 export interface RequestOptions {
-    method?: 'GET' | 'POST' | 'PUT'
+    method?: 'GET' | 'POST' | 'PUT' | 'DELETE'
     body?: unknown
     signal?: AbortSignal
     timeoutMs?: number

src/router/index.ts

@@ -25,6 +25,16 @@ export const router = createRouter({
                     component: () => import('../features/breakdown/BreakdownPage.vue'),
                     meta: { title: '剧本拆解' }
                 },
+                {
+                    path: 'visual-style',
+                    component: () => import('../features/visual-style/VisualStylePage.vue'),
+                    meta: { title: '视觉风格' }
+                },
+                {
+                    path: 'subject-identity',
+                    component: () => import('../features/subject-identity/SubjectIdentityPage.vue'),
+                    meta: { title: '主体身份' }
+                },
                 {
                     path: 'subject-images',
                     component: () => import('../features/subject-images/SubjectImagesPage.vue'),

src/styles.css

@@ -62,6 +62,40 @@
 }
 
 @layer components {
+    .identity-workspace {
+        display: grid;
+        grid-template-columns: 240px minmax(0, 1fr);
+        align-items: start;
+        gap: 20px;
+    }
+    .identity-subject-list {
+        max-height: 560px;
+        overflow-y: auto;
+    }
+    .identity-subject-item {
+        display: block;
+        width: 100%;
+        text-align: left;
+        padding: 12px;
+        border-radius: 4px;
+        font-size: 12px;
+        overflow-wrap: anywhere;
+    }
+    .identity-subject-item:hover {
+        background: #f6f7f3;
+    }
+    .identity-subject-item.selected {
+        background: #f1e8df;
+        color: var(--color-accent);
+    }
+    @media (max-width: 900px) {
+        .identity-workspace {
+            grid-template-columns: minmax(0, 1fr);
+        }
+        .identity-subject-list {
+            max-height: 220px;
+        }
+    }
     .app-shell {
         display: grid;
         grid-template-columns: 216px minmax(0, 1fr);