update

UI测试手册.md

@@ -1,6 +1,6 @@
 # 第 1~4 周功能 · 界面人工测试手册
 
-> 用途：你按这份手册在浏览器里点一遍，验证第 1~4 周的功能（登录、租户/项目、资源 ACL、审计日志、审批中心、写操作自动审计与审批拦截）。
+> 用途：你按这份手册在浏览器里点一遍，验证第 1~4 周的功能（登录、租户/项目、资源 ACL、审计日志、审批中心、数据处理 C 模块、写操作自动审计与审批拦截）。
 > 功能代码层均已联调通过（含此前修复的 `audit.ts`/`approval.ts` 双重解包、审计接口 `/system` 前缀、审批模块 `include_router` 启动崩溃）。下面是给你的人工回归步骤。
 
 ---
@@ -276,7 +276,77 @@ npm run dev
 
 ---
 
-## 6. 验收清单（打勾）
+## 6. 数据处理（C 模块 · 第 2 周）
+
+> 入口：左侧菜单「数据治理 → 数据处理」（`/data-process`），对应后端 `/modelTF/data-process/*`。该模块为本次合并从 `yg_ft1` 并入的可运行子系统（后端 `app/modules/data_process/` + 前端 `views/data-process/`）。
+> 创建任务为**六步向导**：`创建任务 → 大模型选择 → 上传文件 → 数据预览 → 开始生成 → 结果编辑与保存`。后端启动时会自动建表（`002_data_process.sql`），无需手动迁移。
+
+### 6.1 列表与入口
+1. 左侧菜单「数据治理 → 数据处理」
+2. 预期：进入列表页 `/data-process`，展示已有数据处理任务（空状态显示「暂无数据」属正常，不要当成 bug）
+3. 点右上「新建数据处理」进入创建向导
+
+### 6.2 创建任务（六步向导）
+
+**步骤 1 · 创建任务**
+1. 填任务名称（如 `test-dp-manual`）、描述（可选）
+2. 选处理类型：
+   - `结构化数据`（structured）：按行生成 QA 对
+   - `非结构化数据`（unstructured）：文档切片（chunk 方法/大小/重叠/保留表格代码块等）
+   - `外部数据源`（external）：接 PostgreSQL 等外部库
+3. 配置处理选项（预处理、语义增强、数据集切分比例、输出类型、温度等，保持默认即可）
+4. 点「继续：选择大模型」
+
+**步骤 2 · 大模型选择**
+1. 从模型下拉选一个生成模型（需「模型管理」里已登记可用模型；若下拉为空，先到「模型管理」登记一个基座/API 模型）
+2. 设输出要求（生成模型、提示词、输出类型、reasoning、温度、max_tokens、质量过滤开关等）
+3. 点「继续：上传文件」
+
+**步骤 3 · 上传文件**
+- 结构化/非结构化：点「上传」选本地文件（jsonl/csv/pdf/docx/xlsx/txt 等，按处理类型校验扩展名）；或点「使用样例文件」快速载入示例 `finance_qa.jsonl`
+- 外部数据源：填连接（postgresql URL、认证模式、只读 `SELECT` 查询），点「测试连接」→ 联通后「拉取数据」
+1. 选/拉取至少一个源文件，等待上传完成（状态变 `ready`）
+2. 点「继续：数据预览」，系统自动按配置**切分**（进度条，可能耗时；失败会提示原因，可重试）
+
+**步骤 4 · 数据预览**
+1. 切分完成后进入预览，左侧为源文件 / 分片列表，右侧为预览内容
+2. 可编辑预览条目内容、新增 / 删除条目、还原为原文（不影响源文件）
+3. 切换不同源文件核对切分结果
+4. 点「确认预览并继续」
+
+**步骤 5 · 开始生成**
+1. 进入生成页，显示任务摘要（任务名、处理类型、文件、预览条目数、修改条数）
+2. 点「开始生成」启动处理（调用所选大模型）
+3. 等待生成完成：状态 `running → completed`（失败显示原因，可「重新生成」）
+4. 生成成功后点「查看生成结果」
+
+**步骤 6 · 结果编辑与保存**
+1. 进入结果编辑页，逐条检查结果（问题 / 答案等），可编辑字段、单条或批量「重新生成」
+2. 校验通过后点「保存任务」
+3. 预期：提示「生成结果已确认」，跳转到任务详情页 `/data-process/<任务ID>`
+
+### 6.3 任务详情
+1. 列表点目标任务「详情」（或保存后自动进入）
+2. 预期：详情页展示任务信息、处理配置、源文件、预览条目、生成结果
+3. 支持「重新生成」（`/data-process/:id/regenerate`）与「任务进度」（`/data-process/:id/workflow`）两个子页，均复用创建向导
+
+### 6.4 重新生成（可选）
+1. 详情页或列表进入「重新生成」子页
+2. 调整配置（保持原处理类型），保存后按新配置重新切分 / 生成；原生成结果与已发布数据在点击「开始生成」前保持不变
+3. 验证：提交重新生成后，进度可在「任务进度」页查看
+
+### 6.5 发布数据集（第 2 周验收点）
+1. 结果确认后，在详情页点「发布数据集」（对应 `POST /modelTF/data-process/{id}/publish`）
+2. 填目标数据集元信息（名称 / 项目归属等），确认发布
+3. 预期：生成数据集记录，可在「数据集管理」看到该发布数据集，来源链路保留
+
+### 6.6 删除任务（可选）
+1. 列表行点「删除」，二次确认后删除
+2. 预期：列表不再显示该任务
+
+---
+
+## 7. 验收清单（打勾）
 
 **第 1~2 周**
 - [ ] 1. 登录成功，进仪表盘，显示 `admin`
@@ -287,6 +357,14 @@ npm run dev
 - [ ] 6. 归档：点后状态变 `archived`
 - [ ] 7. 用户设置：创建用户（账号/显示名/初始密码/角色/状态/页面权限均可设，列表出现新账号）、重置密码（自定义生效 / 留空回退 `platform123` / `admin` 被拒）、启用禁用账号（禁用后登录失败、启用后恢复）、页面权限精细控制（取消权限后登录即生效）、删除账号（二次确认后列表移除、登录失败）
 
+**第 2 周（数据处理 · C 模块）**
+- [ ] 16. 数据处理入口可见（数据治理 → 数据处理），列表 / 空状态正常
+- [ ] 17. 创建向导六步可走通：创建任务 → 选模型 → 上传/拉取源文件 → 切分预览（可编辑/增删/还原）→ 生成 → 结果保存并跳详情
+- [ ] 18. 结构 / 非结构化 / 外部数据源三类处理类型均可配置并产出预览条目
+- [ ] 19. 生成结果可编辑、可单条 / 批量重新生成，保存后跳详情
+- [ ] 20. 发布数据集成功，「数据集管理」可见且保留来源链路
+- [ ] 21. 全程浏览器控制台（F12 → Console）在数据处理流程中无红色报错
+
 **第 3 周（审计）**
 - [ ] 8. 审计日志：按 动作/操作人/项目 过滤均能返回正确结果
 - [ ] 9. 审计日志：导出 CSV 成功，列与内容正确
@@ -301,19 +379,19 @@ npm run dev
 
 ---
 
-## 7. 我自测已覆盖（你不用重复，除非想验证）
+## 8. 我自测已覆盖（你不用重复，除非想验证）
 
 - 后端真实导入：`import app.main` → `IMPORT_OK`，启动崩溃已修复（`approval/__init__.py` 补 re-export `router`）。
 - 前端 `type-check` 全绿：`audit.ts`/`approval.ts` 双重解包已改 `get/post`；审计接口已加 `/system` 前缀（`/system/audit-logs`、`/system/audit-logs/export`）。
 - 接口链路已用真实代码核对：审计查询/导出（`system`）、审批模板/实例/逐步决策（`approvals`）、项目写操作自动 `record_audit` 与 `_require_no_pending_approval` 拦截均按上述行为实现。
 
-## 8. 已知非 bug / 注意事项（仅供参考）
+## 9. 已知非 bug / 注意事项（仅供参考）
 
 1. 前端由你自己在 **Windows 终端** 跑 `npm run dev` 启动（默认 **16801**）。不要从 WSL 终端启动（会慢 8 倍）。
 2. 「归档项目」当前是**直接执行无确认弹窗**——功能正确，建议后续补个二次确认，避免误操作。
 4. 控制台偶见的 `ERR_ABORTED` 是导航时浏览器正常中止旧 CSS 请求，无害；Google Fonts 外网字体加载失败不影响功能。
 5. 审计「操作人」列 = 登录 token（当前登录用户标识），由前端 `Authorization: Bearer <token>` 透传，非真实姓名。
 
-## 9. 清理测试数据（可选）
+## 10. 清理测试数据（可选）
 
 手动建的 `test-tenant-manual` / `test-project-manual`、审批实例/模板可在对应列表里删除，或告诉我帮你清库（后端连 PostgreSQL，`PlatformStore` 启动时自动建表）。

backend/app/api/v1/router.py

@@ -8,6 +8,9 @@ from app.modules.tenant import router as tenant_router
 from app.modules.project import router as project_router
 from app.modules.resource import router as resource_router
 from app.modules.approval import router as approval_router
+from app.api.v1.endpoints.data_process import router as data_process_router
+
+import logging
 
 api_router = APIRouter()
 api_router.include_router(health_router, tags=["health"])
@@ -18,4 +21,11 @@ api_router.include_router(tenant_router, tags=["tenant"])
 api_router.include_router(project_router, tags=["project"])
 api_router.include_router(resource_router, tags=["resource"])
 api_router.include_router(approval_router, tags=["approval"])
+api_router.include_router(data_process_router, tags=["data_process"])
+
+try:
+    from app.modules.data_process.store import DataProcessStore
+    DataProcessStore().ensure_schema()
+except Exception as exc:  # best-effort at startup; feature degrades if DB unavailable
+    logging.getLogger(__name__).warning("data_process schema ensure failed: %s", exc)
 

backend/app/core/config.py

@@ -23,7 +23,7 @@ class Settings:
     app_env: str = os.getenv("APP_ENV", "local")
     route_prefix: str = os.getenv("MODELTF_ROUTE_PREFIX", "/modelTF")
     app_mode: str = os.getenv("APP_MODE", "local")
-    database_url: str = os.getenv("DATABASE_URL", "postgresql+psycopg://yg_ft:change_me@localhost:15432/yg_ft")
+    database_url: str = os.getenv("DATABASE_URL", "postgresql+psycopg://root:8811614287327Leo@www.caoxiaozhu.com:5432/yg_ft")
     cors_allow_origins: list[str] = None  # type: ignore[assignment]
     compute_mode: str = os.getenv("COMPUTE_MODE", "real")
     compute_status_sync_mode: str = os.getenv("COMPUTE_STATUS_SYNC_MODE", "polling")

backend/app/db/session.py

@@ -8,7 +8,7 @@ from sqlalchemy import create_engine
 from sqlalchemy.orm import Session, sessionmaker
 
 
-DATABASE_URL = os.getenv("DATABASE_URL", "postgresql+psycopg://yg_ft:change_me@localhost:15432/yg_ft")
+DATABASE_URL = os.getenv("DATABASE_URL", "postgresql+psycopg://root:8811614287327Leo@www.caoxiaozhu.com:5432/yg_ft")
 
 engine = create_engine(
     DATABASE_URL,

backend/app/db/sql/002_data_process.sql

@@ -0,0 +1,343 @@
+-- Data processing migration.
+--
+-- IMPORTANT: This file is intentionally NOT wired into application startup.
+-- Apply it explicitly in a controlled deployment, or call
+-- DataProcessStore.ensure_schema() from an administrative command.
+
+BEGIN;
+
+-- This migration targets the current runtime schema created by
+-- 001_platform_runtime.sql. Refuse the UUID/JSONB target-design schema instead
+-- of partially altering it with incompatible TEXT foreign keys.
+DO $$
+DECLARE
+    datasets_id_type TEXT;
+BEGIN
+    SELECT format_type(a.atttypid, a.atttypmod)
+      INTO datasets_id_type
+      FROM pg_attribute a
+      JOIN pg_class c ON c.oid = a.attrelid
+      JOIN pg_namespace n ON n.oid = c.relnamespace
+     WHERE n.nspname = current_schema()
+       AND c.relname = 'datasets'
+       AND a.attname = 'id'
+       AND a.attnum > 0
+       AND NOT a.attisdropped;
+    IF datasets_id_type IS NULL THEN
+        RAISE EXCEPTION '002_data_process.sql requires 001_platform_runtime.sql first';
+    END IF;
+    IF datasets_id_type <> 'text' THEN
+        RAISE EXCEPTION
+          '002_data_process.sql supports only the current TEXT runtime schema; found datasets.id type %',
+          datasets_id_type;
+    END IF;
+END $$;
+
+ALTER TABLE datasets ADD COLUMN IF NOT EXISTS source_task_id TEXT;
+ALTER TABLE datasets ADD COLUMN IF NOT EXISTS size_bytes BIGINT NOT NULL DEFAULT 0;
+ALTER TABLE datasets ADD COLUMN IF NOT EXISTS record_count BIGINT NOT NULL DEFAULT 0;
+ALTER TABLE datasets ADD COLUMN IF NOT EXISTS metadata TEXT NOT NULL DEFAULT '{}';
+ALTER TABLE datasets ADD COLUMN IF NOT EXISTS tenant_id TEXT;
+ALTER TABLE datasets ADD COLUMN IF NOT EXISTS project_id TEXT;
+ALTER TABLE datasets ADD COLUMN IF NOT EXISTS owner_id TEXT;
+ALTER TABLE datasets ADD COLUMN IF NOT EXISTS created_by TEXT;
+ALTER TABLE datasets ADD COLUMN IF NOT EXISTS created_at TIMESTAMPTZ NOT NULL DEFAULT now();
+ALTER TABLE datasets ADD COLUMN IF NOT EXISTS updated_at TIMESTAMPTZ NOT NULL DEFAULT now();
+ALTER TABLE datasets ADD COLUMN IF NOT EXISTS deleted_at TIMESTAMPTZ;
+
+ALTER TABLE dataset_files ADD COLUMN IF NOT EXISTS storage_object_id TEXT;
+ALTER TABLE dataset_files ADD COLUMN IF NOT EXISTS current_version_id TEXT;
+ALTER TABLE dataset_files ADD COLUMN IF NOT EXISTS size_bytes BIGINT NOT NULL DEFAULT 0;
+ALTER TABLE dataset_files ADD COLUMN IF NOT EXISTS record_count BIGINT NOT NULL DEFAULT 0;
+ALTER TABLE dataset_files ADD COLUMN IF NOT EXISTS file_format VARCHAR(40);
+ALTER TABLE dataset_files ADD COLUMN IF NOT EXISTS checksum_sha256 CHAR(64);
+ALTER TABLE dataset_files ADD COLUMN IF NOT EXISTS version_no INTEGER NOT NULL DEFAULT 1;
+ALTER TABLE dataset_files ADD COLUMN IF NOT EXISTS source_task_id TEXT;
+ALTER TABLE dataset_files ADD COLUMN IF NOT EXISTS tenant_id TEXT;
+ALTER TABLE dataset_files ADD COLUMN IF NOT EXISTS project_id TEXT;
+ALTER TABLE dataset_files ADD COLUMN IF NOT EXISTS created_by TEXT;
+ALTER TABLE dataset_files ADD COLUMN IF NOT EXISTS metadata TEXT NOT NULL DEFAULT '{}';
+ALTER TABLE dataset_files ADD COLUMN IF NOT EXISTS created_at TIMESTAMPTZ NOT NULL DEFAULT now();
+ALTER TABLE dataset_files ADD COLUMN IF NOT EXISTS updated_at TIMESTAMPTZ NOT NULL DEFAULT now();
+ALTER TABLE dataset_files ADD COLUMN IF NOT EXISTS deleted_at TIMESTAMPTZ;
+
+CREATE TABLE IF NOT EXISTS data_process_tasks (
+    id TEXT PRIMARY KEY,
+    name VARCHAR(150) NOT NULL,
+    description TEXT,
+    status VARCHAR(20) NOT NULL DEFAULT 'pending'
+        CHECK (status IN ('pending', 'running', 'completed', 'failed', 'stopped')),
+    process_type VARCHAR(20) NOT NULL
+        CHECK (process_type IN ('structured', 'unstructured', 'external')),
+    source_dataset_id TEXT REFERENCES datasets(id) ON DELETE SET NULL,
+    output_dataset_id TEXT REFERENCES datasets(id) ON DELETE SET NULL,
+    config TEXT NOT NULL DEFAULT '{}',
+    progress NUMERIC(5,2) NOT NULL DEFAULT 0 CHECK (progress >= 0 AND progress <= 100),
+    input_count BIGINT NOT NULL DEFAULT 0 CHECK (input_count >= 0),
+    output_count BIGINT NOT NULL DEFAULT 0 CHECK (output_count >= 0),
+    filtered_count BIGINT NOT NULL DEFAULT 0 CHECK (filtered_count >= 0),
+    duplicate_count BIGINT NOT NULL DEFAULT 0 CHECK (duplicate_count >= 0),
+    error_count BIGINT NOT NULL DEFAULT 0 CHECK (error_count >= 0),
+    failure_reason TEXT,
+    generation_run_id TEXT,
+    results_confirmed BOOLEAN NOT NULL DEFAULT TRUE,
+    workflow_step VARCHAR(20) NOT NULL DEFAULT 'create'
+        CHECK (workflow_step IN ('create', 'model', 'upload', 'preview', 'generate', 'results')),
+    preview_status VARCHAR(20) NOT NULL DEFAULT 'idle'
+        CHECK (preview_status IN ('idle', 'queued', 'running', 'completed', 'failed', 'cancelled')),
+    preview_progress NUMERIC(5,2) NOT NULL DEFAULT 0
+        CHECK (preview_progress >= 0 AND preview_progress <= 100),
+    preview_run_id TEXT,
+    preview_failure_reason TEXT,
+    preview_total_files INTEGER NOT NULL DEFAULT 0 CHECK (preview_total_files >= 0),
+    preview_completed_files INTEGER NOT NULL DEFAULT 0 CHECK (preview_completed_files >= 0),
+    tenant_id TEXT,
+    project_id TEXT,
+    owner_id TEXT,
+    approval_status VARCHAR(30) NOT NULL DEFAULT 'not_required',
+    created_by TEXT,
+    updated_by TEXT,
+    deleted_by TEXT,
+    started_at TIMESTAMPTZ,
+    completed_at TIMESTAMPTZ,
+    created_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+    updated_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+    deleted_at TIMESTAMPTZ
+);
+
+ALTER TABLE data_process_tasks ADD COLUMN IF NOT EXISTS generation_run_id TEXT;
+-- 历史任务在引入六步确认流程前已经完成审核，默认保留为已确认；
+-- 新任务由创建接口显式写入 FALSE，并在第六步确认后转为 TRUE。
+ALTER TABLE data_process_tasks
+    ADD COLUMN IF NOT EXISTS results_confirmed BOOLEAN NOT NULL DEFAULT TRUE;
+UPDATE data_process_tasks
+SET results_confirmed=FALSE
+WHERE status <> 'completed' AND results_confirmed=TRUE;
+
+-- 先以可空列接入旧库，才能只回填历史行；随后再收紧默认值与约束。
+ALTER TABLE data_process_tasks ADD COLUMN IF NOT EXISTS workflow_step VARCHAR(20);
+ALTER TABLE data_process_tasks ADD COLUMN IF NOT EXISTS preview_status VARCHAR(20);
+ALTER TABLE data_process_tasks ADD COLUMN IF NOT EXISTS preview_progress NUMERIC(5,2);
+ALTER TABLE data_process_tasks ADD COLUMN IF NOT EXISTS preview_run_id TEXT;
+ALTER TABLE data_process_tasks ADD COLUMN IF NOT EXISTS preview_failure_reason TEXT;
+ALTER TABLE data_process_tasks ADD COLUMN IF NOT EXISTS preview_total_files INTEGER;
+ALTER TABLE data_process_tasks ADD COLUMN IF NOT EXISTS preview_completed_files INTEGER;
+
+CREATE TEMP TABLE data_process_workflow_backfill_ids ON COMMIT DROP AS
+SELECT id FROM data_process_tasks WHERE workflow_step IS NULL;
+
+UPDATE data_process_tasks task
+SET workflow_step = CASE
+    WHEN task.status IN ('running', 'failed', 'stopped') THEN 'generate'
+    WHEN task.status = 'completed' AND task.results_confirmed=FALSE THEN 'generate'
+    WHEN task.status = 'completed' THEN 'results'
+    ELSE 'create'
+END
+WHERE task.workflow_step IS NULL;
+UPDATE data_process_tasks
+SET preview_status='idle', preview_progress=0,
+    preview_total_files=0, preview_completed_files=0
+WHERE preview_status IS NULL OR preview_progress IS NULL
+   OR preview_total_files IS NULL OR preview_completed_files IS NULL;
+
+ALTER TABLE data_process_tasks ALTER COLUMN workflow_step SET DEFAULT 'create';
+ALTER TABLE data_process_tasks ALTER COLUMN workflow_step SET NOT NULL;
+ALTER TABLE data_process_tasks ALTER COLUMN preview_status SET DEFAULT 'idle';
+ALTER TABLE data_process_tasks ALTER COLUMN preview_status SET NOT NULL;
+ALTER TABLE data_process_tasks ALTER COLUMN preview_progress SET DEFAULT 0;
+ALTER TABLE data_process_tasks ALTER COLUMN preview_progress SET NOT NULL;
+ALTER TABLE data_process_tasks ALTER COLUMN preview_total_files SET DEFAULT 0;
+ALTER TABLE data_process_tasks ALTER COLUMN preview_total_files SET NOT NULL;
+ALTER TABLE data_process_tasks ALTER COLUMN preview_completed_files SET DEFAULT 0;
+ALTER TABLE data_process_tasks ALTER COLUMN preview_completed_files SET NOT NULL;
+
+DO $$
+BEGIN
+    IF NOT EXISTS (
+        SELECT 1 FROM pg_constraint
+        WHERE conrelid='data_process_tasks'::regclass
+          AND conname='ck_data_process_tasks_workflow_step'
+    ) THEN
+        ALTER TABLE data_process_tasks ADD CONSTRAINT ck_data_process_tasks_workflow_step
+            CHECK (workflow_step IN ('create', 'model', 'upload', 'preview', 'generate', 'results'));
+    END IF;
+    IF NOT EXISTS (
+        SELECT 1 FROM pg_constraint
+        WHERE conrelid='data_process_tasks'::regclass
+          AND conname='ck_data_process_tasks_preview_status'
+    ) THEN
+        ALTER TABLE data_process_tasks ADD CONSTRAINT ck_data_process_tasks_preview_status
+            CHECK (preview_status IN ('idle', 'queued', 'running', 'completed', 'failed', 'cancelled'));
+    END IF;
+    IF NOT EXISTS (
+        SELECT 1 FROM pg_constraint
+        WHERE conrelid='data_process_tasks'::regclass
+          AND conname='ck_data_process_tasks_preview_progress'
+    ) THEN
+        ALTER TABLE data_process_tasks ADD CONSTRAINT ck_data_process_tasks_preview_progress
+            CHECK (preview_progress >= 0 AND preview_progress <= 100);
+    END IF;
+    IF NOT EXISTS (
+        SELECT 1 FROM pg_constraint
+        WHERE conrelid='data_process_tasks'::regclass
+          AND conname='ck_data_process_tasks_preview_file_counts'
+    ) THEN
+        ALTER TABLE data_process_tasks ADD CONSTRAINT ck_data_process_tasks_preview_file_counts
+            CHECK (preview_total_files >= 0 AND preview_completed_files >= 0
+                   AND preview_completed_files <= preview_total_files);
+    END IF;
+END $$;
+
+CREATE UNIQUE INDEX IF NOT EXISTS uq_data_process_tasks_name_alive
+    ON data_process_tasks(name) WHERE deleted_at IS NULL;
+CREATE INDEX IF NOT EXISTS idx_data_process_tasks_scope_status
+    ON data_process_tasks(tenant_id, project_id, status, created_at DESC)
+    WHERE deleted_at IS NULL;
+CREATE INDEX IF NOT EXISTS idx_data_process_tasks_creator_created
+    ON data_process_tasks(created_by, created_at DESC) WHERE deleted_at IS NULL;
+
+CREATE TABLE IF NOT EXISTS data_process_source_files (
+    id TEXT PRIMARY KEY,
+    task_id TEXT NOT NULL REFERENCES data_process_tasks(id) ON DELETE CASCADE,
+    storage_object_id TEXT,
+    name TEXT NOT NULL,
+    size_bytes BIGINT NOT NULL DEFAULT 0 CHECK (size_bytes >= 0),
+    record_count BIGINT NOT NULL DEFAULT 0 CHECK (record_count >= 0),
+    file_format VARCHAR(40),
+    checksum_sha256 CHAR(64) NOT NULL,
+    version_no INTEGER NOT NULL DEFAULT 1 CHECK (version_no > 0),
+    content TEXT NOT NULL,
+    content_preview TEXT,
+    metadata TEXT NOT NULL DEFAULT '{}',
+    tenant_id TEXT,
+    project_id TEXT,
+    created_by TEXT,
+    created_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+    updated_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+    deleted_at TIMESTAMPTZ
+);
+
+CREATE INDEX IF NOT EXISTS idx_data_process_source_files_task
+    ON data_process_source_files(task_id, created_at) WHERE deleted_at IS NULL;
+CREATE UNIQUE INDEX IF NOT EXISTS uq_data_process_source_checksum_alive
+    ON data_process_source_files(task_id, checksum_sha256) WHERE deleted_at IS NULL;
+
+CREATE TABLE IF NOT EXISTS data_process_preview_items (
+    id TEXT PRIMARY KEY,
+    task_id TEXT NOT NULL REFERENCES data_process_tasks(id) ON DELETE CASCADE,
+    source_file_id TEXT REFERENCES data_process_source_files(id) ON DELETE CASCADE,
+    original_content TEXT NOT NULL DEFAULT '',
+    edited_content TEXT NOT NULL DEFAULT '',
+    source_start INTEGER CHECK (source_start IS NULL OR source_start >= 0),
+    source_end INTEGER CHECK (source_end IS NULL OR source_end >= 0),
+    source_start_line INTEGER CHECK (source_start_line IS NULL OR source_start_line > 0),
+    source_end_line INTEGER CHECK (source_end_line IS NULL OR source_end_line > 0),
+    token_count INTEGER NOT NULL DEFAULT 0 CHECK (token_count >= 0),
+    status VARCHAR(20) NOT NULL DEFAULT 'original'
+        CHECK (status IN ('original', 'modified', 'manual', 'invalid')),
+    quality_score TEXT NOT NULL DEFAULT '{}',
+    created_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+    updated_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+    CHECK (source_start IS NULL OR source_end IS NULL OR source_end >= source_start),
+    CHECK (source_start_line IS NULL OR source_end_line IS NULL OR source_end_line >= source_start_line)
+);
+
+CREATE INDEX IF NOT EXISTS idx_data_process_preview_task_file
+    ON data_process_preview_items(task_id, source_file_id, created_at);
+
+-- 子表在新库中到这里才存在；只修复本次新增 workflow_step 前的历史任务。
+UPDATE data_process_tasks task
+SET workflow_step = CASE
+    WHEN EXISTS (
+        SELECT 1 FROM data_process_preview_items preview
+        WHERE preview.task_id=task.id
+    ) THEN 'preview'
+    WHEN EXISTS (
+        SELECT 1 FROM data_process_source_files source_file
+        WHERE source_file.task_id=task.id AND source_file.deleted_at IS NULL
+    ) THEN 'upload'
+    ELSE task.workflow_step
+END
+WHERE task.id IN (SELECT id FROM data_process_workflow_backfill_ids)
+  AND task.status='pending';
+
+CREATE TABLE IF NOT EXISTS data_process_results (
+    id TEXT PRIMARY KEY,
+    task_id TEXT NOT NULL REFERENCES data_process_tasks(id) ON DELETE CASCADE,
+    preview_item_id TEXT REFERENCES data_process_preview_items(id) ON DELETE SET NULL,
+    instruction TEXT NOT NULL,
+    input TEXT NOT NULL DEFAULT '',
+    output TEXT NOT NULL,
+    original_instruction TEXT,
+    original_input TEXT,
+    original_output TEXT,
+    status VARCHAR(20) NOT NULL DEFAULT 'valid'
+        CHECK (status IN ('valid', 'modified', 'invalid')),
+    error TEXT,
+    split VARCHAR(20) CHECK (split IS NULL OR split IN ('train', 'validation', 'test')),
+    quality_score TEXT NOT NULL DEFAULT '{}',
+    created_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+    updated_at TIMESTAMPTZ NOT NULL DEFAULT now()
+);
+
+CREATE INDEX IF NOT EXISTS idx_data_process_results_task_status
+    ON data_process_results(task_id, status, id);
+CREATE INDEX IF NOT EXISTS idx_data_process_results_task_split
+    ON data_process_results(task_id, split);
+
+CREATE TABLE IF NOT EXISTS dataset_file_versions (
+    id TEXT PRIMARY KEY,
+    dataset_file_id TEXT NOT NULL REFERENCES dataset_files(id) ON DELETE CASCADE,
+    version_no INTEGER NOT NULL CHECK (version_no > 0),
+    storage_object_id TEXT NOT NULL,
+    content_preview TEXT,
+    description TEXT,
+    base_version_id TEXT REFERENCES dataset_file_versions(id) ON DELETE SET NULL,
+    size_bytes BIGINT NOT NULL DEFAULT 0 CHECK (size_bytes >= 0),
+    record_count BIGINT NOT NULL DEFAULT 0 CHECK (record_count >= 0),
+    checksum_sha256 CHAR(64) NOT NULL,
+    source_task_id TEXT REFERENCES data_process_tasks(id) ON DELETE SET NULL,
+    metadata TEXT NOT NULL DEFAULT '{}',
+    created_by TEXT,
+    created_at TIMESTAMPTZ NOT NULL DEFAULT now()
+);
+
+ALTER TABLE dataset_file_versions ADD COLUMN IF NOT EXISTS source_task_id TEXT;
+ALTER TABLE dataset_file_versions ADD COLUMN IF NOT EXISTS metadata TEXT NOT NULL DEFAULT '{}';
+CREATE UNIQUE INDEX IF NOT EXISTS uq_dataset_file_versions_no_002
+    ON dataset_file_versions(dataset_file_id, version_no);
+CREATE INDEX IF NOT EXISTS idx_dataset_file_versions_source_task_002
+    ON dataset_file_versions(source_task_id) WHERE source_task_id IS NOT NULL;
+
+CREATE TABLE IF NOT EXISTS dataset_records (
+    id TEXT PRIMARY KEY,
+    dataset_id TEXT NOT NULL REFERENCES datasets(id) ON DELETE CASCADE,
+    dataset_file_id TEXT REFERENCES dataset_files(id) ON DELETE CASCADE,
+    version_id TEXT REFERENCES dataset_file_versions(id) ON DELETE CASCADE,
+    line_no INTEGER,
+    split VARCHAR(20) CHECK (split IS NULL OR split IN ('train', 'validation', 'test')),
+    instruction TEXT,
+    input TEXT,
+    output TEXT,
+    raw TEXT NOT NULL DEFAULT '{}',
+    status VARCHAR(20) NOT NULL DEFAULT 'valid'
+        CHECK (status IN ('valid', 'modified', 'invalid')),
+    source_task_id TEXT REFERENCES data_process_tasks(id) ON DELETE SET NULL,
+    source_result_id TEXT REFERENCES data_process_results(id) ON DELETE SET NULL,
+    preview_item_id TEXT REFERENCES data_process_preview_items(id) ON DELETE SET NULL,
+    created_at TIMESTAMPTZ NOT NULL DEFAULT now()
+);
+
+ALTER TABLE dataset_records ADD COLUMN IF NOT EXISTS source_task_id TEXT;
+ALTER TABLE dataset_records ADD COLUMN IF NOT EXISTS source_result_id TEXT;
+ALTER TABLE dataset_records ADD COLUMN IF NOT EXISTS preview_item_id TEXT;
+CREATE INDEX IF NOT EXISTS idx_dataset_records_dataset_002
+    ON dataset_records(dataset_id, id);
+CREATE INDEX IF NOT EXISTS idx_dataset_records_source_task_002
+    ON dataset_records(source_task_id, source_result_id);
+CREATE INDEX IF NOT EXISTS idx_datasets_source_task_002
+    ON datasets(source_task_id) WHERE source_task_id IS NOT NULL;
+CREATE INDEX IF NOT EXISTS idx_dataset_files_source_task_002
+    ON dataset_files(source_task_id) WHERE source_task_id IS NOT NULL;
+
+COMMIT;

backend/app/db/sql/002_governance.sql

@@ -35,6 +35,8 @@ CREATE TABLE IF NOT EXISTS tenants (
     code          TEXT NOT NULL UNIQUE,
     status        TEXT NOT NULL,
     owner_user_id TEXT,
+    quota         TEXT,
+    retention_policy_id TEXT,
     create_time   TEXT NOT NULL
 );
 

backend/app/modules/data_process/constants.py

@@ -0,0 +1,7 @@
+"""数据处理模块的共享限制。"""
+
+MAX_QA_PAIRS_PER_ITEM = 50
+MODEL_GENERATION_BATCH_SIZE = 10
+
+
+__all__ = ["MAX_QA_PAIRS_PER_ITEM", "MODEL_GENERATION_BATCH_SIZE"]

backend/app/modules/data_process/dataset_format.py

@@ -0,0 +1,148 @@
+"""Dataset format validation for Alpaca, ShareGPT, DPO, CPT formats.
+
+Used by the training preflight flow to validate that uploaded dataset files
+conform to the declared format before submitting to the compute node.
+"""
+
+from __future__ import annotations
+
+import json
+from typing import Any
+
+
+def _load_sample(path: str | None, content: str | None = None, max_samples: int = 20) -> list[dict[str, Any]]:
+    """Load up to max_samples records from JSONL file path or raw content string."""
+    try:
+        if content is not None:
+            text = content.strip()
+        elif path:
+            with open(path, "r", encoding="utf-8") as fh:
+                text = fh.read().strip()
+        else:
+            return []
+    except Exception:
+        return []
+
+    if not text:
+        return []
+
+    lines = text.splitlines()[:max_samples]
+    records: list[dict[str, Any]] = []
+    for line in lines:
+        line = line.strip()
+        if not line:
+            continue
+        try:
+            record = json.loads(line)
+        except json.JSONDecodeError:
+            continue
+        if isinstance(record, dict):
+            records.append(record)
+    return records
+
+
+def _check_alpaca(records: list[dict[str, Any]]) -> list[str]:
+    """Validate Alpaca format: requires 'instruction' field."""
+    errors: list[str] = []
+    if not records:
+        errors.append("Alpaca 格式数据集无有效记录")
+        return errors
+    missing_instruction = sum(1 for r in records if not r.get("instruction"))
+    if missing_instruction:
+        errors.append(
+            f"Alpaca 格式要求每条记录包含 instruction 字段，"
+            f"前{len(records)}条中有{missing_instruction}条缺失"
+        )
+    return errors
+
+
+def _check_sharegpt(records: list[dict[str, Any]]) -> list[str]:
+    """Validate ShareGPT format: requires 'messages' (list of dicts with role/content)."""
+    errors: list[str] = []
+    if not records:
+        errors.append("ShareGPT 格式数据集无有效记录")
+        return errors
+    bad = 0
+    for r in records:
+        messages = r.get("messages")
+        if not isinstance(messages, list) or not messages:
+            bad += 1
+            continue
+        for msg in messages:
+            if not isinstance(msg, dict) or "role" not in msg or "content" not in msg:
+                bad += 1
+                break
+    if bad:
+        errors.append(
+            f"ShareGPT 格式要求每条记录包含 messages 列表，"
+            f"每条消息需有 role 和 content 字段，前{len(records)}条中有{bad}条不符合"
+        )
+    return errors
+
+
+def _check_dpo(records: list[dict[str, Any]]) -> list[str]:
+    """Validate DPO format: requires 'chosen' and 'rejected' fields."""
+    errors: list[str] = []
+    if not records:
+        errors.append("DPO 格式数据集无有效记录")
+        return errors
+    missing_chosen = sum(1 for r in records if not r.get("chosen"))
+    missing_rejected = sum(1 for r in records if not r.get("rejected"))
+    if missing_chosen:
+        errors.append(f"DPO 格式要求 chosen 字段，前{len(records)}条中有{missing_chosen}条缺失")
+    if missing_rejected:
+        errors.append(f"DPO 格式要求 rejected 字段，前{len(records)}条中有{missing_rejected}条缺失")
+    return errors
+
+
+def _check_cpt(records: list[dict[str, Any]]) -> list[str]:
+    """Validate CPT format: requires 'text' field, should NOT have instruction/output."""
+    errors: list[str] = []
+    if not records:
+        errors.append("CPT 格式数据集无有效记录")
+        return errors
+    missing_text = sum(1 for r in records if not r.get("text"))
+    has_instruction = sum(1 for r in records if r.get("instruction") or r.get("output"))
+    if missing_text:
+        errors.append(f"CPT 格式要求 text 字段，前{len(records)}条中有{missing_text}条缺失")
+    if has_instruction:
+        errors.append(
+            f"CPT 格式不应包含 instruction/output 字段（疑似 Alpaca 格式），"
+            f"前{len(records)}条中有{has_instruction}条包含此类字段"
+        )
+    return errors
+
+
+FORMAT_VALIDATORS = {
+    "alpaca": _check_alpaca,
+    "alpaca_jsonl": _check_alpaca,
+    "sharegpt": _check_sharegpt,
+    "dpo": _check_dpo,
+    "cpt": _check_cpt,
+    "pt": _check_cpt,
+}
+
+
+def validate_dataset_format(
+    dataset_format: str,
+    content: str | None = None,
+    path: str | None = None,
+    max_samples: int = 20,
+) -> list[str]:
+    """Validate dataset content against expected format.
+
+    Args:
+        dataset_format: One of 'alpaca', 'sharegpt', 'dpo', 'cpt'.
+        content: Raw file content (JSONL text). Mutually exclusive with path.
+        path: File path to read content from.
+        max_samples: Maximum records to sample for validation.
+
+    Returns:
+        List of error messages (empty if valid).
+    """
+    fmt = str(dataset_format).lower().strip()
+    validator = FORMAT_VALIDATORS.get(fmt)
+    if not validator:
+        return [f"不支持的数据集格式: {dataset_format}，支持的格式: {', '.join(sorted(FORMAT_VALIDATORS))}"]
+    records = _load_sample(path=path, content=content, max_samples=max_samples)
+    return validator(records)

backend/app/modules/data_process/document_chunking.py

@@ -0,0 +1,444 @@
+"""基于 Docling 与 LlamaIndex 的文档切分实现。"""
+
+from __future__ import annotations
+
+import os
+import re
+import threading
+import unicodedata
+from dataclasses import dataclass
+from functools import lru_cache
+from io import BytesIO
+from typing import Any, Literal
+
+import tiktoken
+from docling_core.transforms.chunker.hierarchical_chunker import ChunkingSerializerProvider
+from llama_index.core import Document
+from llama_index.core.base.embeddings.base import BaseEmbedding
+from llama_index.core.node_parser import SemanticSplitterNodeParser, SentenceSplitter
+
+from app.modules.data_process.algorithms import normalize_text
+
+ChunkMethod = Literal["layout_hybrid", "semantic", "fixed"]
+
+_PAGE_FURNITURE = re.compile(
+    r"(?m)^\s*(?:第\s*\d+\s*页\s*共\s*\d+\s*页|[-—–]?\s*\d+\s*[/／]\s*\d+\s*[-—–]?)\s*$"
+)
+_COMPACT_CHARACTER = re.compile(r"[\w\u3400-\u4dbf\u4e00-\u9fff]", re.UNICODE)
+_CONVERTER_LOCK = threading.Lock()
+
+
+@dataclass(frozen=True, slots=True)
+class DocumentChunk:
+    """切片正文及其在原文件中的可追溯信息。"""
+
+    original_content: str
+    contextualized_content: str
+    source_start: int | None
+    source_end: int | None
+    source_start_line: int | None
+    source_end_line: int | None
+    token_count: int
+    heading_path: tuple[str, ...] = ()
+    source_pages: tuple[int, ...] = ()
+    doc_item_refs: tuple[str, ...] = ()
+    source_bboxes: tuple[dict[str, Any], ...] = ()
+
+
+def _sentence_chunks(text: str) -> list[str]:
+    """提供稳定的中英文句界，避免 LlamaIndex 默认分词器下载额外资源。"""
+
+    boundary = re.compile(
+        r".*?(?:\n\s*\n|[。！？!?；;](?:[\"'”’）】》]*)|\.(?:\s+|$)|$)",
+        re.DOTALL,
+    )
+    return [part for part in boundary.findall(text) if part]
+
+
+@lru_cache(maxsize=1)
+def _tokenizer() -> tiktoken.Encoding:
+    return tiktoken.get_encoding("cl100k_base")
+
+
+def _text_chunks(
+    text: str,
+    *,
+    chunk_size: int,
+    chunk_overlap: int,
+) -> list[DocumentChunk]:
+    normalized = normalize_text(text)
+    if not normalized:
+        return []
+    splitter = SentenceSplitter(
+        chunk_size=chunk_size,
+        chunk_overlap=chunk_overlap,
+        tokenizer=_tokenizer().encode,
+        chunking_tokenizer_fn=_sentence_chunks,
+        include_metadata=False,
+        include_prev_next_rel=False,
+    )
+    nodes = splitter.get_nodes_from_documents([Document(text=normalized)])
+    return _nodes_to_chunks(nodes, normalized)
+
+
+def chunk_fixed_text(
+    text: str,
+    *,
+    chunk_size: int,
+    chunk_overlap: int,
+) -> list[DocumentChunk]:
+    """使用 LlamaIndex SentenceSplitter 按句界控制固定 Token 长度。"""
+
+    return _text_chunks(text, chunk_size=chunk_size, chunk_overlap=chunk_overlap)
+
+
+@lru_cache(maxsize=1)
+def _semantic_embedding_model() -> BaseEmbedding:
+    # 模型可在部署环境覆盖；默认模型体积较小且适合中英文语义边界判断。
+    from llama_index.embeddings.huggingface import HuggingFaceEmbedding
+
+    return HuggingFaceEmbedding(
+        model_name=os.getenv("DATA_PROCESS_EMBEDDING_MODEL", "BAAI/bge-small-zh-v1.5"),
+        device=os.getenv("DATA_PROCESS_EMBEDDING_DEVICE", "cpu"),
+        trust_remote_code=False,
+    )
+
+
+def chunk_semantic_text(
+    text: str,
+    *,
+    chunk_size: int,
+    chunk_overlap: int,
+    breakpoint_percentile_threshold: int,
+    embed_model: BaseEmbedding | None = None,
+) -> list[DocumentChunk]:
+    """使用 LlamaIndex SemanticSplitter 识别主题跳变，再限制最大长度。"""
+
+    normalized = normalize_text(text)
+    if not normalized:
+        return []
+    splitter = SemanticSplitterNodeParser.from_defaults(
+        embed_model=embed_model or _semantic_embedding_model(),
+        breakpoint_percentile_threshold=breakpoint_percentile_threshold,
+        buffer_size=1,
+        sentence_splitter=_sentence_chunks,
+        include_metadata=False,
+        include_prev_next_rel=False,
+    )
+    semantic_nodes = splitter.get_nodes_from_documents([Document(text=normalized)])
+    result: list[DocumentChunk] = []
+    search_from = 0
+    for node in semantic_nodes:
+        content = node.get_content().strip()
+        if not content:
+            continue
+        start = _locate_text(normalized, content, search_from)
+        if start is None:
+            start = _locate_text(normalized, content, 0)
+        if start is None:
+            continue
+        if len(_tokenizer().encode(content)) <= chunk_size:
+            result.append(_make_text_chunk(normalized, start, start + len(content)))
+        else:
+            for child in _text_chunks(
+                content,
+                chunk_size=chunk_size,
+                chunk_overlap=chunk_overlap,
+            ):
+                if child.source_start is None or child.source_end is None:
+                    continue
+                result.append(
+                    _make_text_chunk(
+                        normalized,
+                        start + child.source_start,
+                        start + child.source_end,
+                    )
+                )
+        search_from = start + len(content)
+    return result
+
+
+def _nodes_to_chunks(nodes: list[Any], source_text: str) -> list[DocumentChunk]:
+    chunks: list[DocumentChunk] = []
+    search_from = 0
+    for node in nodes:
+        content = node.get_content().strip()
+        if not content:
+            continue
+        raw_start = getattr(node, "start_char_idx", None)
+        raw_end = getattr(node, "end_char_idx", None)
+        if (
+            isinstance(raw_start, int)
+            and isinstance(raw_end, int)
+            and source_text[raw_start:raw_end].strip() == content
+        ):
+            start = raw_start + len(source_text[raw_start:raw_end]) - len(source_text[raw_start:raw_end].lstrip())
+        else:
+            start = _locate_text(source_text, content, search_from)
+            if start is None:
+                start = _locate_text(source_text, content, 0)
+        if start is None:
+            continue
+        end = start + len(content)
+        chunks.append(_make_text_chunk(source_text, start, end))
+        search_from = max(search_from, end)
+    return chunks
+
+
+def _locate_text(source: str, content: str, start: int) -> int | None:
+    position = source.find(content, start)
+    return position if position >= 0 else None
+
+
+def _make_text_chunk(source: str, start: int, end: int) -> DocumentChunk:
+    content = source[start:end]
+    return DocumentChunk(
+        original_content=content,
+        contextualized_content=content,
+        source_start=start,
+        source_end=end,
+        source_start_line=source.count("\n", 0, start) + 1,
+        source_end_line=source.count("\n", 0, max(start, end - 1)) + 1,
+        token_count=len(_tokenizer().encode(content)),
+    )
+
+
+@lru_cache(maxsize=1)
+def _document_converter():
+    from docling.document_converter import DocumentConverter
+
+    return DocumentConverter()
+
+
+class _MarkdownSerializerProvider(ChunkingSerializerProvider):
+    def get_serializer(self, doc: Any):
+        from docling_core.transforms.chunker.hierarchical_chunker import ChunkingDocSerializer
+        from docling_core.transforms.serializer.markdown import (
+            MarkdownParams,
+            MarkdownTableSerializer,
+        )
+        from docling_core.types.doc import DocItemLabel
+
+        excluded = {
+            DocItemLabel.DOCUMENT_INDEX,
+            DocItemLabel.PAGE_HEADER,
+            DocItemLabel.PAGE_FOOTER,
+        }
+        return ChunkingDocSerializer(
+            doc=doc,
+            table_serializer=MarkdownTableSerializer(),
+            params=MarkdownParams(
+                labels=set(DocItemLabel) - excluded,
+                compact_tables=True,
+                image_placeholder="",
+                escape_html=False,
+                escape_underscores=False,
+            ),
+        )
+
+
+def _clean_layout_text(value: str) -> str:
+    return normalize_text(_PAGE_FURNITURE.sub("", value)).strip()
+
+
+def _compact_with_offsets(value: str) -> tuple[str, list[int]]:
+    compact: list[str] = []
+    offsets: list[int] = []
+    for index, character in enumerate(unicodedata.normalize("NFKC", value)):
+        if _COMPACT_CHARACTER.fullmatch(character):
+            compact.append(character.casefold())
+            offsets.append(index)
+    return "".join(compact), offsets
+
+
+def _project_layout_span(
+    source_text: str,
+    content: str,
+    *,
+    compact_source: str,
+    source_offsets: list[int],
+    compact_start: int,
+) -> tuple[int | None, int | None, int]:
+    compact_content, _ = _compact_with_offsets(content)
+    if len(compact_content) < 4:
+        return None, None, compact_start
+    position = compact_source.find(compact_content, compact_start)
+    if position < 0:
+        position = compact_source.find(compact_content)
+    if position < 0:
+        return None, None, compact_start
+    start = source_offsets[position]
+    end = source_offsets[position + len(compact_content) - 1] + 1
+    while start > 0 and source_text[start - 1] not in "\r\n":
+        start -= 1
+    while end < len(source_text) and source_text[end] not in "\r\n":
+        end += 1
+    return start, end, position + len(compact_content)
+
+
+def chunk_layout_document(
+    raw: bytes,
+    *,
+    filename: str,
+    source_text: str,
+    chunk_size: int,
+) -> list[DocumentChunk]:
+    """使用 Docling HybridChunker 按版面层级、列表与表格边界切分。"""
+
+    from docling.chunking import HybridChunker
+    from docling.datamodel.base_models import DocumentStream
+    from docling.exceptions import BaseError as DoclingError
+    from docling_core.transforms.chunker.tokenizer.openai import OpenAITokenizer
+    from docling_core.types.doc import DocItemLabel
+
+    try:
+        with _CONVERTER_LOCK:
+            conversion = _document_converter().convert(
+                DocumentStream(name=filename, stream=BytesIO(raw))
+            )
+    except DoclingError as exc:
+        raise ValueError(f"文档版面解析失败: {exc}") from exc
+    chunker = HybridChunker(
+        tokenizer=OpenAITokenizer(tokenizer=_tokenizer(), max_tokens=chunk_size),
+        serializer_provider=_MarkdownSerializerProvider(),
+        merge_peers=True,
+        repeat_table_header=True,
+    )
+    compact_source, source_offsets = _compact_with_offsets(source_text)
+    compact_start = 0
+    result: list[DocumentChunk] = []
+    excluded = {
+        DocItemLabel.DOCUMENT_INDEX,
+        DocItemLabel.PAGE_HEADER,
+        DocItemLabel.PAGE_FOOTER,
+    }
+    for raw_chunk in chunker.chunk(conversion.document):
+        doc_items = tuple(raw_chunk.meta.doc_items or ())
+        if doc_items and all(item.label in excluded for item in doc_items):
+            continue
+        content = _clean_layout_text(raw_chunk.text)
+        if not content:
+            continue
+        contextualized = _clean_layout_text(chunker.contextualize(raw_chunk)) or content
+        start, end, compact_start = _project_layout_span(
+            source_text,
+            content,
+            compact_source=compact_source,
+            source_offsets=source_offsets,
+            compact_start=compact_start,
+        )
+        original = source_text[start:end] if start is not None and end is not None else content
+        pages: set[int] = set()
+        refs: list[str] = []
+        bboxes: list[dict[str, Any]] = []
+        for item in doc_items:
+            refs.append(str(item.self_ref))
+            for provenance in item.prov or ():
+                pages.add(int(provenance.page_no))
+                bbox = provenance.bbox
+                bboxes.append(
+                    {
+                        "page": int(provenance.page_no),
+                        "left": float(bbox.l),
+                        "top": float(bbox.t),
+                        "right": float(bbox.r),
+                        "bottom": float(bbox.b),
+                        "origin": str(bbox.coord_origin.value),
+                    }
+                )
+        result.append(
+            DocumentChunk(
+                original_content=original,
+                contextualized_content=contextualized,
+                source_start=start,
+                source_end=end,
+                source_start_line=(source_text.count("\n", 0, start) + 1 if start is not None else None),
+                source_end_line=(
+                    source_text.count("\n", 0, max(start or 0, (end or 1) - 1)) + 1
+                    if end is not None
+                    else None
+                ),
+                token_count=len(_tokenizer().encode(contextualized)),
+                heading_path=tuple(str(item) for item in (raw_chunk.meta.headings or ())),
+                source_pages=tuple(sorted(pages)),
+                doc_item_refs=tuple(refs),
+                source_bboxes=tuple(bboxes),
+            )
+        )
+    return result
+
+
+def merge_short_chunks(
+    chunks: list[DocumentChunk],
+    *,
+    source_text: str,
+    min_token_count: int,
+    max_token_count: int,
+) -> list[DocumentChunk]:
+    """在不突破长度上限的前提下，把过短块并入相邻内容。"""
+
+    result: list[DocumentChunk] = []
+    index = 0
+    while index < len(chunks):
+        current = chunks[index]
+        if current.token_count >= min_token_count:
+            result.append(current)
+            index += 1
+            continue
+        if index + 1 < len(chunks):
+            combined = _combine_chunks(current, chunks[index + 1], source_text)
+            if combined.token_count <= max_token_count:
+                result.append(combined)
+                index += 2
+                continue
+        if result:
+            combined = _combine_chunks(result[-1], current, source_text)
+            if combined.token_count <= max_token_count:
+                result[-1] = combined
+                index += 1
+                continue
+        result.append(current)
+        index += 1
+    return result
+
+
+def _combine_chunks(
+    left: DocumentChunk,
+    right: DocumentChunk,
+    source_text: str,
+) -> DocumentChunk:
+    contextualized = "\n\n".join(
+        part for part in (left.contextualized_content, right.contextualized_content) if part
+    )
+    start = left.source_start
+    end = right.source_end
+    has_contiguous_source = (
+        start is not None
+        and left.source_end is not None
+        and right.source_start is not None
+        and end is not None
+        and left.source_end <= right.source_start
+    )
+    original = (
+        source_text[start:end]
+        if has_contiguous_source and start is not None and end is not None
+        else "\n\n".join(
+            part for part in (left.original_content, right.original_content) if part
+        )
+    )
+    if not has_contiguous_source:
+        start = None
+        end = None
+    return DocumentChunk(
+        original_content=original,
+        contextualized_content=contextualized,
+        source_start=start,
+        source_end=end,
+        source_start_line=left.source_start_line if start is not None else None,
+        source_end_line=right.source_end_line if end is not None else None,
+        token_count=len(_tokenizer().encode(contextualized)),
+        heading_path=left.heading_path or right.heading_path,
+        source_pages=tuple(sorted(set(left.source_pages) | set(right.source_pages))),
+        doc_item_refs=left.doc_item_refs + right.doc_item_refs,
+        source_bboxes=left.source_bboxes + right.source_bboxes,
+    )

backend/app/modules/data_process/generation.py

@@ -0,0 +1,562 @@
+"""数据处理任务的大模型生成适配器。"""
+
+from __future__ import annotations
+
+import hashlib
+import json
+import logging
+import re
+from collections.abc import Callable, Iterable, Mapping
+from typing import Any
+from urllib.parse import urlsplit, urlunsplit
+
+import httpx
+
+from app.modules.data_process.algorithms import normalize_text, stable_split_assignments
+from app.modules.data_process.constants import (
+    MAX_QA_PAIRS_PER_ITEM,
+    MODEL_GENERATION_BATCH_SIZE,
+)
+
+
+class ModelGenerationError(ValueError):
+    """模型配置、响应或调用失败。"""
+
+
+class _TerminalModelGenerationError(ModelGenerationError):
+    """使用相同参数重试也无法恢复的模型响应错误。"""
+
+
+OUTPUT_TYPE_STANDARD = "standard"
+OUTPUT_TYPE_REASONING = "reasoning"
+SUPPORTED_OUTPUT_TYPES = {OUTPUT_TYPE_STANDARD, OUTPUT_TYPE_REASONING}
+REASONING_DETAIL_NORMAL = "normal"
+REASONING_DETAIL_DETAILED = "detailed"
+SUPPORTED_REASONING_DETAILS = {
+    REASONING_DETAIL_NORMAL,
+    REASONING_DETAIL_DETAILED,
+}
+MINIMAX_M3_API_HOSTS = {"api.minimax.io", "api.minimaxi.com"}
+MINIMAX_M3_MIN_COMPLETION_TOKENS = 4096
+logger = logging.getLogger(__name__)
+
+
+def _is_retryable_generation_error(exc: Exception) -> bool:
+    if isinstance(exc, _TerminalModelGenerationError):
+        return False
+    if isinstance(exc, httpx.HTTPStatusError):
+        status_code = exc.response.status_code
+        return status_code in {408, 425, 429} or status_code >= 500
+    if isinstance(exc, httpx.RequestError):
+        return True
+    return isinstance(exc, (json.JSONDecodeError, ModelGenerationError))
+
+
+def _is_official_minimax_m3(endpoint: str, model_name: str) -> bool:
+    host = (urlsplit(endpoint).hostname or "").casefold()
+    return host in MINIMAX_M3_API_HOSTS and model_name.casefold() == "minimax-m3"
+
+
+def chat_completions_url(value: str) -> str:
+    """把域名、基础 URL 或完整地址统一为 chat completions 地址。"""
+
+    raw = (value or "").strip()
+    if not raw:
+        raise ModelGenerationError("generation model api_url is required")
+    if "://" not in raw:
+        raw = f"https://{raw}"
+    parsed = urlsplit(raw)
+    if parsed.scheme not in {"http", "https"} or not parsed.hostname:
+        raise ModelGenerationError("generation model api_url must be an HTTP(S) host or URL")
+    if parsed.username or parsed.password:
+        raise ModelGenerationError("generation model api_url must not contain credentials")
+
+    path = parsed.path.rstrip("/")
+    if path.endswith("/chat/completions"):
+        target_path = path
+    elif path.endswith("/v1"):
+        target_path = f"{path}/chat/completions"
+    elif not path:
+        target_path = "/v1/chat/completions"
+    else:
+        target_path = f"{path}/v1/chat/completions"
+    return urlunsplit((parsed.scheme, parsed.netloc, target_path, "", ""))
+
+
+def _response_choice(payload: Mapping[str, Any]) -> Mapping[str, Any]:
+    try:
+        choice = payload["choices"][0]
+    except (KeyError, IndexError, TypeError) as exc:
+        raise ModelGenerationError("模型响应缺少 choices[0]") from exc
+    if not isinstance(choice, Mapping):
+        raise ModelGenerationError("模型响应 choices[0] 不是对象")
+    return choice
+
+
+def _response_finish_reason(payload: Mapping[str, Any]) -> str:
+    try:
+        return str(_response_choice(payload).get("finish_reason") or "").strip().lower()
+    except ModelGenerationError:
+        return ""
+
+
+def _response_content_length(payload: Mapping[str, Any]) -> int:
+    try:
+        message = _response_choice(payload).get("message")
+        if not isinstance(message, Mapping):
+            return 0
+        content = message.get("content")
+        if isinstance(content, str):
+            return len(content)
+        if isinstance(content, list):
+            return sum(
+                len(str(item.get("text") or ""))
+                for item in content
+                if isinstance(item, Mapping)
+            )
+    except ModelGenerationError:
+        pass
+    return 0
+
+
+def _raise_for_terminal_response(payload: Mapping[str, Any]) -> Mapping[str, Any]:
+    choice = _response_choice(payload)
+    base_response = payload.get("base_resp")
+    status_code: Any = None
+    status_message = ""
+    if isinstance(base_response, Mapping):
+        status_code = base_response.get("status_code")
+        status_message = re.sub(
+            r"\s+", " ", str(base_response.get("status_msg") or "")
+        ).strip()[:200]
+
+    if bool(payload.get("input_sensitive")) or status_code in {1026, "1026"}:
+        raise _TerminalModelGenerationError(
+            f"模型输入触发内容安全拦截（code={status_code or 1026}）"
+        )
+    if bool(payload.get("output_sensitive")) or status_code in {1027, "1027"}:
+        raise _TerminalModelGenerationError(
+            f"模型输出触发内容安全拦截（code={status_code or 1027}）"
+        )
+
+    finish_reason = str(choice.get("finish_reason") or "").strip().lower()
+    if finish_reason == "length":
+        raise _TerminalModelGenerationError(
+            "模型输出因达到 Token 上限被截断（finish_reason=length），"
+            "请提高最大输出长度后重试"
+        )
+    if finish_reason == "content_filter":
+        raise _TerminalModelGenerationError(
+            "模型输出被内容安全策略拦截（finish_reason=content_filter）"
+        )
+    if finish_reason in {"tool_calls", "function_call"}:
+        raise _TerminalModelGenerationError(
+            f"模型返回了当前生成任务不支持的工具调用（finish_reason={finish_reason}）"
+        )
+    if status_code not in {None, "", 0, "0"}:
+        detail = f"：{status_message}" if status_message else ""
+        raise _TerminalModelGenerationError(
+            f"模型服务返回业务错误（code={status_code}）{detail}"
+        )
+    return choice
+
+
+def _message_content(payload: Mapping[str, Any]) -> str:
+    choice = _raise_for_terminal_response(payload)
+    message = choice.get("message")
+    if not isinstance(message, Mapping):
+        raise ModelGenerationError("模型响应缺少 choices[0].message")
+    content = message.get("content")
+    if isinstance(content, str):
+        result = content
+    elif isinstance(content, list):
+        parts = [
+            str(item.get("text") or "")
+            for item in content
+            if isinstance(item, Mapping) and item.get("type") in {None, "text", "output_text"}
+        ]
+        result = "".join(parts)
+    elif content is None:
+        result = ""
+    else:
+        raise ModelGenerationError("模型响应 content 必须是文本")
+    if not result.strip():
+        raise ModelGenerationError("模型返回的最终内容为空，未生成可解析的 JSON")
+    return result
+
+
+def _json_documents(content: str) -> list[Any]:
+    decoder = json.JSONDecoder()
+    documents: list[Any] = []
+    cursor = 0
+    while cursor < len(content):
+        match = re.search(r"[\[{]", content[cursor:])
+        if not match:
+            break
+        start = cursor + match.start()
+        try:
+            value, end = decoder.raw_decode(content[start:])
+        except json.JSONDecodeError:
+            cursor = start + 1
+            continue
+        if isinstance(value, (Mapping, list)):
+            documents.append(value)
+        cursor = start + max(end, 1)
+    return documents
+
+
+def _json_payload(content: str) -> Any:
+    # 只移除模型在 JSON 之前自行输出的思考过程，不能破坏 JSON 字段中的训练内容。
+    cleaned = content.strip()
+    if re.match(r"^\s*<think>", cleaned, flags=re.IGNORECASE) and not re.match(
+        r"^\s*<think>[\s\S]*?</think>", cleaned, flags=re.IGNORECASE
+    ):
+        raise ModelGenerationError("模型思考内容未闭合，响应可能已被截断")
+    cleaned = re.sub(
+        r"^\s*(?:<think>[\s\S]*?</think>\s*)+",
+        "",
+        cleaned,
+        count=1,
+        flags=re.IGNORECASE,
+    ).strip()
+    fenced = re.fullmatch(r"```(?:json)?\s*([\s\S]*?)\s*```", cleaned, flags=re.IGNORECASE)
+    if fenced:
+        cleaned = fenced.group(1).strip()
+    try:
+        return json.loads(cleaned)
+    except json.JSONDecodeError as direct_error:
+        documents = _json_documents(cleaned)
+        if len(documents) == 1:
+            return documents[0]
+        if len(documents) > 1:
+            raise ModelGenerationError("模型响应包含多个 JSON 对象，无法确定应使用哪一个")
+        raise ModelGenerationError(
+            "模型响应中没有找到唯一且完整的 JSON 对象"
+            f"（第 {direct_error.lineno} 行，第 {direct_error.colno} 列）"
+        ) from direct_error
+
+
+def _result_items(payload: Any) -> list[Mapping[str, Any]]:
+    if isinstance(payload, list):
+        values = payload
+    elif isinstance(payload, Mapping):
+        nested = next(
+            (
+                payload[key]
+                for key in ("items", "results", "data", "records")
+                if isinstance(payload.get(key), list)
+            ),
+            None,
+        )
+        values = nested if isinstance(nested, list) else [payload]
+    else:
+        raise ModelGenerationError("model JSON must be an object or array")
+    items = [item for item in values if isinstance(item, Mapping)]
+    if not items:
+        raise ModelGenerationError("model JSON does not contain result objects")
+    return items
+
+
+def _prompt_messages(
+    prompt: str,
+    content: str,
+    count: int,
+    *,
+    start_index: int,
+    total_count: int,
+    output_type: str,
+    reasoning_detail: str,
+) -> list[dict[str, str]]:
+    end_index = start_index + count - 1
+    if output_type == OUTPUT_TYPE_REASONING:
+        schema = '{"items":[{"instruction":"...","input":"...","reasoning":"...","answer":"..."}]}'
+        detail_rule = (
+            "推理详细程度为“详细”：完整展开问题条件、来源依据、中间计算或推导，"
+            "并在得出答案前核对结论；每一步都必须能从来源内容中验证。"
+            if reasoning_detail == REASONING_DETAIL_DETAILED
+            else
+            "推理详细程度为“普通”：只保留得出答案所需的关键依据和必要步骤，"
+            "避免冗长复述、套话和无依据扩展。"
+        )
+        output_rule = (
+            "你正在生成用于训练推理模型的思维链数据，而不是普通问答数据。"
+            "instruction、reasoning 和 answer 均不得为空；reasoning 必须是基于来源内容、"
+            f"可核对的推理过程，answer 只写最终答案。{detail_rule}"
+            "这是思维链输出模式，即使其他提示语要求省略分析，也不得省略 reasoning。"
+            "不要自行添加 <think> 标签，系统会在保存时统一组装。"
+        )
+    else:
+        schema = '{"items":[{"instruction":"...","input":"...","output":"..."}]}'
+        output_rule = (
+            "你正在生成标准监督微调问答数据。instruction 和 output 不得为空；"
+            "output 只写最终答案，禁止输出分析、推理过程或 <think> 标签。"
+        )
+    schema_instruction = (
+        f"必须只返回 JSON 对象，格式为 {schema}；items 必须包含 {count} 条。"
+        f"这是总计 {total_count} 条中的第 {start_index}-{end_index} 条，"
+        "各条必须使用不同的提问角度和表述，避免重复。"
+        f"{output_rule}不要输出 Markdown 代码围栏或 JSON 之外的说明。"
+    )
+    base_prompt = normalize_text(prompt) or "请根据来源内容生成可用于监督微调的问答数据。"
+    if "{{ content }}" in base_prompt:
+        user_prompt = base_prompt.replace("{{ content }}", content)
+        return [
+            {"role": "system", "content": schema_instruction},
+            {"role": "user", "content": user_prompt},
+        ]
+    return [
+        {"role": "system", "content": f"{base_prompt}\n{schema_instruction}"},
+        {"role": "user", "content": f"来源内容：\n{content}"},
+    ]
+
+
+def generate_model_records(
+    preview_items: Iterable[Mapping[str, Any]],
+    *,
+    model: Mapping[str, Any],
+    config: Mapping[str, Any],
+    task_id: str,
+    split: Mapping[str, int],
+    qa_pairs_per_item: int,
+    client: httpx.Client | None = None,
+    on_progress: Callable[[int, int], None] | None = None,
+) -> list[dict[str, Any]]:
+    """调用 OpenAI 兼容接口，将预览切片生成标准训练记录。
+
+    每个切片按安全批次调用模型；失败批次会产生一条可人工修复的
+    invalid 结果，已经成功的批次不会丢失。
+    """
+
+    if not 1 <= qa_pairs_per_item <= MAX_QA_PAIRS_PER_ITEM:
+        raise ModelGenerationError(f"qa_pairs_per_item must be in [1, {MAX_QA_PAIRS_PER_ITEM}]")
+    output_type = str(config.get("output_type") or OUTPUT_TYPE_STANDARD).strip().lower()
+    if output_type not in SUPPORTED_OUTPUT_TYPES:
+        raise ModelGenerationError(f"output_type must be one of {sorted(SUPPORTED_OUTPUT_TYPES)}")
+    reasoning_detail = str(
+        config.get("reasoning_detail") or REASONING_DETAIL_NORMAL
+    ).strip().lower()
+    if reasoning_detail not in SUPPORTED_REASONING_DETAILS:
+        raise ModelGenerationError(
+            f"reasoning_detail must be one of {sorted(SUPPORTED_REASONING_DETAILS)}"
+        )
+    endpoint = chat_completions_url(str(model.get("api_url") or ""))
+    model_name = str(model.get("online_model_name") or model.get("name") or "").strip()
+    if not model_name:
+        raise ModelGenerationError("generation model name is required")
+    is_minimax_m3 = _is_official_minimax_m3(endpoint, model_name)
+
+    temperature = float(config.get("temperature", 0.7))
+    max_tokens = int(config.get("max_tokens", 1024))
+    timeout = max(1.0, min(120.0, float(config.get("request_timeout_seconds", 60))))
+    retries = max(0, min(5, int(config.get("generation_retries", 2))))
+    headers = {"Content-Type": "application/json"}
+    api_key = str(model.get("api_key") or "").strip()
+    if api_key:
+        headers["Authorization"] = f"Bearer {api_key}"
+
+    owns_client = client is None
+    http_client = client or httpx.Client(timeout=timeout)
+    results: list[dict[str, Any]] = []
+    try:
+        preview_list = list(preview_items)
+        total_items = len(preview_list)
+        for item_index, item in enumerate(preview_list):
+            preview_id = str(item.get("id") or f"preview-{item_index + 1}")
+            content = normalize_text(
+                str(item.get("edited_content") or item.get("original_content") or "")
+            )
+            for batch_offset in range(0, qa_pairs_per_item, MODEL_GENERATION_BATCH_SIZE):
+                batch_count = min(
+                    MODEL_GENERATION_BATCH_SIZE,
+                    qa_pairs_per_item - batch_offset,
+                )
+                batch_start = batch_offset + 1
+                batch_end = batch_offset + batch_count
+                request_payload: dict[str, Any] = {
+                    "model": model_name,
+                    "messages": _prompt_messages(
+                        str(config.get("generation_prompt") or ""),
+                        content,
+                        batch_count,
+                        start_index=batch_start,
+                        total_count=qa_pairs_per_item,
+                        output_type=output_type,
+                        reasoning_detail=reasoning_detail,
+                    ),
+                    "temperature": temperature,
+                }
+                if is_minimax_m3:
+                    request_payload.update(
+                        reasoning_split=True,
+                        max_completion_tokens=max(
+                            max_tokens,
+                            MINIMAX_M3_MIN_COMPLETION_TOKENS,
+                        ),
+                    )
+                else:
+                    request_payload["max_tokens"] = max_tokens
+                if bool(config.get("json_mode", False)) and not is_minimax_m3:
+                    request_payload["response_format"] = {"type": "json_object"}
+
+                last_error: Exception | None = None
+                generated_items: list[Mapping[str, Any]] | None = None
+                for _ in range(retries + 1):
+                    try:
+                        response = http_client.post(
+                            endpoint,
+                            headers=headers,
+                            json=request_payload,
+                        )
+                        response.raise_for_status()
+                        body = response.json()
+                        if not isinstance(body, Mapping):
+                            raise ModelGenerationError("model response body must be a JSON object")
+                        try:
+                            candidate_items = _result_items(
+                                _json_payload(_message_content(body))
+                            )
+                        except ModelGenerationError as exc:
+                            logger.warning(
+                                "data process model response rejected task_id=%s model=%s "
+                                "finish_reason=%s response_chars=%s input_sensitive=%s "
+                                "output_sensitive=%s reason=%s",
+                                task_id,
+                                model_name,
+                                _response_finish_reason(body) or "missing",
+                                _response_content_length(body),
+                                bool(body.get("input_sensitive")),
+                                bool(body.get("output_sensitive")),
+                                str(exc),
+                            )
+                            raise
+                        if len(candidate_items) < batch_count:
+                            raise ModelGenerationError(
+                                "model response contains fewer result objects than requested: "
+                                f"expected {batch_count}, got {len(candidate_items)}"
+                            )
+                        generated_items = candidate_items
+                        break
+                    except (
+                        httpx.HTTPError,
+                        json.JSONDecodeError,
+                        ModelGenerationError,
+                    ) as exc:
+                        last_error = exc
+                        if not _is_retryable_generation_error(exc):
+                            break
+
+                if generated_items is None:
+                    error_message = str(last_error or "model generation failed")[:2000]
+                    failure_instruction = (
+                        f"模型生成失败，请人工补充（第 {batch_start}-{batch_end} 条）"
+                    )
+                    result_id = (
+                        "result_"
+                        f"{hashlib.sha256(f'{preview_id}:error:{batch_start}'.encode()).hexdigest()[:16]}"
+                    )
+                    results.append(
+                        {
+                            "id": result_id,
+                            "preview_item_id": preview_id,
+                            "instruction": failure_instruction,
+                            "input": content,
+                            "output": "",
+                            "original_instruction": failure_instruction,
+                            "original_input": content,
+                            "original_output": "",
+                            "status": "invalid",
+                            "error": error_message,
+                            "split": "train",
+                        }
+                    )
+                    continue
+
+                for batch_index, value in enumerate(generated_items[:batch_count]):
+                    variant_index = batch_offset + batch_index
+                    instruction = normalize_text(
+                        str(value.get("instruction") or value.get("question") or "")
+                    )
+                    input_text = normalize_text(
+                        str(value.get("input") or value.get("context") or "")
+                    )
+                    if output_type == OUTPUT_TYPE_REASONING:
+                        reasoning = normalize_text(
+                            re.sub(
+                                r"</?think>",
+                                "",
+                                str(value.get("reasoning") or value.get("analysis") or ""),
+                                flags=re.IGNORECASE,
+                            )
+                        )
+                        answer = normalize_text(
+                            re.sub(
+                                r"</?think>",
+                                "",
+                                str(
+                                    value.get("answer")
+                                    or value.get("final_answer")
+                                    or value.get("output")
+                                    or ""
+                                ),
+                                flags=re.IGNORECASE,
+                            )
+                        )
+                        output = (
+                            f"<think>\n{reasoning}\n</think>\n{answer}"
+                            if reasoning and answer
+                            else answer or (f"<think>\n{reasoning}\n</think>" if reasoning else "")
+                        )
+                        valid = bool(instruction and reasoning and answer)
+                        missing_error = "model result is missing instruction, reasoning or answer"
+                    else:
+                        output = normalize_text(
+                            str(
+                                value.get("output")
+                                or value.get("answer")
+                                or value.get("response")
+                                or ""
+                            )
+                        )
+                        output = normalize_text(
+                            re.sub(
+                                r"<think>[\s\S]*?(?:</think>|$)",
+                                "",
+                                output,
+                                flags=re.IGNORECASE,
+                            )
+                        )
+                        valid = bool(instruction and output)
+                        missing_error = "model result is missing instruction or output"
+                    raw_id = f"{preview_id}:{variant_index + 1}:{instruction}:{output}"
+                    result_id = f"result_{hashlib.sha256(raw_id.encode()).hexdigest()[:16]}"
+                    results.append(
+                        {
+                            "id": result_id,
+                            "preview_item_id": preview_id,
+                            "instruction": instruction,
+                            "input": input_text,
+                            "output": output,
+                            "original_instruction": instruction,
+                            "original_input": input_text,
+                            "original_output": output,
+                            "status": "valid" if valid else "invalid",
+                            "error": (None if valid else missing_error),
+                            "split": "train",
+                        }
+                    )
+            if on_progress:
+                on_progress(item_index + 1, total_items)
+    finally:
+        if owns_client:
+            http_client.close()
+    assignments = stable_split_assignments(
+        [str(result["id"]) for result in results],
+        split,
+        seed=task_id,
+    )
+    for result, assignment in zip(results, assignments, strict=True):
+        result["split"] = assignment
+    return results
+
+
+__all__ = ["ModelGenerationError", "chat_completions_url", "generate_model_records"]

backend/app/modules/data_process/office_preview.py

@@ -0,0 +1,308 @@
+"""Word 与 Excel 原文件的安全、受限预览模型。
+
+预览只返回浏览器绘制所需的结构化数据，不返回或执行 Office 包中的活动内容。
+DOCX 的字符偏移与上传时的正文抽取规则保持一致，供前端定位当前切片。
+"""
+
+from __future__ import annotations
+
+import io
+import re
+from typing import Any
+
+from docx import Document
+from docx.oxml.table import CT_Tbl
+from docx.oxml.text.paragraph import CT_P
+from docx.table import Table
+from docx.text.paragraph import Paragraph
+from openpyxl import load_workbook
+
+from app.modules.data_process.algorithms import (
+    _MAX_WORKBOOK_COLUMNS,
+    _MAX_WORKBOOK_HEADER_SCAN_ROWS,
+    _infer_xlsx_header_region,
+    _normalize_spreadsheet_value,
+    _rewrite_xlsx_workbook_relationships,
+    _validate_office_archive,
+    _xlsx_sheet_merge_ranges,
+    normalize_text,
+)
+
+MAX_DOCX_PREVIEW_BLOCKS = 2_000
+MAX_XLSX_PREVIEW_ROWS = 200
+
+
+def _docx_alignment(paragraph: Paragraph) -> str:
+    value = paragraph.alignment
+    return {
+        0: "left",
+        1: "center",
+        2: "right",
+        3: "justify",
+        4: "distribute",
+        5: "justify",
+        7: "justify",
+        8: "distribute",
+        9: "distribute",
+    }.get(int(value) if value is not None else -1, "left")
+
+
+def _docx_heading_level(paragraph: Paragraph) -> int | None:
+    style = paragraph.style
+    if style is None:
+        return None
+    style_name = str(style.name or "")
+    style_id = str(style.style_id or "")
+    match = re.search(r"(?:heading|标题)\s*([1-6])", f"{style_name} {style_id}", re.IGNORECASE)
+    return int(match.group(1)) if match else None
+
+
+def build_docx_preview(raw: bytes) -> dict[str, Any]:
+    """把 DOCX 转为保留标题、段落和表格顺序的浏览器预览模型。"""
+
+    _validate_office_archive(raw, "docx")
+    try:
+        document = Document(io.BytesIO(raw))
+    except Exception as exc:
+        raise ValueError(f"invalid DOCX file: {exc}") from exc
+
+    blocks: list[dict[str, Any]] = []
+    source_cursor = 0
+    has_source_content = False
+    rendered_blocks = 0
+    truncated = False
+
+    def source_range(value: str) -> tuple[str, int, int] | None:
+        nonlocal source_cursor, has_source_content
+        text = normalize_text(value)
+        if not text:
+            return None
+        if has_source_content:
+            source_cursor += 2
+        start = source_cursor
+        source_cursor += len(text)
+        has_source_content = True
+        return text, start, source_cursor
+
+    for child in document.element.body.iterchildren():
+        if rendered_blocks >= MAX_DOCX_PREVIEW_BLOCKS:
+            truncated = True
+            break
+
+        if isinstance(child, CT_P):
+            paragraph = Paragraph(child, document)
+            located = source_range(paragraph.text)
+            if located is None:
+                continue
+            text, start, end = located
+            style_name = str(paragraph.style.name or "") if paragraph.style else ""
+            blocks.append(
+                {
+                    "type": "paragraph",
+                    "text": text,
+                    "style": style_name,
+                    "heading_level": _docx_heading_level(paragraph),
+                    "alignment": _docx_alignment(paragraph),
+                    "is_list": "list" in style_name.casefold() or "列表" in style_name,
+                    "source_start": start,
+                    "source_end": end,
+                }
+            )
+            rendered_blocks += 1
+            continue
+
+        if not isinstance(child, CT_Tbl):
+            continue
+        table = Table(child, document)
+        preview_rows: list[dict[str, Any]] = []
+        for row in table.rows:
+            if rendered_blocks >= MAX_DOCX_PREVIEW_BLOCKS:
+                truncated = True
+                break
+            cell_values = [normalize_text(cell.text) for cell in row.cells]
+            located = source_range("\t".join(cell_values))
+            if located is None:
+                continue
+            _, start, end = located
+            preview_rows.append(
+                {
+                    "cells": cell_values,
+                    "source_start": start,
+                    "source_end": end,
+                }
+            )
+            rendered_blocks += 1
+        if preview_rows:
+            blocks.append({"type": "table", "rows": preview_rows})
+        if truncated:
+            break
+
+    return {
+        "format": "docx",
+        "blocks": blocks,
+        "truncated": truncated,
+    }
+
+
+def build_xlsx_preview(
+    raw: bytes,
+    *,
+    sheet_index: int = 0,
+    offset: int = 0,
+    limit: int = 100,
+) -> dict[str, Any]:
+    """按工作表分页返回 XLSX 的表头和记录网格。"""
+
+    if sheet_index < 0 or offset < 0:
+        raise ValueError("sheet_index and offset must be non-negative")
+    if limit < 1 or limit > MAX_XLSX_PREVIEW_ROWS:
+        raise ValueError(
+            f"XLSX preview limit must be in [1, {MAX_XLSX_PREVIEW_ROWS}]"
+        )
+
+    _validate_office_archive(raw, "xlsx")
+    merged_by_sheet, normalized_targets = _xlsx_sheet_merge_ranges(raw)
+    workbook_raw = (
+        _rewrite_xlsx_workbook_relationships(raw, normalized_targets)
+        if normalized_targets
+        else raw
+    )
+    try:
+        workbook = load_workbook(
+            io.BytesIO(workbook_raw),
+            read_only=True,
+            data_only=True,
+            keep_links=False,
+        )
+    except Exception as exc:
+        raise ValueError(f"invalid XLSX file: {exc}") from exc
+
+    try:
+        sheets = [
+            {
+                "index": index,
+                "name": worksheet.title,
+                "state": worksheet.sheet_state,
+            }
+            for index, worksheet in enumerate(workbook.worksheets)
+        ]
+        if not sheets:
+            raise ValueError("XLSX workbook contains no worksheets")
+        if sheet_index >= len(sheets):
+            raise ValueError("XLSX worksheet index is out of range")
+
+        worksheet = workbook.worksheets[sheet_index]
+        reset_dimensions = getattr(worksheet, "reset_dimensions", None)
+        if callable(reset_dimensions):
+            reset_dimensions()
+        row_iterator = enumerate(worksheet.iter_rows(values_only=True), start=1)
+        buffered_rows: dict[int, tuple[Any, ...]] = {}
+
+        def normalized_values(row: tuple[Any, ...]) -> list[Any]:
+            values = list(row)
+            while values and values[-1] in {None, ""}:
+                values.pop()
+            if len(values) > _MAX_WORKBOOK_COLUMNS:
+                raise ValueError(
+                    f"XLSX worksheet {worksheet.title!r} exceeds "
+                    f"{_MAX_WORKBOOK_COLUMNS} columns"
+                )
+            return values
+
+        for row_number, row in row_iterator:
+            values = normalized_values(row)
+            if not values or all(value in {None, ""} for value in values):
+                continue
+            buffered_rows[row_number] = tuple(values)
+            if len(buffered_rows) >= _MAX_WORKBOOK_HEADER_SCAN_ROWS:
+                break
+
+        if not buffered_rows:
+            return {
+                "format": "xlsx",
+                "sheets": sheets,
+                "active_sheet": {
+                    "index": sheet_index,
+                    "name": worksheet.title,
+                    "columns": [],
+                    "rows": [],
+                    "offset": offset,
+                    "limit": limit,
+                    "has_more": False,
+                },
+            }
+
+        _, header_end_row, headers = _infer_xlsx_header_region(
+            worksheet.title,
+            buffered_rows,
+            merged_by_sheet.get(worksheet.title, ()),
+        )
+        preview_rows: list[dict[str, Any]] = []
+        record_index = 0
+        has_more = False
+
+        def append_row(row_number: int, values: tuple[Any, ...] | list[Any]) -> bool:
+            nonlocal record_index, has_more
+            row_values = list(values)
+            if len(row_values) > len(headers):
+                raise ValueError(
+                    f"XLSX worksheet {worksheet.title!r} has a row wider than its header"
+                )
+            row_values.extend([None] * (len(headers) - len(row_values)))
+            record = {
+                header: _normalize_spreadsheet_value(value)
+                for header, value in zip(headers, row_values, strict=True)
+            }
+            if not any(value not in {"", None} for value in record.values()):
+                return False
+            current_index = record_index
+            record_index += 1
+            if current_index < offset:
+                return False
+            if len(preview_rows) >= limit:
+                has_more = True
+                return True
+            preview_rows.append(
+                {
+                    "row_number": row_number,
+                    "record_index": current_index,
+                    "values": [record[header] for header in headers],
+                    "record": record,
+                }
+            )
+            return False
+
+        for row_number, values in buffered_rows.items():
+            if row_number > header_end_row and append_row(row_number, values):
+                break
+        else:
+            for row_number, row in row_iterator:
+                values = normalized_values(row)
+                if not values or all(value in {None, ""} for value in values):
+                    continue
+                if append_row(row_number, values):
+                    break
+
+        return {
+            "format": "xlsx",
+            "sheets": sheets,
+            "active_sheet": {
+                "index": sheet_index,
+                "name": worksheet.title,
+                "columns": headers,
+                "rows": preview_rows,
+                "offset": offset,
+                "limit": limit,
+                "has_more": has_more,
+            },
+        }
+    finally:
+        workbook.close()
+
+
+__all__ = [
+    "MAX_DOCX_PREVIEW_BLOCKS",
+    "MAX_XLSX_PREVIEW_ROWS",
+    "build_docx_preview",
+    "build_xlsx_preview",
+]

backend/app/modules/data_process/schema_cli.py

@@ -0,0 +1,76 @@
+"""数据处理运行表的显式检查与安装命令。"""
+
+from __future__ import annotations
+
+import argparse
+from urllib.parse import urlsplit
+
+from app.modules.data_process.store import DataProcessStore
+
+REQUIRED_TASK_COLUMNS = (
+    "generation_run_id",
+    "results_confirmed",
+    "workflow_step",
+    "preview_status",
+    "preview_progress",
+    "preview_run_id",
+    "preview_failure_reason",
+    "preview_total_files",
+    "preview_completed_files",
+)
+
+
+def _target_label(database_url: str) -> str:
+    parsed = urlsplit(database_url)
+    database = parsed.path.strip("/") or "(unknown)"
+    return f"{parsed.hostname or '(unknown)'}:{parsed.port or 5432}/{database}"
+
+
+def _schema_ready(store: DataProcessStore) -> bool:
+    with store.connect() as conn:
+        row = conn.execute(
+            """
+            SELECT COUNT(*) = %s AS ready
+            FROM information_schema.columns
+            WHERE table_schema=current_schema()
+              AND table_name='data_process_tasks'
+              AND column_name = ANY(%s)
+            """,
+            (len(REQUIRED_TASK_COLUMNS), list(REQUIRED_TASK_COLUMNS)),
+        ).fetchone()
+    return bool(row and row["ready"])
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(
+        description="检查或显式安装数据处理运行表（不会由应用启动自动执行）"
+    )
+    action = parser.add_mutually_exclusive_group(required=True)
+    action.add_argument("--check", action="store_true", help="只读检查迁移是否已安装")
+    action.add_argument("--apply", action="store_true", help="执行 002 数据处理迁移")
+    parser.add_argument(
+        "--yes",
+        action="store_true",
+        help="确认允许修改 DATABASE_URL 指向的数据库；与 --apply 同时使用",
+    )
+    args = parser.parse_args()
+
+    store = DataProcessStore()
+    target = _target_label(store.database_url)
+    if args.check:
+        ready = _schema_ready(store)
+        print(f"数据处理 schema：{'已安装' if ready else '未安装'}；目标：{target}")
+        return 0 if ready else 1
+    if not args.yes:
+        parser.error("--apply 必须同时提供 --yes，确认修改目标数据库")
+
+    print(f"正在安装数据处理 schema；目标：{target}")
+    store.ensure_schema()
+    if not _schema_ready(store):
+        raise RuntimeError("迁移执行后仍未检测到 generation_run_id")
+    print("数据处理 schema 安装完成")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

backend/app/modules/data_process/storage.py

@@ -0,0 +1,499 @@
+"""数据处理原始源文件的受控本地对象存储。"""
+
+from __future__ import annotations
+
+import os
+import re
+import stat
+import unicodedata
+import uuid
+from dataclasses import dataclass
+from functools import lru_cache
+from pathlib import Path, PurePosixPath
+from typing import Iterable, Iterator
+from urllib.parse import quote, unquote, urlsplit
+
+
+class DataProcessStorageError(ValueError):
+    """本地对象引用或文件系统状态不安全。"""
+
+
+@dataclass(frozen=True, slots=True)
+class StagedSourceObject:
+    """尚未发布的原始文件；绝对路径仅在存储模块内部流转。"""
+
+    reference: str
+    _temporary_path: Path
+    _relative_path: PurePosixPath
+
+
+def _default_storage_root() -> Path:
+    return Path(__file__).resolve().parents[3] / "storage" / "data-process"
+
+
+def _configured_storage_root() -> Path:
+    configured = os.getenv("DATA_PROCESS_STORAGE_DIR", "").strip()
+    if not configured:
+        return _default_storage_root()
+    path = Path(configured).expanduser()
+    # 相对配置固定以 backend 目录为基准，
+    # 避免从不同 cwd 启动时写入不同位置。
+    return path if path.is_absolute() else Path(__file__).resolve().parents[3] / path
+
+
+def _safe_component(value: str, label: str) -> str:
+    if not value or value in {".", ".."} or len(value) > 128:
+        raise DataProcessStorageError(f"invalid {label}")
+    if not value[0].isalnum() or any(
+        not (character.isalnum() or character in {"-", "_", "."})
+        for character in value
+    ):
+        raise DataProcessStorageError(f"invalid {label}")
+    return value
+
+
+def _safe_basename(value: str) -> str:
+    if not value or len(value.encode("utf-8")) > 255:
+        raise DataProcessStorageError("invalid source file name")
+    if value != Path(value).name or "/" in value or "\\" in value or "\x00" in value:
+        raise DataProcessStorageError("invalid source file name")
+    if value in {".", ".."} or any(
+        unicodedata.category(character).startswith("C") for character in value
+    ):
+        raise DataProcessStorageError("invalid source file name")
+    return value
+
+
+class LocalDataProcessStorage:
+    """只允许访问配置根目录下的版本化原始文件。"""
+
+    def __init__(self, root: str | os.PathLike[str] | Path | None = None) -> None:
+        configured = Path(root) if root is not None else _configured_storage_root()
+        configured = configured.expanduser()
+        if configured.exists() and configured.is_symlink():
+            raise DataProcessStorageError("data process storage root must not be a symlink")
+        configured.mkdir(parents=True, exist_ok=True, mode=0o700)
+        self._root = configured.resolve(strict=True)
+        # StagedSourceObject 本身是普通 dataclass，不能只依赖其中的路径字段判断
+        # 来源；只接受由当前存储实例实际签发的对象，
+        # 避免调用方伪造暂存路径。
+        self._issued_staged_objects: dict[Path, StagedSourceObject] = {}
+        self._ensure_directory(self._root / ".staging")
+
+    @property
+    def root(self) -> Path:
+        """仅供运维和测试检查；API 响应不得序列化该属性。"""
+
+        return self._root
+
+    def new_batch_id(self) -> str:
+        return f"batch-{uuid.uuid4().hex}"
+
+    def stage_bytes(
+        self,
+        *,
+        batch_id: str,
+        task_id: str,
+        source_file_id: str,
+        version: int,
+        name: str,
+        content: bytes,
+    ) -> StagedSourceObject:
+        batch_id = _safe_component(batch_id, "batch id")
+        task_id = _safe_component(task_id, "task id")
+        source_file_id = _safe_component(source_file_id, "source file id")
+        if isinstance(version, bool) or not isinstance(version, int) or version < 1:
+            raise DataProcessStorageError("invalid source file version")
+        basename = _safe_basename(name)
+        if not isinstance(content, bytes):
+            raise TypeError("content must be bytes")
+
+        batch_directory = self._ensure_directory(self._root / ".staging" / batch_id)
+        temporary_path = batch_directory / f"{source_file_id}-{uuid.uuid4().hex}.tmp"
+        flags = os.O_CREAT | os.O_EXCL | os.O_WRONLY
+        if hasattr(os, "O_NOFOLLOW"):
+            flags |= os.O_NOFOLLOW
+        descriptor = os.open(temporary_path, flags, 0o600)
+        try:
+            with os.fdopen(descriptor, "wb", closefd=True) as stream:
+                stream.write(content)
+                stream.flush()
+                os.fsync(stream.fileno())
+        except Exception:
+            temporary_path.unlink(missing_ok=True)
+            raise
+
+        relative_path = PurePosixPath(
+            task_id,
+            source_file_id,
+            f"v{version}",
+            basename,
+        )
+        reference = (
+            "local://data-process/"
+            f"{task_id}/{source_file_id}/v{version}/{quote(basename, safe='')}"
+        )
+        staged = StagedSourceObject(reference, temporary_path, relative_path)
+        self._issued_staged_objects[temporary_path] = staged
+        return staged
+
+    def publish(self, objects: Iterable[StagedSourceObject]) -> None:
+        staged = list(objects)
+        published: list[StagedSourceObject] = []
+        try:
+            seen_temporary_paths: set[Path] = set()
+            for item in staged:
+                self._validate_staged_object(item, require_file=True)
+                if item._temporary_path in seen_temporary_paths:
+                    raise DataProcessStorageError("duplicate staged source object")
+                seen_temporary_paths.add(item._temporary_path)
+            for item in staged:
+                final_path = self._path_for_relative(item._relative_path)
+                self._ensure_directory(final_path.parent)
+                if final_path.exists() or final_path.is_symlink():
+                    raise DataProcessStorageError("source storage object already exists")
+                os.link(item._temporary_path, final_path, follow_symlinks=False)
+                published.append(item)
+                item._temporary_path.unlink()
+                self._fsync_directory(final_path.parent)
+        except Exception:
+            for item in reversed(published):
+                try:
+                    self.delete(item.reference)
+                except Exception:
+                    # 回滚必须尽量处理其余对象，并保留真正的发布异常。
+                    pass
+            for item in staged:
+                try:
+                    self.discard([item])
+                except Exception:
+                    pass
+            raise
+        self.discard(staged)
+
+    def discard(self, objects: Iterable[StagedSourceObject]) -> None:
+        staged = list(objects)
+        for item in staged:
+            self._validate_staged_object(item, require_file=False)
+
+        batch_directories: set[Path] = set()
+        first_error: Exception | None = None
+        for item in staged:
+            temporary_path = item._temporary_path
+            try:
+                temporary_path.unlink(missing_ok=True)
+            except Exception as exc:
+                if first_error is None:
+                    first_error = exc
+            else:
+                self._issued_staged_objects.pop(temporary_path, None)
+            batch_directories.add(temporary_path.parent)
+        for directory in batch_directories:
+            self._remove_empty_directory(directory)
+        if first_error is not None:
+            raise first_error
+
+    def read(self, reference: str) -> bytes | None:
+        """读取 local 引用；旧 ``db://`` 对象返回 ``None`` 由数据库正文兜底。"""
+
+        relative_path = self._relative_from_reference(reference)
+        if relative_path is None:
+            return None
+        descriptor, _ = self._open_read_descriptor(relative_path)
+        with os.fdopen(descriptor, "rb", closefd=True) as stream:
+            return stream.read()
+
+    def file_size(
+        self,
+        reference: str,
+        *,
+        expected_task_id: str,
+        expected_source_file_id: str,
+    ) -> int | None:
+        """返回受控 local 对象大小；旧 ``db://`` 对象没有原始文件。"""
+
+        relative_path = self._relative_from_reference(reference)
+        if relative_path is None:
+            return None
+        self._assert_expected_owner(
+            relative_path,
+            expected_task_id=expected_task_id,
+            expected_source_file_id=expected_source_file_id,
+        )
+        descriptor, info = self._open_read_descriptor(relative_path)
+        os.close(descriptor)
+        return info.st_size
+
+    def iter_bytes(
+        self,
+        reference: str,
+        *,
+        expected_task_id: str,
+        expected_source_file_id: str,
+        expected_size: int,
+        start: int = 0,
+        length: int | None = None,
+        chunk_size: int = 256 * 1024,
+    ) -> Iterator[bytes]:
+        """按范围流式读取原始文件，避免 PDF 预览把大文件整体载入内存。"""
+
+        relative_path = self._relative_from_reference(reference)
+        if relative_path is None:
+            raise DataProcessStorageError("original source object is not available")
+        self._assert_expected_owner(
+            relative_path,
+            expected_task_id=expected_task_id,
+            expected_source_file_id=expected_source_file_id,
+        )
+        if start < 0 or expected_size < 0 or chunk_size < 1:
+            raise DataProcessStorageError("invalid source byte range")
+        descriptor, info = self._open_read_descriptor(relative_path)
+        if info.st_size != expected_size:
+            os.close(descriptor)
+            raise DataProcessStorageError("source object size does not match metadata")
+        remaining = expected_size - start if length is None else length
+        if remaining < 0 or start + remaining > expected_size:
+            os.close(descriptor)
+            raise DataProcessStorageError("invalid source byte range")
+        with os.fdopen(descriptor, "rb", closefd=True) as stream:
+            stream.seek(start)
+            while remaining:
+                chunk = stream.read(min(chunk_size, remaining))
+                if not chunk:
+                    raise DataProcessStorageError("source object ended unexpectedly")
+                remaining -= len(chunk)
+                yield chunk
+
+    def validate_owner(
+        self,
+        reference: str,
+        *,
+        expected_task_id: str,
+        expected_source_file_id: str,
+    ) -> bool:
+        """校验 local 引用归属；旧 ``db://`` 引用无需文件系统处理。"""
+
+        relative_path = self._relative_from_reference(reference)
+        if relative_path is None:
+            return False
+        self._assert_expected_owner(
+            relative_path,
+            expected_task_id=expected_task_id,
+            expected_source_file_id=expected_source_file_id,
+        )
+        return True
+
+    def _open_read_descriptor(
+        self,
+        relative_path: PurePosixPath,
+    ) -> tuple[int, os.stat_result]:
+        path = self._path_for_relative(relative_path)
+        self._assert_controlled_parent(path)
+        try:
+            before_open = path.lstat()
+        except FileNotFoundError as exc:
+            raise DataProcessStorageError("source storage object does not exist") from exc
+        if stat.S_ISLNK(before_open.st_mode) or not stat.S_ISREG(before_open.st_mode):
+            raise DataProcessStorageError("source storage object is not a regular file")
+        flags = os.O_RDONLY
+        if hasattr(os, "O_NOFOLLOW"):
+            flags |= os.O_NOFOLLOW
+        descriptor = os.open(path, flags)
+        after_open = os.fstat(descriptor)
+        if (
+            not stat.S_ISREG(after_open.st_mode)
+            or before_open.st_dev != after_open.st_dev
+            or before_open.st_ino != after_open.st_ino
+        ):
+            os.close(descriptor)
+            raise DataProcessStorageError("source storage object changed while opening")
+        return descriptor, after_open
+
+    def delete(
+        self,
+        reference: str,
+        *,
+        expected_task_id: str | None = None,
+        expected_source_file_id: str | None = None,
+    ) -> bool:
+        """删除受控 local 对象；旧 ``db://`` 引用保持不变。"""
+
+        relative_path = self._relative_from_reference(reference)
+        if relative_path is None:
+            return False
+        if (expected_task_id is None) != (expected_source_file_id is None):
+            raise DataProcessStorageError("both expected storage owner fields are required")
+        if expected_task_id is not None and expected_source_file_id is not None:
+            self._assert_expected_owner(
+                relative_path,
+                expected_task_id=expected_task_id,
+                expected_source_file_id=expected_source_file_id,
+            )
+        path = self._path_for_relative(relative_path)
+        self._assert_controlled_parent(path)
+        try:
+            info = path.lstat()
+        except FileNotFoundError:
+            return False
+        if stat.S_ISLNK(info.st_mode) or not stat.S_ISREG(info.st_mode):
+            raise DataProcessStorageError("refusing to delete a non-regular storage object")
+        path.unlink()
+        self._fsync_directory(path.parent)
+        for directory in (path.parent, path.parent.parent, path.parent.parent.parent):
+            self._remove_empty_directory(directory)
+        return True
+
+    @staticmethod
+    def _assert_expected_owner(
+        relative_path: PurePosixPath,
+        *,
+        expected_task_id: str,
+        expected_source_file_id: str,
+    ) -> None:
+        task_id = _safe_component(expected_task_id, "expected task id")
+        source_file_id = _safe_component(
+            expected_source_file_id,
+            "expected source file id",
+        )
+        if relative_path.parts[:2] != (task_id, source_file_id):
+            raise DataProcessStorageError("source storage object owner mismatch")
+
+    def _relative_from_reference(self, reference: str) -> PurePosixPath | None:
+        if reference.startswith("db://"):
+            return None
+        parsed = urlsplit(reference)
+        if parsed.scheme != "local" or parsed.netloc != "data-process":
+            raise DataProcessStorageError("unsupported source storage reference")
+        if parsed.query or parsed.fragment or "\\" in parsed.path:
+            raise DataProcessStorageError("unsafe source storage reference")
+        raw_parts = parsed.path.lstrip("/").split("/")
+        if len(raw_parts) != 4:
+            raise DataProcessStorageError("unsafe source storage reference")
+        if any(re.search(r"%(?![0-9A-Fa-f]{2})", part) for part in raw_parts):
+            raise DataProcessStorageError("unsafe source storage reference")
+        try:
+            decoded = [unquote(part, encoding="utf-8", errors="strict") for part in raw_parts]
+        except UnicodeDecodeError as exc:
+            raise DataProcessStorageError("unsafe source storage reference") from exc
+        if any("/" in part or "\\" in part for part in decoded):
+            raise DataProcessStorageError("unsafe source storage reference")
+        canonical_parts = [
+            quote(decoded[0], safe="-_."),
+            quote(decoded[1], safe="-_."),
+            quote(decoded[2], safe="-_."),
+            quote(decoded[3], safe=""),
+        ]
+        if canonical_parts != raw_parts:
+            raise DataProcessStorageError("source storage reference is not canonical")
+        task_id = _safe_component(decoded[0], "task id")
+        source_file_id = _safe_component(decoded[1], "source file id")
+        version_component = decoded[2]
+        if not version_component.startswith("v") or not version_component[1:].isdigit():
+            raise DataProcessStorageError("invalid source file version")
+        version = int(version_component[1:])
+        if version < 1:
+            raise DataProcessStorageError("invalid source file version")
+        basename = _safe_basename(decoded[3])
+        return PurePosixPath(task_id, source_file_id, f"v{version}", basename)
+
+    def _path_for_relative(self, relative_path: PurePosixPath) -> Path:
+        if relative_path.is_absolute() or any(
+            part in {"", ".", ".."} for part in relative_path.parts
+        ):
+            raise DataProcessStorageError("storage path escapes the configured root")
+        path = self._root.joinpath(*relative_path.parts)
+        self._assert_controlled_parent(path)
+        return path
+
+    def _validate_staged_object(
+        self,
+        item: StagedSourceObject,
+        *,
+        require_file: bool,
+    ) -> None:
+        if not isinstance(item, StagedSourceObject):
+            raise DataProcessStorageError("invalid staged source object")
+        if self._issued_staged_objects.get(item._temporary_path) is not item:
+            raise DataProcessStorageError("staged source object was not issued by this storage")
+        expected_relative = self._relative_from_reference(item.reference)
+        if expected_relative is None or expected_relative != item._relative_path:
+            raise DataProcessStorageError("staged source object reference mismatch")
+        staging_root = self._root / ".staging"
+        try:
+            relative_temporary = item._temporary_path.relative_to(staging_root)
+        except ValueError as exc:
+            raise DataProcessStorageError("staged source object escapes staging") from exc
+        if len(relative_temporary.parts) != 2:
+            raise DataProcessStorageError("invalid staged source object path")
+        _safe_component(relative_temporary.parts[0], "batch id")
+        _safe_basename(relative_temporary.parts[1])
+        self._assert_controlled_parent(item._temporary_path)
+        try:
+            info = item._temporary_path.lstat()
+        except FileNotFoundError:
+            if require_file:
+                raise DataProcessStorageError("staged source object does not exist") from None
+            return
+        if stat.S_ISLNK(info.st_mode) or not stat.S_ISREG(info.st_mode):
+            raise DataProcessStorageError("staged source object is not a regular file")
+
+    def _ensure_directory(self, directory: Path) -> Path:
+        try:
+            relative = directory.relative_to(self._root)
+        except ValueError as exc:
+            raise DataProcessStorageError("storage path escapes the configured root") from exc
+        current = self._root
+        for component in relative.parts:
+            current = current / component
+            try:
+                current.mkdir(mode=0o700)
+            except FileExistsError:
+                pass
+            info = current.lstat()
+            if stat.S_ISLNK(info.st_mode) or not stat.S_ISDIR(info.st_mode):
+                raise DataProcessStorageError("storage path contains a symlink or non-directory")
+        return directory
+
+    def _assert_controlled_parent(self, path: Path) -> None:
+        try:
+            relative_parent = path.parent.relative_to(self._root)
+        except ValueError as exc:
+            raise DataProcessStorageError("storage path escapes the configured root") from exc
+        current = self._root
+        for component in relative_parent.parts:
+            current = current / component
+            if not current.exists():
+                continue
+            info = current.lstat()
+            if stat.S_ISLNK(info.st_mode) or not stat.S_ISDIR(info.st_mode):
+                raise DataProcessStorageError("storage path contains a symlink or non-directory")
+
+    @staticmethod
+    def _fsync_directory(directory: Path) -> None:
+        descriptor = os.open(directory, os.O_RDONLY)
+        try:
+            os.fsync(descriptor)
+        finally:
+            os.close(descriptor)
+
+    def _remove_empty_directory(self, directory: Path) -> None:
+        if directory in {self._root, self._root / ".staging"}:
+            return
+        self._assert_controlled_parent(directory / "placeholder")
+        try:
+            directory.rmdir()
+        except (FileNotFoundError, OSError):
+            return
+
+
+@lru_cache
+def get_data_process_storage() -> LocalDataProcessStorage:
+    return LocalDataProcessStorage()
+
+
+__all__ = [
+    "DataProcessStorageError",
+    "LocalDataProcessStorage",
+    "StagedSourceObject",
+    "get_data_process_storage",
+]

backend/app/schemas/data_process.py

@@ -0,0 +1,373 @@
+from __future__ import annotations
+
+from enum import StrEnum
+from typing import Any, Literal
+
+from pydantic import BaseModel, ConfigDict, Field, field_validator, model_validator
+
+from app.modules.data_process.constants import MAX_QA_PAIRS_PER_ITEM
+
+
+def _config_value(config: dict[str, Any], snake_name: str, camel_name: str, default: Any) -> Any:
+    if snake_name in config:
+        return config[snake_name]
+    return config.get(camel_name, default)
+
+
+def _validate_process_config(config: dict[str, Any]) -> None:
+    chunk_method = _config_value(config, "chunk_method", "chunkMethod", "layout_hybrid")
+    if not isinstance(chunk_method, str) or chunk_method not in {
+        "layout_hybrid",
+        "semantic",
+        "fixed",
+    }:
+        raise ValueError("chunk_method must be one of: layout_hybrid, semantic, fixed")
+
+    semantic_percentile = _config_value(
+        config,
+        "semantic_breakpoint_percentile",
+        "semanticBreakpointPercentile",
+        95,
+    )
+    if (
+        isinstance(semantic_percentile, bool)
+        or not isinstance(semantic_percentile, int)
+        or not 1 <= semantic_percentile <= 99
+    ):
+        raise ValueError("semantic_breakpoint_percentile must be an integer in [1, 99]")
+
+    split = _config_value(config, "dataset_split", "datasetSplit", None)
+    if split is not None:
+        if not isinstance(split, dict) or set(split) != {"train", "validation", "test"}:
+            raise ValueError("dataset_split must contain train, validation and test")
+        values = list(split.values())
+        if any(isinstance(value, bool) or not isinstance(value, int) for value in values):
+            raise ValueError("dataset_split values must be integers")
+        if any(value < 0 or value > 100 for value in values) or sum(values) != 100:
+            raise ValueError("dataset_split values must be in [0, 100] and total 100")
+
+    chunk_fields = {
+        "chunk_size",
+        "chunkSize",
+        "chunk_overlap",
+        "chunkOverlap",
+        "min_chunk_size",
+        "minChunkSize",
+    }
+    if chunk_fields.intersection(config):
+        chunk_size = _config_value(config, "chunk_size", "chunkSize", 800)
+        overlap = _config_value(config, "chunk_overlap", "chunkOverlap", 100)
+        minimum = _config_value(config, "min_chunk_size", "minChunkSize", 100)
+        if any(
+            isinstance(value, bool) or not isinstance(value, int)
+            for value in (chunk_size, overlap, minimum)
+        ):
+            raise ValueError("chunk_size, chunk_overlap and min_chunk_size must be integers")
+        if not 16 <= chunk_size <= 32_768:
+            raise ValueError("chunk_size must be in [16, 32768]")
+        if overlap < 0 or overlap >= chunk_size:
+            raise ValueError("chunk_overlap must be in [0, chunk_size)")
+        if minimum <= 0 or minimum > chunk_size or overlap + minimum > chunk_size:
+            raise ValueError("min_chunk_size and chunk_overlap exceed chunk_size")
+
+    temperature = _config_value(config, "temperature", "temperature", None)
+    if temperature is not None:
+        if isinstance(temperature, bool) or not isinstance(temperature, (int, float)):
+            raise ValueError("temperature must be a number")
+        if not 0 <= float(temperature) <= 2:
+            raise ValueError("temperature must be in [0, 2]")
+
+    max_tokens = _config_value(config, "max_tokens", "maxTokens", None)
+    if max_tokens is not None:
+        if isinstance(max_tokens, bool) or not isinstance(max_tokens, int):
+            raise ValueError("max_tokens must be an integer")
+        if not 1 <= max_tokens <= 32_768:
+            raise ValueError("max_tokens must be in [1, 32768]")
+
+    for snake_name, camel_name in (
+        ("qa_pairs_per_row", "qaPairsPerRow"),
+        ("qa_pairs_per_chunk", "qaPairsPerChunk"),
+    ):
+        pairs = _config_value(config, snake_name, camel_name, None)
+        if pairs is None:
+            continue
+        if (
+            isinstance(pairs, bool)
+            or not isinstance(pairs, int)
+            or not 1 <= pairs <= MAX_QA_PAIRS_PER_ITEM
+        ):
+            raise ValueError(
+                f"{snake_name} must be an integer in [1, {MAX_QA_PAIRS_PER_ITEM}]"
+            )
+
+
+class DataProcessStatus(StrEnum):
+    pending = "pending"
+    running = "running"
+    completed = "completed"
+    failed = "failed"
+    stopped = "stopped"
+
+
+class DataProcessWorkflowStep(StrEnum):
+    create = "create"
+    model = "model"
+    upload = "upload"
+    preview = "preview"
+    generate = "generate"
+    results = "results"
+
+
+class DataProcessPreviewStatus(StrEnum):
+    idle = "idle"
+    queued = "queued"
+    running = "running"
+    completed = "completed"
+    failed = "failed"
+    cancelled = "cancelled"
+
+
+class ProcessType(StrEnum):
+    structured = "structured"
+    unstructured = "unstructured"
+    external = "external"
+
+
+class DataProcessTaskCreate(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    name: str = Field(min_length=1, max_length=150)
+    description: str = ""
+    process_type: ProcessType
+    source_dataset_id: str | None = None
+    config: dict[str, Any] = Field(default_factory=dict)
+
+    @field_validator("name")
+    @classmethod
+    def normalize_name(cls, value: str) -> str:
+        value = value.strip()
+        if not value:
+            raise ValueError("task name cannot be empty")
+        return value
+
+    @model_validator(mode="after")
+    def validate_config(self) -> "DataProcessTaskCreate":
+        _validate_process_config(self.config)
+        return self
+
+
+class DataProcessTaskUpdate(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    name: str | None = Field(default=None, min_length=1, max_length=150)
+    description: str | None = None
+    process_type: ProcessType | None = None
+    source_dataset_id: str | None = None
+    config: dict[str, Any] | None = None
+
+    @field_validator("name")
+    @classmethod
+    def normalize_name(cls, value: str | None) -> str | None:
+        if value is None:
+            return None
+        value = value.strip()
+        if not value:
+            raise ValueError("task name cannot be empty")
+        return value
+
+    @model_validator(mode="after")
+    def validate_config(self) -> "DataProcessTaskUpdate":
+        if self.config is not None:
+            _validate_process_config(self.config)
+        return self
+
+
+class DataProcessWorkflowStepUpdate(BaseModel):
+    """仅保存创建向导位置，不修改配置或使下游产物失效。"""
+
+    model_config = ConfigDict(extra="forbid")
+
+    workflow_step: DataProcessWorkflowStep
+
+
+class DataProcessRegenerateRequest(BaseModel):
+    """以一份完整配置准备任务重新生成。
+
+    ``expected_updated_at`` 用于防止详情页的旧快照覆盖其他人刚刚
+    保存的配置。重新生成不允许改变处理类型，避免旧源文件在新解析
+    规则下被静默误用。
+    """
+
+    model_config = ConfigDict(extra="forbid")
+
+    name: str = Field(min_length=1, max_length=150)
+    description: str
+    process_type: ProcessType
+    config: dict[str, Any]
+    expected_updated_at: str = Field(min_length=1)
+
+    @field_validator("name")
+    @classmethod
+    def normalize_name(cls, value: str) -> str:
+        value = value.strip()
+        if not value:
+            raise ValueError("task name cannot be empty")
+        return value
+
+    @model_validator(mode="after")
+    def validate_config(self) -> "DataProcessRegenerateRequest":
+        _validate_process_config(self.config)
+        return self
+
+
+class PreviewBuildRequest(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    replace_existing: Literal[True] = True
+    source_file_ids: list[str] | None = None
+    source_file_id: str | None = None
+
+    @model_validator(mode="after")
+    def validate_source_file_selection(self) -> "PreviewBuildRequest":
+        if self.source_file_ids is not None and self.source_file_id is not None:
+            raise ValueError("source_file_id and source_file_ids cannot be used together")
+        values = self.source_file_ids
+        if values is None and self.source_file_id is not None:
+            values = [self.source_file_id]
+        if values is None:
+            return self
+        normalized = list(dict.fromkeys(str(value).strip() for value in values))
+        if not normalized or any(not value for value in normalized):
+            raise ValueError("at least one non-empty source file id is required")
+        self.source_file_ids = normalized
+        self.source_file_id = None
+        return self
+
+
+class PreviewItemCreate(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    source_file_id: str | None = None
+    original_content: str = ""
+    edited_content: str = ""
+    source_start: int | None = Field(default=None, ge=0)
+    source_end: int | None = Field(default=None, ge=0)
+    source_start_line: int | None = Field(default=None, ge=1)
+    source_end_line: int | None = Field(default=None, ge=1)
+
+    @model_validator(mode="after")
+    def validate_ranges(self) -> "PreviewItemCreate":
+        if self.source_start is not None and self.source_end is not None:
+            if self.source_end < self.source_start:
+                raise ValueError("source_end must be greater than or equal to source_start")
+        if self.source_start_line is not None and self.source_end_line is not None:
+            if self.source_end_line < self.source_start_line:
+                raise ValueError(
+                    "source_end_line must be greater than or equal to source_start_line"
+                )
+        return self
+
+
+class PreviewItemUpdate(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    edited_content: str
+    expected_updated_at: str | None = None
+
+
+class GenerateRequest(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    replace_existing: Literal[True] = True
+
+
+class ExternalSourceRequest(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    type: str = Field(min_length=1, max_length=30)
+    url: str = Field(min_length=1, max_length=2048)
+    auth_mode: Literal["none", "basic"] = "none"
+    username: str | None = Field(default=None, max_length=150)
+    password: str | None = Field(default=None, max_length=500)
+    limit: int = Field(default=1000, ge=1, le=100_000)
+
+
+class ExternalPullRequest(ExternalSourceRequest):
+    query: str | None = Field(default=None, max_length=20_000)
+    file_name: str = Field(default="external-data.jsonl", min_length=1, max_length=255)
+
+    @field_validator("file_name")
+    @classmethod
+    def validate_file_name(cls, value: str) -> str:
+        name = value.strip()
+        if not name.lower().endswith((".jsonl", ".ndjson")):
+            raise ValueError("external pull file_name must end with .jsonl or .ndjson")
+        return name
+
+
+class ResultUpdate(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    instruction: str | None = None
+    input: str | None = None
+    output: str | None = None
+    expected_updated_at: str | None = None
+
+
+class ResultRegenerateRequest(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    expected_updated_at: str = Field(min_length=1, max_length=100)
+
+
+class ResultBatchRegenerateItem(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    result_id: str = Field(min_length=1, max_length=100)
+    expected_updated_at: str = Field(min_length=1, max_length=100)
+
+
+class ResultBatchRegenerateRequest(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    items: list[ResultBatchRegenerateItem] = Field(min_length=1, max_length=100)
+
+    @model_validator(mode="after")
+    def validate_unique_results(self) -> "ResultBatchRegenerateRequest":
+        result_ids = [item.result_id for item in self.items]
+        if len(result_ids) != len(set(result_ids)):
+            raise ValueError("result_id values must be unique")
+        return self
+
+
+class DatasetSplit(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    train: int = Field(default=80, ge=0, le=100)
+    validation: int = Field(default=10, ge=0, le=100)
+    test: int = Field(default=10, ge=0, le=100)
+
+    @model_validator(mode="after")
+    def validate_total(self) -> "DatasetSplit":
+        if self.train + self.validation + self.test != 100:
+            raise ValueError("dataset split must total 100")
+        return self
+
+
+class PublishRequest(BaseModel):
+    model_config = ConfigDict(extra="forbid")
+
+    dataset_name: str = Field(min_length=1, max_length=150)
+    dataset_type: Literal["train", "test", "eval", "val", "other"] = "train"
+    storage_type: Literal["local"] = "local"
+    split: DatasetSplit = Field(default_factory=DatasetSplit)
+    format: Literal["alpaca_jsonl", "jsonl"] = "alpaca_jsonl"
+    description: str = ""
+
+    @field_validator("dataset_name")
+    @classmethod
+    def normalize_dataset_name(cls, value: str) -> str:
+        value = value.strip()
+        if not value:
+            raise ValueError("dataset name cannot be empty")
+        return value

backend/pyproject.toml

@@ -16,6 +16,14 @@ dependencies = [
   "PyJWT>=2.8.0",
   "passlib[bcrypt]>=1.7.4",
   "python-dotenv>=1.0.1",
+  "pypdf[crypto]>=5.0.0",
+  "python-docx>=1.1.2",
+  "openpyxl>=3.1.5",
+  "python-pptx>=1.0.2",
+  "llama-index-core==0.14.23",
+  "llama-index-embeddings-huggingface==0.6.1",
+  "docling==2.115.0",
+  "tiktoken>=0.7.0",
 ]
 
 [project.optional-dependencies]

backend/requirements.txt

@@ -10,3 +10,11 @@ httpx>=0.27.0
 PyJWT>=2.8.0
 passlib[bcrypt]>=1.7.4
 python-dotenv>=1.0.1
+pypdf[crypto]>=5.0.0
+python-docx>=1.1.2
+openpyxl>=3.1.5
+python-pptx>=1.0.2
+llama-index-core==0.14.23
+llama-index-embeddings-huggingface==0.6.1
+docling==2.115.0
+tiktoken>=0.7.0

backend/storage/data-process/dpt_3ed8684864a34688b8ee/dpsf_73355fb704644768b557/v1/111.json

@@ -0,0 +1,16 @@
+ [   {
+        "instruction": "数字钱包的“零余额”管理在日末结算时有何具体要求？",
+        "input": "",
+        "output": "数字钱包严格执行“零余额”管理，在日末结算时，若结算钱包存在余额，必须将数字货币兑回基本账户，以确保日末数字钱包无余额。"
+    },
+    {
+        "instruction": "用印登记表中包含哪些具体的印章类型？",
+        "input": "",
+        "output": "根据用印登记表的内容显示，表中明确列出了五种具体的印章类型，分别是财务专用章、法人名章、法人授权人名章1、法人授权人名章2以及其他。这些印章类型被详细划分在“用印数量”这一栏目下，用于记录不同印章的具体使用情况。"
+    },
+    {
+        "instruction": "密钥交接记录表中需要哪些角色签字确认？",
+        "input": "",
+        "output": "密钥交接记录表中需要三个角色的签字确认，分别是交接人签字、接交人签字以及监交人签字。交接人负责移交密钥，接交人负责接收密钥，而监交人则负责对整个密钥交接过程进行监督，这三个角色的共同签字确认能够确保密钥交接流程的规范性与安全性。"
+    }
+]

compute/requirements.txt

@@ -1,5 +1,6 @@
 fastapi>=0.111.0
 uvicorn[standard]>=0.30.0
+python-multipart>=0.0.9
 pydantic>=2.7.0
 python-dotenv>=1.0.1
 httpx>=0.27.0

devserver.err

@@ -1,47 +0,0 @@
-Traceback (most recent call last):
-  File "<frozen runpy>", line 198, in _run_module_as_main
-  File "<frozen runpy>", line 88, in _run_code
-  File "/mnt/e/yg_ft/projects/backend/venv/lib/python3.12/site-packages/uvicorn/__main__.py", line 4, in <module>
-    uvicorn.main()
-  File "/mnt/e/yg_ft/projects/backend/venv/lib/python3.12/site-packages/click/core.py", line 1569, in __call__
-    return self.main(*args, **kwargs)
-           ^^^^^^^^^^^^^^^^^^^^^^^^^^
-  File "/mnt/e/yg_ft/projects/backend/venv/lib/python3.12/site-packages/click/core.py", line 1490, in main
-    rv = self.invoke(ctx)
-         ^^^^^^^^^^^^^^^^
-  File "/mnt/e/yg_ft/projects/backend/venv/lib/python3.12/site-packages/click/core.py", line 1353, in invoke
-    return ctx.invoke(self.callback, **ctx.params)
-           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-  File "/mnt/e/yg_ft/projects/backend/venv/lib/python3.12/site-packages/click/core.py", line 907, in invoke
-    return callback(*args, **kwargs)
-           ^^^^^^^^^^^^^^^^^^^^^^^^^
-  File "/mnt/e/yg_ft/projects/backend/venv/lib/python3.12/site-packages/uvicorn/main.py", line 440, in main
-    run(
-  File "/mnt/e/yg_ft/projects/backend/venv/lib/python3.12/site-packages/uvicorn/main.py", line 609, in run
-    config.load_app()
-  File "/mnt/e/yg_ft/projects/backend/venv/lib/python3.12/site-packages/uvicorn/config.py", line 427, in load_app
-    return import_from_string(self.app)
-           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-  File "/mnt/e/yg_ft/projects/backend/venv/lib/python3.12/site-packages/uvicorn/importer.py", line 22, in import_from_string
-    raise exc from None
-  File "/mnt/e/yg_ft/projects/backend/venv/lib/python3.12/site-packages/uvicorn/importer.py", line 19, in import_from_string
-    module = importlib.import_module(module_str)
-             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-  File "/usr/lib/python3.12/importlib/__init__.py", line 90, in import_module
-    return _bootstrap._gcd_import(name[level:], package, level)
-           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-  File "<frozen importlib._bootstrap>", line 1387, in _gcd_import
-  File "<frozen importlib._bootstrap>", line 1360, in _find_and_load
-  File "<frozen importlib._bootstrap>", line 1331, in _find_and_load_unlocked
-  File "<frozen importlib._bootstrap>", line 935, in _load_unlocked
-  File "<frozen importlib._bootstrap_external>", line 995, in exec_module
-  File "<frozen importlib._bootstrap>", line 488, in _call_with_frames_removed
-  File "/mnt/e/yg_ft/backend/app/main.py", line 4, in <module>
-    from app.api.v1.router import api_router
-  File "/mnt/e/yg_ft/backend/app/api/v1/router.py", line 3, in <module>
-    from app.api.v1.endpoints.platform import router as platform_router
-  File "/mnt/e/yg_ft/backend/app/api/v1/endpoints/platform.py", line 10, in <module>
-    from app.db.platform_store import get_platform_store
-  File "/mnt/e/yg_ft/backend/app/db/platform_store.py", line 15, in <module>
-    import psycopg
-ModuleNotFoundError: No module named 'psycopg'

frontend/package-lock.json

@@ -16,6 +16,7 @@
         "element-plus": "^2.9.1",
         "marked": "^15.0.5",
         "md-editor-v3": "^5.1.4",
+        "pdfjs-dist": "^6.2.108",
         "pinia": "^2.3.0",
         "vue": "^3.5.13",
         "vue-echarts": "^8.0.1",
@@ -1213,6 +1214,256 @@
       "integrity": "sha512-FY+MKLBoTsLNJF/eLWaOsXGdz6uh3Iu1axjPf6TUq92IYumcTcXWHoS747JARLkcdlJ/Waiaxc5wQfFO8jC6NA==",
       "license": "MIT"
     },
+    "node_modules/@napi-rs/canvas": {
+      "version": "1.0.3",
+      "resolved": "https://registry.npmmirror.com/@napi-rs/canvas/-/canvas-1.0.3.tgz",
+      "integrity": "sha512-OlI657a5XXvKGFX7kNeIzJ8rO7IXt87Mqu2H8rXE46viAuOfum/JA7ysX7+eBhxNKznT+RCZh418mndlcFX3+w==",
+      "license": "MIT",
+      "optional": true,
+      "workspaces": [
+        "e2e/*"
+      ],
+      "engines": {
+        "node": ">= 10"
+      },
+      "funding": {
+        "type": "github",
+        "url": "https://github.com/sponsors/Brooooooklyn"
+      },
+      "optionalDependencies": {
+        "@napi-rs/canvas-android-arm64": "1.0.3",
+        "@napi-rs/canvas-darwin-arm64": "1.0.3",
+        "@napi-rs/canvas-darwin-x64": "1.0.3",
+        "@napi-rs/canvas-linux-arm-gnueabihf": "1.0.3",
+        "@napi-rs/canvas-linux-arm64-gnu": "1.0.3",
+        "@napi-rs/canvas-linux-arm64-musl": "1.0.3",
+        "@napi-rs/canvas-linux-riscv64-gnu": "1.0.3",
+        "@napi-rs/canvas-linux-x64-gnu": "1.0.3",
+        "@napi-rs/canvas-linux-x64-musl": "1.0.3",
+        "@napi-rs/canvas-win32-arm64-msvc": "1.0.3",
+        "@napi-rs/canvas-win32-x64-msvc": "1.0.3"
+      }
+    },
+    "node_modules/@napi-rs/canvas-android-arm64": {
+      "version": "1.0.3",
+      "resolved": "https://registry.npmmirror.com/@napi-rs/canvas-android-arm64/-/canvas-android-arm64-1.0.3.tgz",
+      "integrity": "sha512-7kSCdUhoXiO+AaIMXdBGdtp6EctZNkmF62Rea/BmVQlwKaM3bBhOzyGUzxyxz9dv5vdBfpyAaxhSRSJF4kqK4A==",
+      "cpu": [
+        "arm64"
+      ],
+      "license": "MIT",
+      "optional": true,
+      "os": [
+        "android"
+      ],
+      "engines": {
+        "node": ">= 10"
+      },
+      "funding": {
+        "type": "github",
+        "url": "https://github.com/sponsors/Brooooooklyn"
+      }
+    },
+    "node_modules/@napi-rs/canvas-darwin-arm64": {
+      "version": "1.0.3",
+      "resolved": "https://registry.npmmirror.com/@napi-rs/canvas-darwin-arm64/-/canvas-darwin-arm64-1.0.3.tgz",
+      "integrity": "sha512-ds14V1BPagLszQyaDTeggny5fNeTCqsUQ5QhFj9VDxSEfzrVxXtdbR0LoFyKa0Siaaw8KvqSk4t7k/WoZJwvbg==",
+      "cpu": [
+        "arm64"
+      ],
+      "license": "MIT",
+      "optional": true,
+      "os": [
+        "darwin"
+      ],
+      "engines": {
+        "node": ">= 10"
+      },
+      "funding": {
+        "type": "github",
+        "url": "https://github.com/sponsors/Brooooooklyn"
+      }
+    },
+    "node_modules/@napi-rs/canvas-darwin-x64": {
+      "version": "1.0.3",
+      "resolved": "https://registry.npmmirror.com/@napi-rs/canvas-darwin-x64/-/canvas-darwin-x64-1.0.3.tgz",
+      "integrity": "sha512-qof3LRAAycmkV2I1izZo9RoSHF8kCQr5O05sFwv0jK8rSdYV6KHVwimo6Qb7RxZj40WHKbLHm5JDaUF0o5XUAA==",
+      "cpu": [
+        "x64"
+      ],
+      "license": "MIT",
+      "optional": true,
+      "os": [
+        "darwin"
+      ],
+      "engines": {
+        "node": ">= 10"
+      },
+      "funding": {
+        "type": "github",
+        "url": "https://github.com/sponsors/Brooooooklyn"
+      }
+    },
+    "node_modules/@napi-rs/canvas-linux-arm-gnueabihf": {
+      "version": "1.0.3",
+      "resolved": "https://registry.npmmirror.com/@napi-rs/canvas-linux-arm-gnueabihf/-/canvas-linux-arm-gnueabihf-1.0.3.tgz",
+      "integrity": "sha512-FU2kKZLmolHA9+KcUA+l1+xH3WTLUUTQDU/kLv9SEUr2TrRPu94aytOeizFJDHPs/QBcw4QL1mCQhetQXYBbag==",
+      "cpu": [
+        "arm"
+      ],
+      "license": "MIT",
+      "optional": true,
+      "os": [
+        "linux"
+      ],
+      "engines": {
+        "node": ">= 10"
+      },
+      "funding": {
+        "type": "github",
+        "url": "https://github.com/sponsors/Brooooooklyn"
+      }
+    },
+    "node_modules/@napi-rs/canvas-linux-arm64-gnu": {
+      "version": "1.0.3",
+      "resolved": "https://registry.npmmirror.com/@napi-rs/canvas-linux-arm64-gnu/-/canvas-linux-arm64-gnu-1.0.3.tgz",
+      "integrity": "sha512-GVSjntxKeA+/y/ZKf1F+cmUw1WeIkE5aMRPqnZUlBTBvBcrvgWccJAWuYCKPX4QJQwZILIIwhgdAbl51yj6fpA==",
+      "cpu": [
+        "arm64"
+      ],
+      "license": "MIT",
+      "optional": true,
+      "os": [
+        "linux"
+      ],
+      "engines": {
+        "node": ">= 10"
+      },
+      "funding": {
+        "type": "github",
+        "url": "https://github.com/sponsors/Brooooooklyn"
+      }
+    },
+    "node_modules/@napi-rs/canvas-linux-arm64-musl": {
+      "version": "1.0.3",
+      "resolved": "https://registry.npmmirror.com/@napi-rs/canvas-linux-arm64-musl/-/canvas-linux-arm64-musl-1.0.3.tgz",
+      "integrity": "sha512-J51oK/axyZ13kxycumSMfLiDZMdWdOVvqDFI28BpuViZHE3A0bQfr8B5vg8YnPEnqLD3BSn1hkdlh2buspEcNQ==",
+      "cpu": [
+        "arm64"
+      ],
+      "license": "MIT",
+      "optional": true,
+      "os": [
+        "linux"
+      ],
+      "engines": {
+        "node": ">= 10"
+      },
+      "funding": {
+        "type": "github",
+        "url": "https://github.com/sponsors/Brooooooklyn"
+      }
+    },
+    "node_modules/@napi-rs/canvas-linux-riscv64-gnu": {
+      "version": "1.0.3",
+      "resolved": "https://registry.npmmirror.com/@napi-rs/canvas-linux-riscv64-gnu/-/canvas-linux-riscv64-gnu-1.0.3.tgz",
+      "integrity": "sha512-CtQgQjoVTX67jS9XuCTtJ40Sl7wRLMguoFnnGnfDmCWf7kzKFZVwj5ynqUOIGKFMSB61ZCuQlwPvVNxYTTseaw==",
+      "cpu": [
+        "riscv64"
+      ],
+      "license": "MIT",
+      "optional": true,
+      "os": [
+        "linux"
+      ],
+      "engines": {
+        "node": ">= 10"
+      },
+      "funding": {
+        "type": "github",
+        "url": "https://github.com/sponsors/Brooooooklyn"
+      }
+    },
+    "node_modules/@napi-rs/canvas-linux-x64-gnu": {
+      "version": "1.0.3",
+      "resolved": "https://registry.npmmirror.com/@napi-rs/canvas-linux-x64-gnu/-/canvas-linux-x64-gnu-1.0.3.tgz",
+      "integrity": "sha512-jtfzAHFp+FRaR7zGT4jyCe6wUgAG/dVb5A4Apd8FY9jKarntDfUAlJXscugiH7ZF5kKnu7/lHFk9LaDPcrGEVQ==",
+      "cpu": [
+        "x64"
+      ],
+      "license": "MIT",
+      "optional": true,
+      "os": [
+        "linux"
+      ],
+      "engines": {
+        "node": ">= 10"
+      },
+      "funding": {
+        "type": "github",
+        "url": "https://github.com/sponsors/Brooooooklyn"
+      }
+    },
+    "node_modules/@napi-rs/canvas-linux-x64-musl": {
+      "version": "1.0.3",
+      "resolved": "https://registry.npmmirror.com/@napi-rs/canvas-linux-x64-musl/-/canvas-linux-x64-musl-1.0.3.tgz",
+      "integrity": "sha512-xTzaUCKUHTY4bCGadeeRZggbRVbGUT1petg7Z8r9AJR2+D9Bqu6nQAgqBGC6D47tA70LjaaaLTrJ7wNY1T74dg==",
+      "cpu": [
+        "x64"
+      ],
+      "license": "MIT",
+      "optional": true,
+      "os": [
+        "linux"
+      ],
+      "engines": {
+        "node": ">= 10"
+      },
+      "funding": {
+        "type": "github",
+        "url": "https://github.com/sponsors/Brooooooklyn"
+      }
+    },
+    "node_modules/@napi-rs/canvas-win32-arm64-msvc": {
+      "version": "1.0.3",
+      "resolved": "https://registry.npmmirror.com/@napi-rs/canvas-win32-arm64-msvc/-/canvas-win32-arm64-msvc-1.0.3.tgz",
+      "integrity": "sha512-ktVLuBkI6QVOm5BwO/WbdGwxgeetAMJa7TTmR8qBarXF0OU2NKjvjUtPJAl2y8t+zBRczJl/1VOl9gua6WcK2g==",
+      "cpu": [
+        "arm64"
+      ],
+      "license": "MIT",
+      "optional": true,
+      "os": [
+        "win32"
+      ],
+      "engines": {
+        "node": ">= 10"
+      },
+      "funding": {
+        "type": "github",
+        "url": "https://github.com/sponsors/Brooooooklyn"
+      }
+    },
+    "node_modules/@napi-rs/canvas-win32-x64-msvc": {
+      "version": "1.0.3",
+      "resolved": "https://registry.npmmirror.com/@napi-rs/canvas-win32-x64-msvc/-/canvas-win32-x64-msvc-1.0.3.tgz",
+      "integrity": "sha512-SGhlQ8bDjL1Cz2KnsKMasr/5sTcwG/SZkB6WCJxLsmSm/3aS2C+3p39bA7iZ2/94+NkVDySZfbiGoaSZSFHYxA==",
+      "cpu": [
+        "x64"
+      ],
+      "license": "MIT",
+      "optional": true,
+      "os": [
+        "win32"
+      ],
+      "engines": {
+        "node": ">= 10"
+      },
+      "funding": {
+        "type": "github",
+        "url": "https://github.com/sponsors/Brooooooklyn"
+      }
+    },
     "node_modules/@nodelib/fs.scandir": {
       "version": "2.1.5",
       "resolved": "https://registry.npmjs.org/@nodelib/fs.scandir/-/fs.scandir-2.1.5.tgz",
@@ -3443,6 +3694,18 @@
       "dev": true,
       "license": "MIT"
     },
+    "node_modules/pdfjs-dist": {
+      "version": "6.2.108",
+      "resolved": "https://registry.npmmirror.com/pdfjs-dist/-/pdfjs-dist-6.2.108.tgz",
+      "integrity": "sha512-YxFb+SQcodN2rnX9Tn3dHYlqfb7NjlzzfONPpJd+AKoKtUjEdevTfbC07d5TcczzOK6261auRkP/M8OBHs9vFQ==",
+      "license": "Apache-2.0",
+      "engines": {
+        "node": ">=22.13.0 || >=24"
+      },
+      "optionalDependencies": {
+        "@napi-rs/canvas": "^1.0.0"
+      }
+    },
     "node_modules/picocolors": {
       "version": "1.1.1",
       "resolved": "https://registry.npmjs.org/picocolors/-/picocolors-1.1.1.tgz",

frontend/package.json

@@ -36,6 +36,7 @@
     "element-plus": "^2.9.1",
     "marked": "^15.0.5",
     "md-editor-v3": "^5.1.4",
+    "pdfjs-dist": "^6.2.108",
     "pinia": "^2.3.0",
     "vue": "^3.5.13",
     "vue-echarts": "^8.0.1",

frontend/src/api/modules/dataProcess.ts

@@ -0,0 +1,306 @@
+import { del, get, post, put } from '../request'
+import type {
+  DataProcessExternalSourcePayload,
+  DataProcessExternalTestResult,
+  DataProcessPage,
+  DataProcessOfficePreview,
+  DataProcessPdfPages,
+  DataProcessPreviewBuildPayload,
+  DataProcessPreviewBuildResult,
+  DataProcessPreviewProgress,
+  DataProcessPreviewCreatePayload,
+  DataProcessPreviewItem,
+  DataProcessPreviewUpdatePayload,
+  DataProcessProgress,
+  DataProcessRegeneratePayload,
+  DataProcessRegenerateResult,
+  DataProcessPublishPayload,
+  DataProcessPublishResult,
+  DataProcessQualityScore,
+  DataProcessResult,
+  DataProcessResultBatchRegeneratePayload,
+  DataProcessResultBatchRegenerateResult,
+  DataProcessResultRegeneratePayload,
+  DataProcessResultUpdatePayload,
+  DataProcessSourceContent,
+  DataProcessSourceFile,
+  DataProcessTask,
+  DataProcessTaskCreatePayload,
+  DataProcessTaskUpdatePayload,
+  DataProcessWorkflowStep,
+} from '@/types/dataProcess'
+
+export type {
+  DataProcessConfig,
+  DataProcessDatasetSplit,
+  DataProcessExternalSourcePayload,
+  DataProcessExternalTestResult,
+  DataProcessPage,
+  DataProcessDocxParagraph,
+  DataProcessDocxPreview,
+  DataProcessDocxTable,
+  DataProcessDocxTableRow,
+  DataProcessOfficePreview,
+  DataProcessPdfPageRange,
+  DataProcessPdfPages,
+  DataProcessPreviewBuildPayload,
+  DataProcessPreviewBuildFileResult,
+  DataProcessPreviewBuildResult,
+  DataProcessPreviewProgress,
+  DataProcessPreviewStatus,
+  DataProcessPreviewFileBuildProgress,
+  DataProcessPreviewFileStatus,
+  DataProcessPreviewCreatePayload,
+  DataProcessPreviewItem,
+  DataProcessPreviewUpdatePayload,
+  DataProcessProgress,
+  DataProcessRegeneratePayload,
+  DataProcessRegenerateResult,
+  DataProcessPublishPayload,
+  DataProcessPublishResult,
+  DataProcessQualityScore,
+  DataProcessResult,
+  DataProcessResultBatchRegenerateFailure,
+  DataProcessResultBatchRegenerateItem,
+  DataProcessResultBatchRegeneratePayload,
+  DataProcessResultBatchRegenerateResult,
+  DataProcessResultRegeneratePayload,
+  DataProcessResultStatus,
+  DataProcessResultUpdatePayload,
+  DataProcessSplit,
+  DataProcessSourceContent,
+  DataProcessSourceFile,
+  DataProcessStatus,
+  DataProcessTask,
+  DataProcessTaskCreatePayload,
+  DataProcessTaskUpdatePayload,
+  DataProcessType,
+  DataProcessWorkflowStep,
+  DataProcessXlsxActiveSheet,
+  DataProcessXlsxPreview,
+  DataProcessXlsxPreviewRow,
+  DataProcessXlsxSheet,
+} from '@/types/dataProcess'
+
+export function getDataProcessTasks(params: {
+  page?: number
+  page_size?: number
+  keyword?: string
+  status?: string
+  process_type?: string
+} = {}) {
+  return get<DataProcessPage<DataProcessTask>>('/data-process', params)
+}
+
+export const getDataProcessTask = (taskId: string | number) =>
+  get<DataProcessTask>(`/data-process/${encodeURIComponent(taskId)}`)
+
+export const createDataProcessTask = (payload: DataProcessTaskCreatePayload) =>
+  post<DataProcessTask>(`/data-process`, payload)
+
+export const updateDataProcessTask = (taskId: string | number, payload: DataProcessTaskUpdatePayload) =>
+  put<DataProcessTask>(`/data-process/${encodeURIComponent(taskId)}`, payload)
+
+export const updateDataProcessWorkflowStep = (
+  taskId: string | number,
+  workflowStep: DataProcessWorkflowStep,
+) => put<DataProcessTask>(
+  `/data-process/${encodeURIComponent(taskId)}/workflow-step`,
+  { workflow_step: workflowStep },
+)
+
+export const regenerateDataProcessTask = (
+  taskId: string | number,
+  payload: DataProcessRegeneratePayload,
+) => post<DataProcessRegenerateResult>(
+  `/data-process/${encodeURIComponent(taskId)}/regenerate`,
+  payload,
+)
+
+export const deleteDataProcessTask = (taskId: string | number) =>
+  del<{ deleted: string | number }>(`/data-process/${encodeURIComponent(taskId)}`)
+
+export function uploadDataProcessSourceFiles(
+  taskId: string | number,
+  files: File[],
+  onProgress?: (progress: number) => void,
+) {
+  const formData = new FormData()
+  files.forEach((file) => formData.append('files', file))
+  return post<{ files: DataProcessSourceFile[] }>(
+    `/data-process/${encodeURIComponent(taskId)}/source-files`,
+    formData,
+    {
+      timeout: 5 * 60 * 1000,
+      onUploadProgress: (event) => {
+        if (!event.total) return
+        // 发送完成不等于服务端解析完成；收到成功响应前最多展示 99%。
+        const progress = Math.round((event.loaded / event.total) * 100)
+        onProgress?.(Math.min(99, Math.max(0, progress)))
+      },
+    },
+  )
+}
+
+export const deleteDataProcessSourceFile = (taskId: string | number, fileId: string | number) =>
+  del<{ deleted: string | number }>(
+    `/data-process/${encodeURIComponent(taskId)}/source-files/${encodeURIComponent(fileId)}`,
+  )
+
+export const getDataProcessSourceContent = (
+  taskId: string | number,
+  fileId: string | number,
+  params: { start_line?: number; line_count?: number } = {},
+) => get<DataProcessSourceContent>(
+  `/data-process/${encodeURIComponent(taskId)}/source-files/${encodeURIComponent(fileId)}/content`,
+  params,
+)
+
+export const getDataProcessSourceRawUrl = (
+  taskId: string | number,
+  fileId: string | number,
+) => `/data-process/${encodeURIComponent(taskId)}/source-files/${encodeURIComponent(fileId)}/raw`
+
+export const getDataProcessPdfPages = (
+  taskId: string | number,
+  fileId: string | number,
+) => get<DataProcessPdfPages>(
+  `/data-process/${encodeURIComponent(taskId)}/source-files/${encodeURIComponent(fileId)}/pdf-pages`,
+)
+
+export const getDataProcessOfficePreview = (
+  taskId: string | number,
+  fileId: string | number,
+  params: { sheet_index?: number; offset?: number; limit?: number } = {},
+) => get<DataProcessOfficePreview>(
+  `/data-process/${encodeURIComponent(taskId)}/source-files/${encodeURIComponent(fileId)}/office-preview`,
+  params,
+  { timeout: 60_000 },
+)
+
+export const testDataProcessExternalSource = (
+  taskId: string | number,
+  payload: DataProcessExternalSourcePayload,
+) => {
+  const { query: _query, file_name: _fileName, ...connection } = payload
+  return post<DataProcessExternalTestResult>(
+    `/data-process/${encodeURIComponent(taskId)}/external/test`,
+    connection,
+  )
+}
+
+export const pullDataProcessExternalSource = (
+  taskId: string | number,
+  payload: DataProcessExternalSourcePayload,
+) => post<{ files: DataProcessSourceFile[] }>(
+  `/data-process/${encodeURIComponent(taskId)}/external/pull`,
+  payload,
+)
+
+export const buildDataProcessPreview = (
+  taskId: string | number,
+  payload: DataProcessPreviewBuildPayload = {},
+) => post<DataProcessPreviewBuildResult>(
+  `/data-process/${encodeURIComponent(taskId)}/preview/build`,
+  payload,
+  { timeout: 5 * 60 * 1000 },
+)
+
+export const startDataProcessPreview = (
+  taskId: string | number,
+  payload: DataProcessPreviewBuildPayload = {},
+) => post<DataProcessPreviewProgress>(
+  `/data-process/${encodeURIComponent(taskId)}/preview/start`,
+  payload,
+)
+
+export const getDataProcessPreviewProgress = (taskId: string | number) =>
+  get<DataProcessPreviewProgress>(
+    `/data-process/${encodeURIComponent(taskId)}/preview/progress`,
+  )
+
+export function getDataProcessPreview(
+  taskId: string | number,
+  params: { source_file_id?: string | number; page?: number; page_size?: number; keyword?: string } = {},
+) {
+  return get<DataProcessPage<DataProcessPreviewItem>>(
+    `/data-process/${encodeURIComponent(taskId)}/preview`,
+    params,
+  )
+}
+
+export const updateDataProcessPreview = (
+  taskId: string | number,
+  previewId: string | number,
+  payload: DataProcessPreviewUpdatePayload,
+) => put<DataProcessPreviewItem>(
+  `/data-process/${encodeURIComponent(taskId)}/preview/${encodeURIComponent(previewId)}`,
+  payload,
+)
+
+export const createDataProcessPreview = (
+  taskId: string | number,
+  payload: DataProcessPreviewCreatePayload,
+) => post<DataProcessPreviewItem>(`/data-process/${encodeURIComponent(taskId)}/preview`, payload)
+
+export const deleteDataProcessPreview = (
+  taskId: string | number,
+  previewId: string | number,
+) => del<Record<string, never>>(
+  `/data-process/${encodeURIComponent(taskId)}/preview/${encodeURIComponent(previewId)}`,
+)
+
+export const generateDataProcess = (taskId: string | number) =>
+  post<DataProcessProgress>(`/data-process/${encodeURIComponent(taskId)}/generate`)
+
+export const getDataProcessProgress = (taskId: string | number) =>
+  get<DataProcessProgress>(`/data-process/${encodeURIComponent(taskId)}/progress`)
+
+export const confirmDataProcessResults = (taskId: string | number) =>
+  post<DataProcessTask>(`/data-process/${encodeURIComponent(taskId)}/confirm-results`)
+
+export function getDataProcessResults(
+  taskId: string | number,
+  params: { page?: number; page_size?: number; keyword?: string; status?: string; split?: string } = {},
+) {
+  return get<DataProcessPage<DataProcessResult>>(
+    `/data-process/${encodeURIComponent(taskId)}/results`,
+    params,
+  )
+}
+
+export const updateDataProcessResult = (
+  taskId: string | number,
+  resultId: string | number,
+  payload: DataProcessResultUpdatePayload,
+) => put<DataProcessResult>(
+  `/data-process/${encodeURIComponent(taskId)}/results/${encodeURIComponent(resultId)}`,
+  payload,
+)
+
+export const restoreDataProcessResult = (taskId: string | number, resultId: string | number) =>
+  post<DataProcessResult>(
+    `/data-process/${encodeURIComponent(taskId)}/results/${encodeURIComponent(resultId)}/restore`,
+  )
+
+export const regenerateDataProcessResult = (
+  taskId: string | number,
+  resultId: string | number,
+  payload: DataProcessResultRegeneratePayload,
+) => post<DataProcessResult>(
+  `/data-process/${encodeURIComponent(taskId)}/results/${encodeURIComponent(resultId)}/regenerate`,
+  payload,
+  { timeout: 210_000 },
+)
+
+export const regenerateDataProcessResults = (
+  taskId: string | number,
+  payload: DataProcessResultBatchRegeneratePayload,
+) => post<DataProcessResultBatchRegenerateResult>(
+  `/data-process/${encodeURIComponent(taskId)}/results/regenerate-batch`,
+  payload,
+  { timeout: 240_000 },
+)
+
+export const publishDataProcess = (taskId: string | number, payload: DataProcessPublishPayload) =>
+  post<DataProcessPublishResult>(`/data-process/${encodeURIComponent(taskId)}/publish`, payload)

frontend/src/api/modules/dataset.ts

@@ -89,10 +89,10 @@ export const downloadFileUrl = (
   fileId: string | number,
   versionId?: string,
 ) => {
-  const baseUrl = `/modelTF/dataset-manage/download/${datasetId}/${fileId}`
+  const baseUrl = `/dataset-manage/download/${datasetId}/${fileId}`
   return versionId ? `${baseUrl}?version_id=${encodeURIComponent(versionId)}` : baseUrl
 }
 
 /** 打包下载数据集 URL */
 export const downloadDatasetUrl = (datasetId: string | number) =>
-  `/modelTF/dataset-manage/download/${datasetId}`
+  `/dataset-manage/download/${datasetId}`

frontend/src/api/modules/model.ts

@@ -45,4 +45,4 @@ export const mergeModel = (data: {
 
 /** 导出已训练模型权重 */
 export const exportModelUrl = (modelName: string) =>
-  `/modelTF/model-manage/trained-models/${encodeURIComponent(modelName)}/export`
+  `/model-manage/trained-models/${encodeURIComponent(modelName)}/export`

frontend/src/router/index.ts

@@ -144,7 +144,19 @@ const routes: RouteRecordRaw[] = [
         path: 'data-process/:id',
         name: 'data-process-detail',
         component: () => import('@/views/data-process/DataProcessDetailView.vue'),
-        meta: { title: '数据处理详情' },
+        meta: { title: '数据处理详情', backRouteName: 'data-process' },
+      },
+      {
+        path: 'data-process/:id/regenerate',
+        name: 'data-process-regenerate',
+        component: () => import('@/views/data-process/DataProcessCreateView.vue'),
+        meta: { title: '重新生成数据处理任务', pageSurface: 'self', backRouteName: 'data-process' },
+      },
+      {
+        path: 'data-process/:id/workflow',
+        name: 'data-process-workflow',
+        component: () => import('@/views/data-process/DataProcessCreateView.vue'),
+        meta: { title: '数据处理任务进度', pageSurface: 'self', backRouteName: 'data-process' },
       },
       // 数据集管理
       {

frontend/src/types/dataProcess.ts

@@ -0,0 +1,398 @@
+/** 数据处理模块的前后端契约。API 字段统一使用 snake_case。 */
+
+export type DataProcessStatus = 'pending' | 'running' | 'completed' | 'failed' | 'stopped'
+export type DataProcessType = 'structured' | 'unstructured' | 'external'
+export type DataProcessResultStatus = 'valid' | 'modified' | 'invalid'
+export type DataProcessSplit = 'train' | 'validation' | 'test'
+export type DataProcessOutputType = 'standard' | 'reasoning'
+export type DataProcessReasoningDetail = 'normal' | 'detailed'
+export type DataProcessWorkflowStep = 'create' | 'model' | 'upload' | 'preview' | 'generate' | 'results'
+export type DataProcessPreviewStatus = 'idle' | 'queued' | 'running' | 'completed' | 'failed' | 'cancelled'
+
+export interface DataProcessPage<T> {
+  items: T[]
+  total: number
+  page: number
+  page_size: number
+}
+
+export interface DataProcessDatasetSplit {
+  train: number
+  validation: number
+  test: number
+}
+
+export type DataProcessConfig = Record<string, unknown> & {
+  dataset_split?: DataProcessDatasetSplit
+  output_type?: DataProcessOutputType
+  reasoning_detail?: DataProcessReasoningDetail
+}
+
+export interface DataProcessTask {
+  id: string | number
+  name: string
+  description?: string
+  status: DataProcessStatus
+  process_type: DataProcessType
+  config?: DataProcessConfig
+  progress?: number
+  source_dataset_id?: string | number | null
+  source_dataset_name?: string | null
+  source_dataset?: string | null
+  output_dataset_id?: string | number | null
+  output_dataset_name?: string | null
+  output_dataset?: string | null
+  output_datasets?: DataProcessOutputDataset[]
+  source_file_count?: number
+  preview_count?: number
+  input_count?: number
+  output_count?: number
+  filtered_count?: number
+  duplicate_count?: number
+  error_count?: number
+  creator_name?: string | null
+  creator?: string | null
+  created_by?: string | number | null
+  create_time?: string
+  created_at?: string
+  start_time?: string | null
+  started_at?: string | null
+  complete_time?: string | null
+  completed_at?: string | null
+  updated_at: string
+  duration?: string | null
+  duration_seconds?: number | null
+  failure_reason?: string | null
+  results_confirmed?: boolean
+  workflow_step?: DataProcessWorkflowStep
+  preview_status?: DataProcessPreviewStatus
+  preview_progress?: number
+  preview_failure_reason?: string | null
+  preview_total_files?: number
+  preview_completed_files?: number
+  preview_run_id?: string | null
+  source_files?: DataProcessSourceFile[]
+}
+
+export interface DataProcessOutputDataset {
+  id: string | number
+  name: string
+  type: 'train' | 'val' | 'test' | string
+  count?: number
+  dataset_split?: DataProcessSplit
+}
+
+export interface DataProcessTaskCreatePayload {
+  name: string
+  description?: string
+  process_type: DataProcessType
+  config: DataProcessConfig
+}
+
+export interface DataProcessRegeneratePayload extends DataProcessTaskCreatePayload {
+  expected_updated_at: string
+}
+
+export interface DataProcessRegenerateResult {
+  task: DataProcessTask
+  preview_invalidated: boolean
+  published_outputs_preserved: boolean
+}
+
+export type DataProcessTaskUpdatePayload = Partial<DataProcessTaskCreatePayload>
+
+export interface DataProcessSourceFile {
+  id: string | number
+  task_id?: string | number
+  name: string
+  size_bytes: number
+  record_count: number
+  file_format?: string
+  checksum_sha256?: string
+  status?: string
+  create_time?: string
+}
+
+export interface DataProcessSourceContent {
+  file_id?: string | number
+  file?: DataProcessSourceFile
+  content: string
+  start_line?: number
+  end_line?: number
+  line_count?: number
+  total_lines?: number
+  has_more?: boolean
+  truncated?: boolean
+  offset?: number
+  limit?: number
+  total_chars?: number
+}
+
+export interface DataProcessPdfPageRange {
+  page_number: number
+  source_start: number
+  source_end: number
+}
+
+export interface DataProcessPdfPages {
+  page_count: number
+  pages: DataProcessPdfPageRange[]
+}
+
+export interface DataProcessDocxParagraph {
+  type: 'paragraph'
+  text: string
+  style: string
+  heading_level: number | null
+  alignment: 'left' | 'center' | 'right' | 'justify' | 'distribute'
+  is_list: boolean
+  source_start: number
+  source_end: number
+}
+
+export interface DataProcessDocxTableRow {
+  cells: string[]
+  source_start: number
+  source_end: number
+}
+
+export interface DataProcessDocxTable {
+  type: 'table'
+  rows: DataProcessDocxTableRow[]
+}
+
+export interface DataProcessDocxPreview {
+  format: 'docx'
+  file_name: string
+  blocks: Array<DataProcessDocxParagraph | DataProcessDocxTable>
+  truncated: boolean
+}
+
+export interface DataProcessXlsxSheet {
+  index: number
+  name: string
+  state: string
+}
+
+export interface DataProcessXlsxPreviewRow {
+  row_number: number
+  record_index: number
+  values: unknown[]
+  record: Record<string, unknown>
+}
+
+export interface DataProcessXlsxActiveSheet {
+  index: number
+  name: string
+  columns: string[]
+  rows: DataProcessXlsxPreviewRow[]
+  offset: number
+  limit: number
+  has_more: boolean
+}
+
+export interface DataProcessXlsxPreview {
+  format: 'xlsx'
+  file_name: string
+  sheets: DataProcessXlsxSheet[]
+  active_sheet: DataProcessXlsxActiveSheet
+}
+
+export type DataProcessOfficePreview = DataProcessDocxPreview | DataProcessXlsxPreview
+
+export interface DataProcessExternalSourcePayload {
+  type: 'postgresql'
+  url: string
+  auth_mode: 'none' | 'basic'
+  username?: string
+  password?: string
+  limit: number
+  query?: string
+  file_name?: string
+}
+
+export interface DataProcessExternalTestResult {
+  connected: boolean
+  latency_ms?: number
+  message?: string
+}
+
+export interface DataProcessPreviewItem {
+  id: string | number
+  source_file_id: string | number
+  original_content: string
+  edited_content: string
+  source_start: number | null
+  source_end: number | null
+  source_start_line: number | null
+  source_end_line: number | null
+  token_count: number
+  status: 'original' | 'modified' | 'manual' | 'invalid'
+  quality_score?: DataProcessQualityScore | null
+  updated_at?: string
+}
+
+export interface DataProcessPreviewBuildPayload {
+  replace_existing?: true
+  source_file_ids?: Array<string | number>
+}
+
+export type DataProcessPreviewFileStatus = 'waiting' | 'processing' | 'success' | 'failed'
+
+export interface DataProcessPreviewBuildFileResult {
+  source_file_id: string | number
+  preview_count: number
+  status: 'completed' | 'empty'
+}
+
+export interface DataProcessPreviewFileBuildProgress {
+  source_file_id: string | number
+  status: Exclude<DataProcessPreviewFileStatus, 'waiting'>
+  progress: number
+  preview_count?: number
+  error?: string
+}
+
+export interface DataProcessPreviewBuildResult extends DataProcessPage<DataProcessPreviewItem> {
+  file_counts?: Record<string, number>
+  files?: DataProcessPreviewBuildFileResult[]
+}
+
+/** 后台切分任务级进度。完成后应重新读取预览条目及源文件统计。 */
+export interface DataProcessPreviewProgress {
+  task_id: string | number
+  workflow_step: DataProcessWorkflowStep
+  preview_status: DataProcessPreviewStatus
+  preview_progress: number
+  preview_failure_reason?: string | null
+  preview_total_files?: number
+  preview_completed_files?: number
+  preview_run_id?: string | null
+}
+
+export interface DataProcessPreviewCreatePayload {
+  source_file_id?: string | number | null
+  original_content?: string
+  edited_content: string
+  source_start?: number | null
+  source_end?: number | null
+  source_start_line?: number | null
+  source_end_line?: number | null
+  token_count?: number
+  status?: DataProcessPreviewItem['status']
+}
+
+export interface DataProcessPreviewUpdatePayload {
+  edited_content: string
+  status?: DataProcessPreviewItem['status']
+  expected_updated_at?: string
+}
+
+export interface DataProcessProgress {
+  task_id: string | number
+  status: DataProcessStatus
+  stage?: string
+  progress: number
+  message?: string
+  processed_count?: number
+  total_count?: number
+  input_count?: number
+  output_count?: number
+  filtered_count?: number
+  duplicate_count?: number
+  error_count?: number
+  failure_reason?: string | null
+  results_confirmed?: boolean
+  updated_at?: string
+}
+
+export interface DataProcessResult {
+  id: string | number
+  preview_item_id?: string | number | null
+  instruction: string
+  input: string
+  output: string
+  original_instruction?: string | null
+  original_input?: string | null
+  original_output?: string | null
+  status: DataProcessResultStatus
+  error?: string | null
+  split?: DataProcessSplit | null
+  quality_score?: DataProcessQualityScore | null
+  updated_at?: string
+}
+
+export interface DataProcessResultUpdatePayload {
+  instruction: string
+  input: string
+  output: string
+  expected_updated_at?: string
+}
+
+export interface DataProcessResultRegeneratePayload {
+  expected_updated_at: string
+}
+
+export interface DataProcessResultBatchRegenerateItem {
+  result_id: string
+  expected_updated_at: string
+}
+
+export interface DataProcessResultBatchRegeneratePayload {
+  items: DataProcessResultBatchRegenerateItem[]
+}
+
+export interface DataProcessResultBatchRegenerateFailure {
+  result_id: string
+  code: 'conflict' | 'skipped' | 'generation_failed' | 'internal_error'
+  message: string
+}
+
+export interface DataProcessResultBatchRegenerateResult {
+  batch_id: string
+  total: number
+  succeeded: number
+  failed: number
+  remaining_invalid_count: number
+  duration_ms: number
+  items: DataProcessResult[]
+  failures: DataProcessResultBatchRegenerateFailure[]
+}
+
+export interface DataProcessQualityScore {
+  overall?: number
+  completeness?: number
+  length?: number
+  readability?: number
+  relevance?: number
+  duplicate?: number
+  is_valid?: boolean
+  flags?: string[]
+  fingerprint?: string
+  [key: string]: unknown
+}
+
+export interface DataProcessPublishPayload {
+  dataset_name: string
+  dataset_type: 'train' | 'test' | 'eval' | 'val' | 'other'
+  storage_type: 'local'
+  split: DataProcessDatasetSplit
+  format: 'alpaca_jsonl' | 'jsonl'
+}
+
+export interface DataProcessPublishResult {
+  dataset_id?: string | number
+  output_dataset_id?: string | number
+  dataset_name?: string
+  record_count?: number
+  dataset?: {
+    id: string | number
+    name?: string
+    record_count?: number
+    count?: number
+    [key: string]: unknown
+  }
+  datasets?: DataProcessOutputDataset[]
+  output_datasets?: DataProcessOutputDataset[]
+  created?: boolean
+}

frontend/src/utils/dataProcessStatus.d.ts

@@ -0,0 +1,13 @@
+import type { DataProcessStatus } from '@/types/dataProcess'
+
+export type DataProcessStatusTagType = 'primary' | 'success' | 'warning' | 'danger' | 'info'
+
+export interface DataProcessStatusLike {
+  status?: DataProcessStatus
+  output_dataset_id?: string | number | null
+}
+
+export function dataProcessDisplayStatus(task: DataProcessStatusLike): {
+  label: string
+  type: DataProcessStatusTagType
+}

frontend/src/utils/dataProcessStatus.js

@@ -0,0 +1,24 @@
+/**
+ * 将数据处理任务的执行状态与当前轮发布指针合并为用户可理解的状态。
+ * 上一轮保留的 output_datasets 不参与判断，避免将尚未重新发布的当前轮误标为已发布。
+ *
+ * @param {{ status?: string, preview_status?: string, output_dataset_id?: string | number | null, results_confirmed?: boolean }} task
+ * @returns {{ label: string, type: 'primary' | 'success' | 'warning' | 'danger' | 'info' }}
+ */
+export function dataProcessDisplayStatus(task) {
+  if (task.preview_status === 'running' || task.preview_status === 'queued') {
+    return { label: '切分中', type: 'primary' }
+  }
+  if (task.preview_status === 'failed') return { label: '切分失败', type: 'danger' }
+  if (task.preview_status === 'cancelled') return { label: '切分已取消', type: 'warning' }
+  if (task.status === 'running') return { label: '生成中', type: 'primary' }
+  if (task.status === 'completed') {
+    if (task.results_confirmed === false) return { label: '待确认', type: 'warning' }
+    return task.output_dataset_id
+      ? { label: '已发布', type: 'success' }
+      : { label: '已生成', type: 'warning' }
+  }
+  if (task.status === 'failed') return { label: '生成失败', type: 'danger' }
+  if (task.status === 'stopped') return { label: '已停止', type: 'info' }
+  return { label: '待生成', type: 'info' }
+}

frontend/src/views/dashboard/DashboardView.vue

@@ -53,6 +53,14 @@ const trainingTasks = ref<DashboardTask[]>([])
 const loginDurationStats = ref<LoginDurationStat[]>([])
 const recentLoginUsers = ref<RecentLoginUser[]>([])
 const training7d = ref<{ date: string; train: number; gpu: number; accuracy: number | null }[]>([])
+
+const onlineServicesHint = computed(() => {
+  if (onlineServices.value === 0) return '暂无在线服务'
+  const abnormal = serviceStatuses.value.filter(
+    (s) => s.state === 'busy' || s.state === 'error'
+  ).length
+  return abnormal > 0 ? `${abnormal} 个异常` : '全部在线'
+})
 const operationDistribution = ref<{ name: string; value: number }[]>([])
 
 const serviceIcon: Record<string, string> = {
@@ -319,8 +327,8 @@ function viewTask(task: DashboardTask) {
         <div class="overview-metrics">
           <div class="overview-metric">
             <span>在线服务</span>
-            <strong>{{ onlineServices }}</strong>
-            <small>全部在线</small>
+          <strong>{{ onlineServices }}</strong>
+          <small>{{ onlineServicesHint }}</small>
           </div>
           <div class="overview-metric">
             <span>运行中任务</span>

frontend/src/views/data-process/DataProcessListView.vue

@@ -1,88 +1,108 @@
 <script setup lang="ts">
-import { ref } from 'vue'
+import { onMounted, ref } from 'vue'
 import { useRouter } from 'vue-router'
-import { ElMessage } from 'element-plus'
+import { ElMessage, ElMessageBox } from 'element-plus'
 import DataTablePage from '@/components/DataTablePage.vue'
-import ModelStatusTag from '@/components/ModelStatusTag.vue'
+import { deleteDataProcessTask, getDataProcessTasks } from '@/api/modules/dataProcess'
+import { usePolling } from '@/composables/usePolling'
+import type { DataProcessTask, DataProcessType } from '@/types/dataProcess'
+import { dataProcessDisplayStatus } from '@/utils/dataProcessStatus.js'
 
-import type { ProcessType } from './create/types'
-
-/** 数据处理任务类型 */
-interface DataProcessTask {
-  id: number | string
-  name: string
-  status: string
-  process_type: ProcessType
-  source_dataset: string
-  output_dataset?: string
-  create_time?: string
-}
-
-const processTypeMap: Record<ProcessType, string> = {
+const processTypeMap: Record<DataProcessType, string> = {
   structured: '结构化数据',
   unstructured: '非结构化数据',
   external: '外来数据源拉取',
 }
 
-// TODO: 接入真实接口前，先用本地 mock 数据
 const router = useRouter()
-const dataList = ref<DataProcessTask[]>([
-  {
-    id: 183921,
-    name: '客服问答数据清洗',
-    status: 'completed',
-    process_type: 'structured',
-    source_dataset: '客服对话原始集',
-    output_dataset: '客服对话清洗集',
-    create_time: '2026-07-08 14:23:00',
-  },
-  {
-    id: 492015,
-    name: '指令微调数据构造',
-    status: 'running',
-    process_type: 'unstructured',
-    source_dataset: '通用语料库',
-    output_dataset: 'SFT 指令集',
-    create_time: '2026-07-09 09:10:00',
-  },
-  {
-    id: 731948,
-    name: '敏感信息脱敏处理',
-    status: 'pending',
-    process_type: 'structured',
-    source_dataset: '用户反馈数据',
-    create_time: '2026-07-09 16:45:00',
-  },
-  {
-    id: 582012,
-    name: '多轮对话拼接',
-    status: 'failed',
-    process_type: 'structured',
-    source_dataset: '单轮问答集',
-    create_time: '2026-07-10 08:30:00',
-  },
-])
+const dataList = ref<DataProcessTask[]>([])
+const loading = ref(false)
+const deletingId = ref<string | number | null>(null)
+const loadError = ref('')
 
-/** 新建数据处理任务 */
-function handleCreate() {
-  router.push('/data-process/create')
+async function loadData(silent = false) {
+  if (!silent) loading.value = true
+  loadError.value = ''
+  try {
+    const response = await getDataProcessTasks({ page: 1, page_size: 200 })
+    dataList.value = response.items
+    if (dataList.value.some((item) => (
+      item.status === 'running'
+      || item.preview_status === 'running'
+      || item.preview_status === 'queued'
+    ))) startPolling()
+    else stopPolling()
+  } catch {
+    loadError.value = '数据处理任务加载失败，请稍后重试。'
+  } finally {
+    if (!silent) loading.value = false
+  }
 }
 
 /** 查看任务详情 */
 function viewDetail(row: unknown) {
-  const taskId = (row as { id: string | number }).id
-  router.push({ name: 'data-process-detail', params: { id: taskId } })
+  const task = row as DataProcessTask
+  const routeName = task.status === 'completed' && task.results_confirmed !== false
+    ? 'data-process-detail'
+    : 'data-process-workflow'
+  router.push({ name: routeName, params: { id: task.id } })
 }
 
-/** 删除任务（功能开发中） */
-function handleDelete(_row: unknown) {
-  ElMessage.info('删除功能开发中...')
+/** 删除由后端再次校验任务状态以及发布锁。 */
+async function handleDelete(row: DataProcessTask) {
+  try {
+    await ElMessageBox.confirm(
+      `确定删除数据处理任务“${row.name}”吗？删除后无法恢复。`,
+      '确认删除',
+      {
+        type: 'warning',
+        confirmButtonText: '删除',
+        cancelButtonText: '取消',
+        confirmButtonClass: 'el-button--danger',
+      },
+    )
+  } catch {
+    return
+  }
+
+  deletingId.value = row.id
+  try {
+    await deleteDataProcessTask(row.id)
+    dataList.value = dataList.value.filter((item) => item.id !== row.id)
+    ElMessage.success('数据处理任务已删除')
+  } catch {
+    // 统一请求层已展示后端返回的失败原因。
+  } finally {
+    deletingId.value = null
+  }
 }
 
 function formatDateTime(value?: string) {
   if (!value) return '-'
-  return new Date(value).toLocaleString('zh-CN', { hour12: false })
+  const date = new Date(value)
+  return Number.isNaN(date.getTime()) ? value : date.toLocaleString('zh-CN', { hour12: false })
 }
+
+function safeCount(value: unknown) {
+  const count = Number(value)
+  return Number.isFinite(count) ? Math.max(0, Math.trunc(count)) : 0
+}
+
+function documentCountLabel(task: DataProcessTask) {
+  return `${safeCount(task.source_file_count)} 个`
+}
+
+function generatedCountLabel(task: DataProcessTask) {
+  return `${safeCount(task.output_count)} 条`
+}
+
+const { start: startPolling, stop: stopPolling } = usePolling(
+  () => loadData(true),
+  3000,
+  { immediate: false },
+)
+
+onMounted(() => void loadData())
 </script>
 
 <template>
@@ -90,37 +110,43 @@ function formatDateTime(value?: string) {
     <DataTablePage
       title=""
       :data="dataList"
+      :loading="loading"
       searchable
       :search-fields="['name']"
       create-text="新建数据处理"
       create-to="/data-process/create"
       row-key="id"
       :page-size="10"
+      :empty-text="loadError || '暂无数据处理任务'"
     >
       <template #columns>
         <el-table-column label="任务ID" prop="id" align="center" width="100" />
         <el-table-column label="任务名称" prop="name" align="center" show-overflow-tooltip />
         <el-table-column label="任务状态" align="center" width="110">
           <template #default="{ row }">
-            <ModelStatusTag :status="row.status" />
+            <el-tag
+              :type="dataProcessDisplayStatus(row as DataProcessTask).type"
+              size="small"
+              effect="light"
+            >{{ dataProcessDisplayStatus(row as DataProcessTask).label }}</el-tag>
           </template>
         </el-table-column>
         <el-table-column label="处理类型" align="center" width="140">
           <template #default="{ row }">
             <el-tag v-if="row.process_type" size="small" type="info" effect="plain">
-              {{ processTypeMap[row.process_type as ProcessType] || row.process_type }}
+              {{ processTypeMap[row.process_type as DataProcessType] || row.process_type }}
             </el-tag>
             <span v-else>-</span>
           </template>
         </el-table-column>
-        <el-table-column label="源数据集" align="center" show-overflow-tooltip>
-          <template #default="{ row }">{{ row.source_dataset || '-' }}</template>
+        <el-table-column label="文档数量" align="center" width="120">
+          <template #default="{ row }">{{ documentCountLabel(row as DataProcessTask) }}</template>
         </el-table-column>
-        <el-table-column label="输出数据集" align="center" show-overflow-tooltip>
-          <template #default="{ row }">{{ row.output_dataset || '-' }}</template>
+        <el-table-column label="生成个数" align="center" width="120">
+          <template #default="{ row }">{{ generatedCountLabel(row as DataProcessTask) }}</template>
         </el-table-column>
         <el-table-column label="创建时间" align="center" width="190">
-          <template #default="{ row }">{{ formatDateTime(row.create_time) }}</template>
+          <template #default="{ row }">{{ formatDateTime(row.create_time || row.created_at) }}</template>
         </el-table-column>
       </template>
 
@@ -129,7 +155,13 @@ function formatDateTime(value?: string) {
           <el-button type="primary" link size="small" @click="viewDetail(row)">
             <i class="fa fa-file-text-o" style="margin-right: 4px" />详情
           </el-button>
-          <el-button type="danger" link size="small" @click="handleDelete(row)">
+          <el-button
+            type="danger"
+            link
+            size="small"
+            :loading="deletingId === row.id"
+            @click="handleDelete(row as DataProcessTask)"
+          >
             <i class="fa fa-trash-o" style="margin-right: 4px" />删除
           </el-button>
         </div>

frontend/src/views/data-process/create/GenerationOptionsPanel.vue

@@ -1,7 +1,11 @@
 <script setup lang="ts">
-import { ref } from 'vue'
+import { ref, onMounted, onBeforeUnmount } from 'vue'
 import type { ModelItem } from '@/types'
 import type { GenerationControlOptions } from './types'
+import {
+  defaultGenerationPrompt,
+  isBuiltInGenerationPrompt,
+} from './dataProcessCreateState'
 
 const props = defineProps<{
   options: GenerationControlOptions
@@ -14,6 +18,43 @@ const emit = defineEmits<{
   'update:options': [value: GenerationControlOptions]
 }>()
 
+const promptInputRef = ref<any>(null)
+const promptHighlightRef = ref<HTMLElement | null>(null)
+
+function splitPrompt(prompt: string) {
+  const parts = (prompt || '').split('{{ content }}')
+  const result: Array<{ text: string, isVar: boolean }> = []
+  parts.forEach((part, index) => {
+    if (part) {
+      result.push({ text: part, isVar: false })
+    }
+    if (index < parts.length - 1) {
+      result.push({ text: '{{ content }}', isVar: true })
+    }
+  })
+  return result
+}
+
+function syncScroll(e: Event) {
+  if (promptHighlightRef.value) {
+    promptHighlightRef.value.scrollTop = (e.target as HTMLElement).scrollTop
+  }
+}
+
+onMounted(() => {
+  const textarea = promptInputRef.value?.textarea
+  if (textarea) {
+    textarea.addEventListener('scroll', syncScroll)
+  }
+})
+
+onBeforeUnmount(() => {
+  const textarea = promptInputRef.value?.textarea
+  if (textarea) {
+    textarea.removeEventListener('scroll', syncScroll)
+  }
+})
+
 function updateField<K extends keyof GenerationControlOptions>(
   field: K,
   value: GenerationControlOptions[K],
@@ -21,6 +62,10 @@ function updateField<K extends keyof GenerationControlOptions>(
   emit('update:options', { ...props.options, [field]: value })
 }
 
+function resetPrompt() {
+  updateField('generationPrompt', defaultGenerationPrompt(props.options.outputType))
+}
+
 function updateQualityRules(value: Array<string | number>) {
   const rules = value.filter((item): item is string => typeof item === 'string')
   emit('update:options', {
@@ -30,6 +75,21 @@ function updateQualityRules(value: Array<string | number>) {
   })
 }
 
+function updateOutputType(value: string | number | boolean | undefined) {
+  const outputType = value === 'reasoning' ? 'reasoning' : 'standard'
+  emit('update:options', {
+    ...props.options,
+    outputType,
+    generationPrompt: isBuiltInGenerationPrompt(props.options.generationPrompt)
+      ? defaultGenerationPrompt(outputType)
+      : props.options.generationPrompt,
+  })
+}
+
+function updateReasoningDetail(value: string | number | boolean | undefined) {
+  updateField('reasoningDetail', value === 'detailed' ? 'detailed' : 'normal')
+}
+
 const selectedQualityRules = () => [
   props.options.filterLowQuality ? 'low_quality' : '',
   props.options.filterShortContent ? 'short_content' : '',
@@ -41,6 +101,15 @@ function sectionValidationMessage() {
   if (props.section === 'model') return isModelMessage ? props.validationMessage : ''
   return isModelMessage ? '' : props.validationMessage
 }
+
+function modelMeta(model: ModelItem) {
+  const source = model.model_source === 'api' || model.model_source === 'online'
+    ? '在线模型'
+    : model.model_source === 'local'
+      ? '本地模型'
+      : model.model_source || '模型管理'
+  return model.type ? `${source} · ${model.type}` : source
+}
 </script>
 
 <template>
@@ -66,34 +135,50 @@ function sectionValidationMessage() {
         >
           <div class="model-option">
             <span>{{ model.name }}</span>
-            <small>{{ model.model_source === 'api' ? '在线模型' : '本地模型' }}</small>
+            <small>{{ modelMeta(model) }}</small>
           </div>
         </el-option>
         <template #empty>
-          <div class="model-empty">暂无可用的大模型，请先在模型管理中添加</div>
+          <div class="model-empty">暂无模型，请先在模型管理中添加</div>
         </template>
       </el-select>
     </div>
 
     <div class="model-field">
       <div class="field-copy">
-        <strong>默认提示语</strong>
-        <small>用于约束生成内容的格式、语气和完整性，可按任务需要修改</small>
+        <div class="field-header" style="display: flex; justify-content: space-between; align-items: center; width: 100%;">
+          <strong>默认提示语</strong>
+          <el-button link type="primary" @click="resetPrompt" style="font-size: 12px; padding: 0;">恢复默认</el-button>
+        </div>
+        <small v-if="options.outputType === 'reasoning'">
+          当前使用思维链专用提示语；系统还会按所选详细程度约束推理结构
+        </small>
+        <small v-else>当前使用标准回答提示语，只要求问题和最终答案</small>
+      </div>
+      <div class="prompt-input-wrapper">
+        <div ref="promptHighlightRef" class="prompt-highlight-layer" aria-hidden="true">
+          <template v-for="(part, index) in splitPrompt(options.generationPrompt)" :key="index">
+            <span v-if="part.isVar" class="highlight-var">{{ part.text }}</span>
+            <span v-else>{{ part.text }}</span>
+          </template>
+        </div>
+        <el-input
+          ref="promptInputRef"
+          class="prompt-input"
+          :model-value="options.generationPrompt"
+          type="textarea"
+          :rows="6"
+          maxlength="2000"
+          show-word-limit
+          resize="vertical"
+          placeholder="请输入模型生成内容时需要遵循的要求"
+          aria-label="模型默认提示语"
+          @update:model-value="updateField('generationPrompt', $event)"
+        />
       </div>
-      <el-input
-        class="prompt-input"
-        :model-value="options.generationPrompt"
-        type="textarea"
-        :rows="4"
-        maxlength="500"
-        show-word-limit
-        resize="vertical"
-        placeholder="请输入模型生成内容时需要遵循的要求"
-        aria-label="模型默认提示语"
-        @update:model-value="updateField('generationPrompt', $event)"
-      />
       <div class="prompt-variables-hint">
-        提示：可在文本中通过 <code v-text="'{{ content }}'" /> 引用当前正在处理的数据内容。
+        默认提示语已包含 <code v-text="'{{ content }}'" />，运行时会替换为当前数据内容；
+        可移动该占位符来调整内容在提示语中的位置。
       </div>
     </div>
 
@@ -148,6 +233,49 @@ function sectionValidationMessage() {
   </div>
 
   <div v-else class="generation-config-group quality-config-group">
+    <div class="output-type-field">
+      <div class="output-type-row">
+        <div class="field-copy">
+          <strong>输出类型</strong>
+          <small>控制答案是否包含可用于推理模型训练的思维链内容</small>
+        </div>
+        <el-select
+          class="output-type-select"
+          :model-value="options.outputType"
+          aria-label="输出类型"
+          @update:model-value="updateOutputType"
+        >
+          <el-option label="标准回答" value="standard" />
+          <el-option label="思维链回答" value="reasoning" />
+        </el-select>
+      </div>
+      <div v-if="options.outputType === 'reasoning'" class="output-type-row">
+        <div class="field-copy">
+          <strong>推理详细程度</strong>
+          <small>控制推理过程保留关键步骤，或完整展开依据与中间推导</small>
+        </div>
+        <el-select
+          class="output-type-select"
+          :model-value="options.reasoningDetail"
+          aria-label="推理详细程度"
+          @update:model-value="updateReasoningDetail"
+        >
+          <el-option label="普通推理（推荐）" value="normal" />
+          <el-option label="详细推理" value="detailed" />
+        </el-select>
+      </div>
+      <p class="output-type-hint">
+        <template v-if="options.outputType === 'reasoning'">
+          <template v-if="options.reasoningDetail === 'detailed'">
+            完整展开条件、来源依据、中间推导和结论核对，
+          </template>
+          <template v-else>保留关键依据与必要步骤，</template>
+          最终按 <code>&lt;think&gt;推理过程&lt;/think&gt;</code> 加最终答案保存。
+        </template>
+        <template v-else>仅保存最终答案，不包含推理过程。</template>
+      </p>
+    </div>
+
     <div class="quality-switch-row">
       <div class="config-group-heading">
         <div>
@@ -279,6 +407,50 @@ function sectionValidationMessage() {
 
 .model-select { width: 100%; }
 .prompt-input { width: 100%; }
+
+.prompt-input-wrapper {
+  position: relative;
+  width: 100%;
+}
+
+.prompt-highlight-layer {
+  position: absolute;
+  top: 1px;
+  left: 1px;
+  right: 1px;
+  bottom: 1px;
+  padding: 5px 11px;
+  box-sizing: border-box;
+  pointer-events: none;
+  font-family: inherit;
+  font-size: 14px;
+  line-height: 1.5;
+  white-space: pre-wrap;
+  word-wrap: break-word;
+  color: #344054;
+  overflow-y: auto;
+  z-index: 1;
+}
+
+.prompt-highlight-layer::-webkit-scrollbar {
+  display: none;
+}
+
+.highlight-var {
+  color: #5b50f2;
+  background: #f0f0ff;
+  border-radius: 4px;
+  padding: 0 4px;
+}
+
+.prompt-input :deep(.el-textarea__inner) {
+  color: transparent !important;
+  caret-color: #344054;
+  background: transparent !important;
+  z-index: 2;
+  position: relative;
+}
+
 .model-option { width: 100%; display: flex; align-items: center; justify-content: space-between; gap: 16px; }
 .model-option small { color: #98a2b3; font-size: 11px; }
 .model-empty { padding: 12px; color: #8a93a3; font-size: 12px; text-align: center; }
@@ -295,6 +467,42 @@ function sectionValidationMessage() {
   }
 }
 
+.output-type-row {
+  display: grid;
+  grid-template-columns: minmax(0, 1fr) 150px;
+  align-items: center;
+  gap: 24px;
+}
+
+.output-type-select {
+  width: 150px;
+}
+
+.output-type-field {
+  display: flex;
+  flex-direction: column;
+  gap: 8px;
+  margin-bottom: 14px;
+  padding-bottom: 14px;
+  border-bottom: 1px solid #e7eaf0;
+}
+
+.output-type-hint {
+  min-height: 20px;
+  margin: 0;
+  color: #667085;
+  font-size: 12px;
+  line-height: 1.6;
+
+  code {
+    padding: 2px 5px;
+    color: #5b50f2;
+    background: #f0f0ff;
+    border-radius: 4px;
+    font-family: inherit;
+  }
+}
+
 .advanced-settings-grid {
   display: grid;
   grid-template-columns: repeat(2, minmax(0, 1fr));
@@ -376,6 +584,8 @@ function sectionValidationMessage() {
 
 @media (max-width: 900px) {
   .advanced-settings-grid { grid-template-columns: 1fr; }
+  .output-type-row { grid-template-columns: 1fr; gap: 10px; }
+  .output-type-select { width: 100%; }
   .quality-rule-list { grid-template-columns: 1fr; }
   .minimum-length-field { grid-template-columns: 1fr; }
 }

frontend/src/views/data-process/create/GenerationStep.vue

@@ -11,7 +11,6 @@ defineProps<{
 }>()
 
 const emit = defineEmits<{
-  stop: []
   retry: []
 }>()
 </script>
@@ -44,7 +43,7 @@ const emit = defineEmits<{
           {{ generation.status === 'idle' ? '准备开始处理'
             : generation.status === 'running' ? '正在生成数据'
               : generation.status === 'success' ? '数据生成完成'
-                : '生成已停止' }}
+                : '生成失败' }}
         </h3>
         <p>{{ generation.message }}</p>
         <el-progress
@@ -58,9 +57,6 @@ const emit = defineEmits<{
           <span>应用预览修改</span>
           <span>生成标准结果</span>
         </div>
-        <el-button v-if="generation.status === 'running'" plain type="warning" @click="emit('stop')">
-          停止生成
-        </el-button>
         <el-button v-if="generation.status === 'failed'" plain type="primary" @click="emit('retry')">
           重新生成
         </el-button>

frontend/src/views/data-process/create/OfficeSourceViewer.vue

@@ -0,0 +1,622 @@
+<script setup lang="ts">
+import { computed, nextTick, ref, shallowRef, watch } from 'vue'
+import {
+  getDataProcessOfficePreview,
+  getDataProcessSourceRawUrl,
+  type DataProcessDocxPreview,
+  type DataProcessDocxTableRow,
+  type DataProcessOfficePreview,
+  type DataProcessXlsxPreview,
+  type DataProcessXlsxPreviewRow,
+} from '@/api/modules/dataProcess'
+import type { PreviewItem } from './types'
+
+const props = defineProps<{
+  taskId: string | number | null
+  sourceFileId: string | number | null
+  fileName: string
+  fileFormat?: string
+  selectedItem: PreviewItem | null
+}>()
+
+const XLSX_PAGE_SIZE = 100
+const scrollRef = ref<HTMLElement | null>(null)
+const preview = shallowRef<DataProcessOfficePreview | null>(null)
+const loading = ref(false)
+const errorMessage = ref('')
+const activeSheetIndex = ref(0)
+const pageOffset = ref(0)
+let loadSequence = 0
+const prefersReducedMotion = window.matchMedia('(prefers-reduced-motion: reduce)').matches
+
+const normalizedFormat = computed(() => (
+  props.fileFormat?.toLowerCase().replace(/^\./, '')
+  || props.fileName.split('.').pop()?.toLowerCase()
+  || ''
+))
+const isDocx = computed(() => normalizedFormat.value === 'docx')
+const docxPreview = computed((): DataProcessDocxPreview | null => (
+  preview.value?.format === 'docx' ? preview.value : null
+))
+const xlsxPreview = computed((): DataProcessXlsxPreview | null => (
+  preview.value?.format === 'xlsx' ? preview.value : null
+))
+const sourceUrl = computed(() => (
+  props.taskId != null && props.sourceFileId != null
+    ? getDataProcessSourceRawUrl(props.taskId, props.sourceFileId)
+    : ''
+))
+const visibleRowRange = computed(() => {
+  const sheet = xlsxPreview.value?.active_sheet
+  if (!sheet || !sheet.rows.length) return '当前工作表没有可预览记录'
+  const start = sheet.offset + 1
+  const end = sheet.offset + sheet.rows.length
+  return `第 ${start}–${end} 条记录`
+})
+
+function overlapsSelection(start: number, end: number) {
+  const item = props.selectedItem
+  if (!item || item.sourceStart == null || item.sourceEnd == null) return false
+  return end > item.sourceStart && start < item.sourceEnd
+}
+
+function tableRowHighlighted(row: DataProcessDocxTableRow) {
+  return overlapsSelection(row.source_start, row.source_end)
+}
+
+function stableValue(value: unknown): unknown {
+  if (Array.isArray(value)) return value.map(stableValue)
+  if (value && typeof value === 'object') {
+    return Object.fromEntries(
+      Object.entries(value as Record<string, unknown>)
+        .sort(([left], [right]) => left.localeCompare(right))
+        .map(([key, item]) => [key, stableValue(item)]),
+    )
+  }
+  return value
+}
+
+function recordKey(value: unknown) {
+  try {
+    return JSON.stringify(stableValue(value))
+  } catch {
+    return ''
+  }
+}
+
+const selectedRecordKey = computed(() => {
+  const content = props.selectedItem?.originalContent
+  if (!content) return ''
+  try {
+    return recordKey(JSON.parse(content))
+  } catch {
+    return ''
+  }
+})
+
+function xlsxRowHighlighted(row: DataProcessXlsxPreviewRow) {
+  return Boolean(selectedRecordKey.value && recordKey(row.record) === selectedRecordKey.value)
+}
+
+function displayCell(value: unknown) {
+  if (value == null || value === '') return '—'
+  if (typeof value === 'object') return JSON.stringify(value)
+  return String(value)
+}
+
+async function locateSelectedItem() {
+  await nextTick()
+  const selected = scrollRef.value?.querySelector<HTMLElement>(
+    '.docx-block.is-highlighted, .docx-table-row.is-highlighted, .xlsx-row.is-highlighted',
+  )
+  selected?.scrollIntoView({
+    block: 'center',
+    inline: 'nearest',
+    behavior: prefersReducedMotion ? 'auto' : 'smooth',
+  })
+}
+
+async function loadPreview(options: { reset?: boolean } = {}) {
+  const sequence = ++loadSequence
+  if (options.reset) {
+    activeSheetIndex.value = 0
+    pageOffset.value = 0
+    preview.value = null
+  }
+  errorMessage.value = ''
+  if (props.taskId == null || props.sourceFileId == null) {
+    errorMessage.value = '缺少原文件标识，无法加载预览'
+    return
+  }
+
+  loading.value = true
+  try {
+    const result = await getDataProcessOfficePreview(
+      props.taskId,
+      props.sourceFileId,
+      isDocx.value
+        ? {}
+        : {
+            sheet_index: activeSheetIndex.value,
+            offset: pageOffset.value,
+            limit: XLSX_PAGE_SIZE,
+          },
+    )
+    if (sequence !== loadSequence) return
+    preview.value = result
+    if (result.format === 'xlsx') activeSheetIndex.value = result.active_sheet.index
+    await locateSelectedItem()
+  } catch (error) {
+    if (sequence !== loadSequence) return
+    errorMessage.value = error instanceof Error ? error.message : 'Office 原文件预览加载失败'
+  } finally {
+    if (sequence === loadSequence) loading.value = false
+  }
+}
+
+function changeSheet(value: string | number) {
+  activeSheetIndex.value = Number(value)
+  pageOffset.value = 0
+  void loadPreview()
+}
+
+function previousPage() {
+  pageOffset.value = Math.max(0, pageOffset.value - XLSX_PAGE_SIZE)
+  void loadPreview()
+}
+
+function nextPage() {
+  if (!xlsxPreview.value?.active_sheet.has_more) return
+  pageOffset.value += XLSX_PAGE_SIZE
+  void loadPreview()
+}
+
+watch(
+  () => [props.taskId, props.sourceFileId, normalizedFormat.value],
+  () => void loadPreview({ reset: true }),
+  { immediate: true },
+)
+
+watch(
+  () => props.selectedItem?.id,
+  () => void locateSelectedItem(),
+)
+</script>
+
+<template>
+  <div
+    class="office-source-viewer"
+    :aria-label="`${normalizedFormat.toUpperCase()} 预览：${fileName}`"
+  >
+    <div class="office-toolbar">
+      <template v-if="xlsxPreview">
+        <div class="sheet-selector">
+          <span>工作表</span>
+          <el-select
+            :model-value="activeSheetIndex"
+            size="small"
+            aria-label="选择 Excel 工作表"
+            @update:model-value="changeSheet"
+          >
+            <el-option
+              v-for="sheet in xlsxPreview.sheets"
+              :key="sheet.index"
+              :label="sheet.name"
+              :value="sheet.index"
+            />
+          </el-select>
+        </div>
+        <span>{{ visibleRowRange }}</span>
+      </template>
+      <template v-else>
+        <span>Word 网页版式预览</span>
+        <span v-if="docxPreview?.truncated">文档较长，仅展示前 2,000 个内容块</span>
+      </template>
+      <span v-if="loading && preview" class="toolbar-loading" role="status">
+        <i class="fa fa-spinner fa-spin" /> 正在更新预览
+      </span>
+      <a
+        v-if="sourceUrl"
+        class="source-file-link"
+        :href="sourceUrl"
+        target="_blank"
+        rel="noopener noreferrer"
+      >
+        <i class="fa fa-external-link" /> 打开原文件
+      </a>
+    </div>
+
+    <div v-if="loading && !preview" class="office-state" role="status">
+      <i class="fa fa-spinner fa-spin" />
+      <strong>正在加载原文件预览</strong>
+      <span>{{ isDocx ? '正在还原 Word 文档结构' : '正在读取 Excel 工作表' }}</span>
+    </div>
+
+    <div v-else-if="errorMessage" class="office-state is-error" role="alert">
+      <i class="fa fa-exclamation-circle" />
+      <strong>{{ isDocx ? 'Word 预览失败' : 'Excel 预览失败' }}</strong>
+      <span>{{ errorMessage }}</span>
+      <div>
+        <el-button type="primary" size="small" @click="loadPreview()">重试</el-button>
+        <el-button v-if="sourceUrl" tag="a" :href="sourceUrl" target="_blank" size="small">
+          打开原文件
+        </el-button>
+      </div>
+    </div>
+
+    <div v-else-if="docxPreview" ref="scrollRef" class="docx-scroll">
+      <article class="docx-page">
+        <template v-for="(block, index) in docxPreview.blocks" :key="index">
+          <component
+            :is="block.heading_level ? `h${block.heading_level}` : 'p'"
+            v-if="block.type === 'paragraph'"
+            class="docx-block"
+            :class="{
+              'is-highlighted': overlapsSelection(block.source_start, block.source_end),
+              'is-list': block.is_list,
+            }"
+            :style="{ textAlign: block.alignment }"
+            :data-source-start="block.source_start"
+          >
+            {{ block.text }}
+          </component>
+          <div v-else class="docx-table-wrap">
+            <table class="docx-table">
+              <tbody>
+                <tr
+                  v-for="(row, rowIndex) in block.rows"
+                  :key="rowIndex"
+                  class="docx-table-row"
+                  :class="{ 'is-highlighted': tableRowHighlighted(row) }"
+                  :data-source-start="row.source_start"
+                >
+                  <td v-for="(cell, cellIndex) in row.cells" :key="cellIndex">{{ cell || ' ' }}</td>
+                </tr>
+              </tbody>
+            </table>
+          </div>
+        </template>
+        <div v-if="!docxPreview.blocks.length" class="office-empty">文档中没有可预览的正文</div>
+      </article>
+    </div>
+
+    <template v-else-if="xlsxPreview">
+      <div ref="scrollRef" class="xlsx-scroll">
+        <table v-if="xlsxPreview.active_sheet.columns.length" class="xlsx-grid">
+          <thead>
+            <tr>
+              <th class="row-number-cell">#</th>
+              <th
+                v-for="column in xlsxPreview.active_sheet.columns"
+                :key="column"
+                :title="column"
+              >
+                {{ column }}
+              </th>
+            </tr>
+          </thead>
+          <tbody>
+            <tr
+              v-for="row in xlsxPreview.active_sheet.rows"
+              :key="row.row_number"
+              class="xlsx-row"
+              :class="{ 'is-highlighted': xlsxRowHighlighted(row) }"
+            >
+              <th class="row-number-cell">{{ row.row_number }}</th>
+              <td
+                v-for="(value, cellIndex) in row.values"
+                :key="cellIndex"
+                :title="displayCell(value)"
+              >
+                {{ displayCell(value) }}
+              </td>
+            </tr>
+          </tbody>
+        </table>
+        <div v-else class="office-empty">当前工作表没有可预览记录</div>
+      </div>
+      <div class="xlsx-pagination">
+        <el-button
+          size="small"
+          :disabled="pageOffset === 0 || loading"
+          aria-label="上一页工作表记录"
+          @click="previousPage"
+        >
+          <i class="fa fa-angle-left" /> 上一页
+        </el-button>
+        <span>{{ visibleRowRange }}</span>
+        <el-button
+          size="small"
+          :disabled="!xlsxPreview.active_sheet.has_more || loading"
+          aria-label="下一页工作表记录"
+          @click="nextPage"
+        >
+          下一页 <i class="fa fa-angle-right" />
+        </el-button>
+      </div>
+    </template>
+  </div>
+</template>
+
+<style scoped lang="scss">
+.office-source-viewer {
+  display: flex;
+  min-width: 0;
+  min-height: 0;
+  flex: 1;
+  flex-direction: column;
+  background: #f4f6f9;
+}
+
+.office-toolbar {
+  display: flex;
+  min-height: 42px;
+  flex: none;
+  align-items: center;
+  gap: 12px;
+  padding: 6px 12px;
+  color: #7d8798;
+  background: #fff;
+  border-bottom: 1px solid #e5e8ee;
+  font-size: 11px;
+
+}
+
+.sheet-selector {
+  display: flex;
+  min-width: 0;
+  align-items: center;
+  gap: 8px;
+
+  > span {
+    flex: none;
+  }
+
+  :deep(.el-select) {
+    width: min(220px, 28vw);
+  }
+}
+
+.source-file-link {
+  flex: none;
+  margin-left: auto;
+  color: #5147df;
+  text-decoration: none;
+
+  &:hover,
+  &:focus-visible {
+    text-decoration: underline;
+  }
+}
+
+.toolbar-loading {
+  color: #5b50f2;
+  white-space: nowrap;
+}
+
+.office-state {
+  display: flex;
+  flex: 1;
+  align-items: center;
+  justify-content: center;
+  flex-direction: column;
+  gap: 9px;
+  padding: 28px;
+  color: #667085;
+  text-align: center;
+
+  > i {
+    color: #5b50f2;
+    font-size: 28px;
+  }
+
+  > strong {
+    color: #344054;
+    font-size: 14px;
+  }
+
+  > span {
+    max-width: 420px;
+    font-size: 12px;
+    line-height: 1.6;
+  }
+
+  &.is-error > i {
+    color: #d92d20;
+  }
+}
+
+.docx-scroll {
+  flex: 1;
+  min-height: 0;
+  padding: 22px;
+  overflow: auto;
+  scroll-behavior: smooth;
+}
+
+.docx-page {
+  width: min(760px, 100%);
+  min-height: calc(100% - 2px);
+  padding: 54px clamp(30px, 7%, 68px);
+  margin: 0 auto;
+  color: #262b34;
+  background: #fff;
+  border: 1px solid #dfe3e9;
+  box-shadow: 0 2px 10px rgb(15 23 42 / 8%);
+  font-family: "Songti SC", SimSun, serif;
+  font-size: 13px;
+  line-height: 1.8;
+}
+
+.docx-block {
+  padding: 2px 6px;
+  margin: 0 0 10px;
+  border-radius: 3px;
+  white-space: pre-wrap;
+  word-break: break-word;
+  transition: background-color 0.18s ease, box-shadow 0.18s ease;
+
+  &.is-list {
+    padding-left: 22px;
+  }
+
+  &.is-highlighted {
+    background: #fff0b8;
+    box-shadow: inset 3px 0 #f0b429;
+  }
+}
+
+h1.docx-block { font-size: 22px; }
+h2.docx-block { font-size: 19px; }
+h3.docx-block { font-size: 17px; }
+h4.docx-block,
+h5.docx-block,
+h6.docx-block { font-size: 15px; }
+
+.docx-table-wrap {
+  max-width: 100%;
+  margin: 12px 0 18px;
+  overflow-x: auto;
+}
+
+.docx-table {
+  width: 100%;
+  border-collapse: collapse;
+  table-layout: fixed;
+
+  td {
+    padding: 7px 9px;
+    border: 1px solid #9da5b2;
+    vertical-align: top;
+    white-space: pre-wrap;
+    word-break: break-word;
+  }
+}
+
+.docx-table-row.is-highlighted td {
+  background: #fff0b8;
+}
+
+.xlsx-scroll {
+  flex: 1;
+  min-width: 0;
+  min-height: 0;
+  overflow: auto;
+  background: #fff;
+  scroll-behavior: smooth;
+}
+
+.xlsx-grid {
+  min-width: 100%;
+  color: #344054;
+  border-spacing: 0;
+  border-collapse: separate;
+  table-layout: auto;
+  font-size: 11px;
+
+  th,
+  td {
+    min-width: 120px;
+    max-width: 320px;
+    height: 36px;
+    padding: 7px 10px;
+    overflow: hidden;
+    text-align: left;
+    text-overflow: ellipsis;
+    white-space: nowrap;
+    border-right: 1px solid #e4e7ec;
+    border-bottom: 1px solid #e4e7ec;
+  }
+
+  thead th {
+    position: sticky;
+    z-index: 2;
+    top: 0;
+    color: #475467;
+    background: #f2f4f7;
+    font-weight: 650;
+  }
+
+  tbody tr:hover td,
+  tbody tr:hover th {
+    background: #f9fafb;
+  }
+}
+
+.row-number-cell {
+  position: sticky;
+  z-index: 1;
+  left: 0;
+  min-width: 54px !important;
+  width: 54px;
+  color: #98a2b3;
+  text-align: center !important;
+  background: #f8fafc;
+  font-variant-numeric: tabular-nums;
+}
+
+thead .row-number-cell {
+  z-index: 3;
+}
+
+.xlsx-row.is-highlighted {
+  td,
+  th {
+    background: #fff0b8;
+    box-shadow: inset 0 2px #f0b429, inset 0 -2px #f0b429;
+  }
+}
+
+.xlsx-pagination {
+  display: flex;
+  min-height: 46px;
+  flex: none;
+  align-items: center;
+  justify-content: flex-end;
+  gap: 12px;
+  padding: 7px 12px;
+  color: #7d8798;
+  background: #fff;
+  border-top: 1px solid #e5e8ee;
+  font-size: 11px;
+}
+
+.office-empty {
+  display: flex;
+  min-height: 180px;
+  align-items: center;
+  justify-content: center;
+  color: #98a2b3;
+  font-size: 12px;
+}
+
+@media (max-width: 900px) {
+  .office-source-viewer {
+    height: 360px;
+    flex: none;
+  }
+
+  .office-toolbar {
+    flex-wrap: wrap;
+  }
+
+  .docx-scroll {
+    padding: 12px;
+  }
+
+  .docx-page {
+    padding: 34px 24px;
+  }
+}
+
+@media (prefers-reduced-motion: reduce) {
+  .docx-scroll,
+  .xlsx-scroll {
+    scroll-behavior: auto;
+  }
+
+  .docx-block {
+    transition: none;
+  }
+}
+</style>

frontend/src/views/data-process/create/PdfSourceViewer.vue

@@ -0,0 +1,667 @@
+<script setup lang="ts">
+import {
+  GlobalWorkerOptions,
+  TextLayer,
+  getDocument,
+  type PDFDocumentLoadingTask,
+  type PDFDocumentProxy,
+  type RenderTask,
+} from 'pdfjs-dist'
+import pdfWorkerUrl from 'pdfjs-dist/build/pdf.worker.min.mjs?url'
+import { computed, nextTick, onBeforeUnmount, ref, shallowRef, watch } from 'vue'
+import {
+  getDataProcessPdfPages,
+  getDataProcessSourceRawUrl,
+  type DataProcessPdfPageRange,
+} from '@/api/modules/dataProcess'
+import type { PreviewItem } from './types'
+
+GlobalWorkerOptions.workerSrc = pdfWorkerUrl
+
+const props = defineProps<{
+  taskId: string | number | null
+  sourceFileId: string | number | null
+  fileName: string
+  selectedItem: PreviewItem | null
+}>()
+
+const scrollRef = ref<HTMLElement | null>(null)
+const pageRef = ref<HTMLElement | null>(null)
+const canvasRef = ref<HTMLCanvasElement | null>(null)
+const textLayerRef = ref<HTMLElement | null>(null)
+const documentRef = shallowRef<PDFDocumentProxy | null>(null)
+const pageRanges = ref<DataProcessPdfPageRange[]>([])
+const currentPage = ref(1)
+const pageCount = ref(0)
+const zoom = ref(1)
+const renderedScale = ref(1)
+const loading = ref(false)
+const rendering = ref(false)
+const errorMessage = ref('')
+const highlightState = ref<'idle' | 'highlighted' | 'unmatched' | 'manual'>('idle')
+
+let loadingTask: PDFDocumentLoadingTask | null = null
+let renderTask: RenderTask | null = null
+let textLayer: TextLayer | null = null
+let loadSequence = 0
+let renderSequence = 0
+let renderedPageNumber = 0
+let resizeFrame = 0
+
+const sourceUrl = computed(() => (
+  props.taskId != null && props.sourceFileId != null
+    ? getDataProcessSourceRawUrl(props.taskId, props.sourceFileId)
+    : ''
+))
+
+const pageStyle = computed(() => ({
+  '--total-scale-factor': String(renderedScale.value),
+}))
+
+const locationText = computed(() => {
+  if (highlightState.value === 'highlighted') return `切片原文已在第 ${currentPage.value} 页高亮`
+  if (highlightState.value === 'unmatched') return `已定位第 ${currentPage.value} 页，未匹配到可高亮文字`
+  if (highlightState.value === 'manual') return '手动新增切片没有原文位置'
+  return pageCount.value ? `第 ${currentPage.value} / ${pageCount.value} 页` : '正在读取 PDF'
+})
+
+function pageForItem(item: PreviewItem | null) {
+  if (!item) return null
+  if (item.sourceStart == null || item.sourceEnd == null) {
+    const pageNumber = item.sourcePages?.[0]
+    return pageNumber == null
+      ? null
+      : pageRanges.value.find((page) => page.page_number === pageNumber) ?? null
+  }
+  return pageRanges.value.find((page) => (
+    item.sourceStart! >= page.source_start && item.sourceStart! < page.source_end
+  )) ?? pageRanges.value.find((page) => (
+    item.sourceStart! < page.source_end && item.sourceEnd! > page.source_start
+  )) ?? null
+}
+
+function selectedTextForPage(page: DataProcessPdfPageRange) {
+  const item = props.selectedItem
+  if (!item) return ''
+  if (item.sourceStart == null || item.sourceEnd == null) {
+    return item.sourcePages?.includes(page.page_number) ? item.originalContent : ''
+  }
+  const intersectionStart = Math.max(item.sourceStart, page.source_start)
+  const intersectionEnd = Math.min(item.sourceEnd, page.source_end)
+  if (intersectionEnd <= intersectionStart) return ''
+  const relativeStart = Math.max(0, intersectionStart - item.sourceStart)
+  const relativeEnd = Math.max(relativeStart, intersectionEnd - item.sourceStart)
+  return item.originalContent.slice(relativeStart, relativeEnd)
+}
+
+function normalizeLocatorText(value: string) {
+  return value.normalize('NFKC').replace(/[^\p{L}\p{N}]+/gu, '').toLocaleLowerCase()
+}
+
+function nearestOccurrence(haystack: string, needle: string, expectedIndex: number) {
+  let nearest = -1
+  let nearestDistance = Number.POSITIVE_INFINITY
+  let cursor = haystack.indexOf(needle)
+  while (cursor >= 0) {
+    const distance = Math.abs(cursor - expectedIndex)
+    if (distance < nearestDistance) {
+      nearest = cursor
+      nearestDistance = distance
+    }
+    cursor = haystack.indexOf(needle, cursor + 1)
+  }
+  return nearest
+}
+
+function findHighlight(
+  itemStrings: string[],
+  selectedText: string,
+  page: DataProcessPdfPageRange,
+) {
+  const itemRanges: Array<{ start: number; end: number }> = []
+  let pageText = ''
+  for (const item of itemStrings) {
+    const start = pageText.length
+    pageText += normalizeLocatorText(item)
+    itemRanges.push({ start, end: pageText.length })
+  }
+
+  const target = normalizeLocatorText(selectedText)
+  if (!pageText || !target) return []
+  const itemStart = props.selectedItem?.sourceStart ?? page.source_start
+  const sourceLength = Math.max(1, page.source_end - page.source_start)
+  const expectedRatio = Math.min(1, Math.max(0, (itemStart - page.source_start) / sourceLength))
+  const expectedIndex = Math.round(pageText.length * expectedRatio)
+  const anchorLengths = [target.length, 120, 80, 48, 24, 12, 8]
+    .map((length) => Math.min(length, target.length))
+    .filter((length, index, values) => length >= 4 && values.indexOf(length) === index)
+
+  for (const anchorLength of anchorLengths) {
+    const anchor = target.slice(0, anchorLength)
+    const matchStart = nearestOccurrence(pageText, anchor, expectedIndex)
+    if (matchStart < 0) continue
+    const matchEnd = matchStart + anchor.length
+    return itemRanges.reduce<number[]>((matches, range, index) => {
+      if (range.end > matchStart && range.start < matchEnd) matches.push(index)
+      return matches
+    }, [])
+  }
+  return []
+}
+
+function applySelectedHighlight(page: DataProcessPdfPageRange | null) {
+  const item = props.selectedItem
+  for (const element of textLayer?.textDivs ?? []) {
+    element.classList.remove('is-slice-highlighted')
+  }
+  if (!item || (
+    (item.sourceStart == null || item.sourceEnd == null)
+    && !item.sourcePages?.length
+  )) {
+    highlightState.value = item ? 'manual' : 'idle'
+    return
+  }
+  if (!page || page.page_number !== currentPage.value || !textLayer) {
+    highlightState.value = 'idle'
+    return
+  }
+  const matches = findHighlight(
+    textLayer.textContentItemsStr,
+    selectedTextForPage(page),
+    page,
+  )
+  for (const index of matches) {
+    textLayer.textDivs[index]?.classList.add('is-slice-highlighted')
+  }
+  highlightState.value = matches.length ? 'highlighted' : 'unmatched'
+}
+
+async function renderCurrentPage(force = false) {
+  const document = documentRef.value
+  const pageContainer = pageRef.value
+  const canvas = canvasRef.value
+  const layerContainer = textLayerRef.value
+  const scroller = scrollRef.value
+  if (!document || !pageContainer || !canvas || !layerContainer || !scroller) return
+
+  const selectedPage = pageForItem(props.selectedItem)
+  if (!force && renderedPageNumber === currentPage.value && textLayer) {
+    applySelectedHighlight(selectedPage)
+    return
+  }
+
+  const sequence = ++renderSequence
+  renderTask?.cancel()
+  textLayer?.cancel()
+  renderTask = null
+  textLayer = null
+  rendering.value = true
+  errorMessage.value = ''
+
+  try {
+    const page = await document.getPage(currentPage.value)
+    if (sequence !== renderSequence) return
+    const baseViewport = page.getViewport({ scale: 1 })
+    const availableWidth = Math.max(280, scroller.clientWidth - 36)
+    const scale = (availableWidth / baseViewport.width) * zoom.value
+    const viewport = page.getViewport({ scale })
+    const outputScale = Math.max(1, window.devicePixelRatio || 1)
+    const context = canvas.getContext('2d')
+    if (!context) throw new Error('浏览器无法创建 PDF 画布')
+
+    renderedScale.value = scale
+    pageContainer.style.width = `${viewport.width}px`
+    pageContainer.style.height = `${viewport.height}px`
+    canvas.width = Math.floor(viewport.width * outputScale)
+    canvas.height = Math.floor(viewport.height * outputScale)
+    canvas.style.width = `${viewport.width}px`
+    canvas.style.height = `${viewport.height}px`
+    layerContainer.replaceChildren()
+
+    const activeRenderTask = page.render({
+      canvas,
+      viewport,
+      transform: outputScale === 1 ? undefined : [outputScale, 0, 0, outputScale, 0, 0],
+    })
+    renderTask = activeRenderTask
+    // 立即挂接取消处理，避免 ResizeObserver 触发重绘时产生未处理的取消异常。
+    const canvasRenderPromise = activeRenderTask.promise.catch((error: unknown) => {
+      if (
+        sequence !== renderSequence
+        || (error instanceof Error && error.name === 'RenderingCancelledException')
+      ) return
+      throw error
+    })
+    // WebKit 兼容：PDF.js 的 getTextContent() 依赖 ReadableStream 异步迭代，
+    // 部分 Safari/WKWebView 未实现该接口。TextLayer 可直接通过 getReader() 消费文本流。
+    const textContentSource = page.streamTextContent()
+    if (sequence !== renderSequence) return
+    const activeTextLayer = new TextLayer({
+      textContentSource,
+      container: layerContainer,
+      viewport,
+    })
+    textLayer = activeTextLayer
+    await Promise.all([canvasRenderPromise, activeTextLayer.render()])
+    if (sequence !== renderSequence) return
+    renderedPageNumber = currentPage.value
+    applySelectedHighlight(selectedPage)
+    await nextTick()
+    pageContainer.querySelector<HTMLElement>('.is-slice-highlighted')
+      ?.scrollIntoView({ block: 'center', inline: 'center', behavior: 'smooth' })
+  } catch (error) {
+    if (sequence !== renderSequence) return
+    errorMessage.value = error instanceof Error ? error.message : 'PDF 页面渲染失败'
+  } finally {
+    if (sequence === renderSequence) rendering.value = false
+  }
+}
+
+async function locateSelectedItem() {
+  if (!documentRef.value) return
+  const item = props.selectedItem
+  if (!item || (
+    (item.sourceStart == null || item.sourceEnd == null)
+    && !item.sourcePages?.length
+  )) {
+    applySelectedHighlight(null)
+    return
+  }
+  const page = pageForItem(item)
+  if (!page) {
+    highlightState.value = 'unmatched'
+    return
+  }
+  if (currentPage.value !== page.page_number) {
+    currentPage.value = page.page_number
+    await renderCurrentPage(true)
+    return
+  }
+  await renderCurrentPage(false)
+}
+
+async function loadPdf() {
+  const sequence = ++loadSequence
+  ++renderSequence
+  renderedPageNumber = 0
+  renderTask?.cancel()
+  textLayer?.cancel()
+  await loadingTask?.destroy()
+  loadingTask = null
+  documentRef.value = null
+  pageRanges.value = []
+  pageCount.value = 0
+  currentPage.value = 1
+  errorMessage.value = ''
+  highlightState.value = 'idle'
+  if (!sourceUrl.value || props.taskId == null || props.sourceFileId == null) return
+
+  loading.value = true
+  try {
+    const task = getDocument({ url: sourceUrl.value })
+    loadingTask = task
+    const [mapping, document] = await Promise.all([
+      getDataProcessPdfPages(props.taskId, props.sourceFileId),
+      task.promise,
+    ])
+    if (sequence !== loadSequence) {
+      await task.destroy()
+      return
+    }
+    documentRef.value = document
+    pageRanges.value = mapping.pages
+    pageCount.value = document.numPages
+    await nextTick()
+    await locateSelectedItem()
+    if (!props.selectedItem) await renderCurrentPage(true)
+  } catch (error) {
+    if (sequence !== loadSequence) return
+    errorMessage.value = error instanceof Error ? error.message : 'PDF 原文件加载失败'
+  } finally {
+    if (sequence === loadSequence) loading.value = false
+  }
+}
+
+async function changePage(delta: number) {
+  const nextPage = Math.min(pageCount.value, Math.max(1, currentPage.value + delta))
+  if (nextPage === currentPage.value) return
+  currentPage.value = nextPage
+  await renderCurrentPage(true)
+}
+
+async function changeZoom(delta: number) {
+  const nextZoom = Math.min(2, Math.max(0.6, Number((zoom.value + delta).toFixed(1))))
+  if (nextZoom === zoom.value) return
+  zoom.value = nextZoom
+  await renderCurrentPage(true)
+}
+
+function scheduleResizeRender() {
+  window.cancelAnimationFrame(resizeFrame)
+  resizeFrame = window.requestAnimationFrame(() => {
+    if (documentRef.value) void renderCurrentPage(true)
+  })
+}
+
+const resizeObserver = new ResizeObserver(scheduleResizeRender)
+watch(scrollRef, (element, previous) => {
+  if (previous) resizeObserver.unobserve(previous)
+  if (element) resizeObserver.observe(element)
+})
+watch(sourceUrl, () => void loadPdf(), { immediate: true })
+watch(() => props.selectedItem?.id, () => void locateSelectedItem())
+
+onBeforeUnmount(() => {
+  ++loadSequence
+  ++renderSequence
+  window.cancelAnimationFrame(resizeFrame)
+  resizeObserver.disconnect()
+  renderTask?.cancel()
+  textLayer?.cancel()
+  void loadingTask?.destroy()
+})
+</script>
+
+<template>
+  <div class="pdf-source-viewer" aria-label="PDF 原文件预览">
+    <div v-if="sourceUrl" class="pdf-toolbar">
+      <div class="toolbar-group">
+        <button type="button" aria-label="上一页" :disabled="currentPage <= 1 || loading" @click="changePage(-1)">
+          <i class="fa fa-chevron-left" aria-hidden="true" />
+        </button>
+        <span class="page-indicator">{{ currentPage }} / {{ pageCount || '—' }}</span>
+        <button type="button" aria-label="下一页" :disabled="currentPage >= pageCount || loading" @click="changePage(1)">
+          <i class="fa fa-chevron-right" aria-hidden="true" />
+        </button>
+      </div>
+      <span class="location-state" :class="`is-${highlightState}`" role="status">
+        <i v-if="highlightState === 'highlighted'" class="fa fa-map-marker" aria-hidden="true" />
+        {{ locationText }}
+      </span>
+      <div class="toolbar-group">
+        <button type="button" aria-label="缩小" :disabled="zoom <= 0.6 || loading" @click="changeZoom(-0.1)">−</button>
+        <span class="zoom-indicator">{{ Math.round(zoom * 100) }}%</span>
+        <button type="button" aria-label="放大" :disabled="zoom >= 2 || loading" @click="changeZoom(0.1)">+</button>
+      </div>
+    </div>
+
+    <div v-if="sourceUrl" ref="scrollRef" class="pdf-scroll" tabindex="0" :aria-label="`PDF 预览：${fileName}`">
+      <div
+        ref="pageRef"
+        class="pdf-page"
+        :style="pageStyle"
+        :data-page-number="currentPage"
+      >
+        <canvas ref="canvasRef" class="pdf-canvas" />
+        <div ref="textLayerRef" class="pdf-text-layer" />
+      </div>
+      <div v-if="loading || rendering" class="pdf-loading" role="status">
+        <i class="fa fa-spinner fa-spin" aria-hidden="true" />
+        {{ loading ? '正在加载 PDF…' : '正在渲染页面…' }}
+      </div>
+      <div v-if="errorMessage" class="pdf-error" role="alert">
+        <i class="fa fa-exclamation-circle" aria-hidden="true" />
+        <strong>PDF 预览失败</strong>
+        <span>{{ errorMessage }}</span>
+        <a :href="sourceUrl" target="_blank" rel="noopener noreferrer">在新窗口打开原件</a>
+      </div>
+    </div>
+    <div v-else class="pdf-unavailable" role="status">
+      <i class="fa fa-file-pdf-o" aria-hidden="true" />
+      <strong>PDF 原文件暂不可预览</strong>
+      <span>请重新上传该文件后再试。</span>
+    </div>
+  </div>
+</template>
+
+<style scoped lang="scss">
+.pdf-source-viewer {
+  display: flex;
+  min-height: 0;
+  flex: 1;
+  flex-direction: column;
+  overflow: hidden;
+  background: #525659;
+}
+
+.pdf-toolbar {
+  display: grid;
+  min-height: 44px;
+  flex: 0 0 auto;
+  grid-template-columns: auto minmax(0, 1fr) auto;
+  align-items: center;
+  gap: 12px;
+  padding: 0 12px;
+  color: #f2f4f7;
+  background: #323639;
+  border-bottom: 1px solid #1f2427;
+}
+
+.toolbar-group {
+  display: flex;
+  align-items: center;
+  gap: 5px;
+
+  button {
+    display: inline-flex;
+    width: 30px;
+    height: 30px;
+    align-items: center;
+    justify-content: center;
+    color: #f2f4f7;
+    font: inherit;
+    font-size: 16px;
+    background: transparent;
+    border: 0;
+    border-radius: 4px;
+    cursor: pointer;
+
+    &:hover:not(:disabled) {
+      background: rgb(255 255 255 / 12%);
+    }
+
+    &:disabled {
+      color: #7d8387;
+      cursor: not-allowed;
+    }
+  }
+}
+
+.page-indicator,
+.zoom-indicator {
+  min-width: 54px;
+  color: #e4e7ec;
+  font-size: 12px;
+  text-align: center;
+}
+
+.location-state {
+  min-width: 0;
+  overflow: hidden;
+  color: #d0d5dd;
+  font-size: 12px;
+  text-align: center;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+
+  &.is-highlighted {
+    color: #ffd666;
+  }
+
+  &.is-unmatched {
+    color: #fdb022;
+  }
+}
+
+.pdf-scroll {
+  position: relative;
+  display: flex;
+  min-height: 538px;
+  flex: 1;
+  align-items: flex-start;
+  justify-content: center;
+  overflow: auto;
+  padding: 18px;
+  outline: none;
+}
+
+.pdf-page {
+  position: relative;
+  flex: 0 0 auto;
+  overflow: hidden;
+  background: #fff;
+  box-shadow: 0 2px 12px rgb(0 0 0 / 34%);
+}
+
+.pdf-canvas {
+  position: absolute;
+  inset: 0;
+  display: block;
+}
+
+.pdf-text-layer {
+  --min-font-size: 1;
+  --text-scale-factor: calc(var(--total-scale-factor) * var(--min-font-size));
+  --min-font-size-inv: calc(1 / var(--min-font-size));
+
+  position: absolute;
+  z-index: 1;
+  inset: 0;
+  overflow: clip;
+  color-scheme: only light;
+  line-height: 1;
+  letter-spacing: normal;
+  word-spacing: normal;
+  text-align: initial;
+  text-size-adjust: none;
+  forced-color-adjust: none;
+  transform-origin: 0 0;
+  caret-color: CanvasText;
+}
+
+.pdf-text-layer :deep(span),
+.pdf-text-layer :deep(br) {
+  position: absolute;
+  color: transparent;
+  white-space: pre;
+  cursor: text;
+  user-select: text;
+  transform-origin: 0 0;
+}
+
+.pdf-text-layer > :deep(:not(.markedContent)),
+.pdf-text-layer :deep(.markedContent span:not(.markedContent)) {
+  --font-height: 0;
+  --scale-x: 1;
+  --rotate: 0deg;
+
+  z-index: 1;
+  font-size: calc(var(--text-scale-factor) * var(--font-height));
+  transform: rotate(var(--rotate)) scaleX(var(--scale-x)) scale(var(--min-font-size-inv));
+}
+
+.pdf-text-layer :deep(.markedContent) {
+  display: contents;
+}
+
+.pdf-text-layer :deep(.is-slice-highlighted) {
+  margin: -2px;
+  padding: 2px;
+  background: rgb(255 202 40 / 48%);
+  border-radius: 3px;
+  box-shadow: 0 0 0 1px rgb(245 158 11 / 38%);
+}
+
+.pdf-text-layer :deep(::selection) {
+  color: transparent;
+  background: rgb(37 99 235 / 30%);
+}
+
+.pdf-loading,
+.pdf-error {
+  position: absolute;
+  z-index: 3;
+  top: 50%;
+  left: 50%;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  transform: translate(-50%, -50%);
+}
+
+.pdf-loading {
+  gap: 8px;
+  padding: 10px 14px;
+  color: #f2f4f7;
+  font-size: 13px;
+  background: rgb(31 36 39 / 86%);
+  border-radius: 6px;
+}
+
+.pdf-error {
+  width: min(360px, calc(100% - 32px));
+  flex-direction: column;
+  gap: 8px;
+  padding: 20px;
+  color: #667085;
+  text-align: center;
+  background: #fff;
+  border-radius: 8px;
+  box-shadow: 0 4px 20px rgb(0 0 0 / 22%);
+
+  i {
+    color: #d92d20;
+    font-size: 28px;
+  }
+
+  strong {
+    color: #344054;
+  }
+
+  span,
+  a {
+    font-size: 12px;
+  }
+}
+
+.pdf-unavailable {
+  display: flex;
+  width: 100%;
+  flex: 1;
+  flex-direction: column;
+  align-items: center;
+  justify-content: center;
+  gap: 8px;
+  color: #667085;
+  background: #f8f9fb;
+
+  i {
+    color: #d92d20;
+    font-size: 36px;
+  }
+
+  strong {
+    color: #344054;
+    font-size: 14px;
+  }
+
+  span {
+    font-size: 12px;
+  }
+}
+
+@media (max-width: 900px) {
+  .pdf-toolbar {
+    gap: 5px;
+    padding: 0 6px;
+  }
+
+  .location-state {
+    font-size: 11px;
+  }
+
+  .pdf-scroll {
+    min-height: 420px;
+  }
+}
+</style>

frontend/src/views/data-process/create/PreviewCompareStep.vue

@@ -1,5 +1,7 @@
 <script setup lang="ts">
 import { computed, nextTick, ref, watch } from 'vue'
+import OfficeSourceViewer from './OfficeSourceViewer.vue'
+import PdfSourceViewer from './PdfSourceViewer.vue'
 import { sourceLines } from './previewModel'
 import type { PreviewItem, ProcessType } from './types'
 
@@ -9,6 +11,9 @@ const props = defineProps<{
   selectedId: string | null
   processType: ProcessType
   fileName: string
+  fileFormat?: string
+  taskId: string | number | null
+  sourceFileId: string | number | null
   files: { id: string; name: string; count: number; modifiedCount: number }[]
   selectedFileId: string | null
 }>()
@@ -17,18 +22,27 @@ const emit = defineEmits<{
   'update:selectedId': [value: string]
   'update:selectedFileId': [value: string]
   'update:item-content': [id: string, value: string]
+  'restore:item': [id: string]
+  'add:item': []
   'remove:item': [id: string]
 }>()
 
 const sourceViewerRef = ref<HTMLElement | null>(null)
 const search = ref('')
 const currentPage = ref(1)
-const PREVIEW_PAGE_SIZE = 6
+const PREVIEW_PAGE_SIZE = 10
 const editingItemId = ref<string | null>(null)
 const editorDraft = ref('')
 const lines = computed(() => sourceLines(props.sourceText))
 const selectedItem = computed(() => props.items.find((item) => item.id === props.selectedId) ?? props.items[0])
 const editingItem = computed(() => props.items.find((item) => item.id === editingItemId.value))
+const normalizedFileFormat = computed(() => (
+  props.fileFormat?.toLowerCase().replace(/^\./, '')
+  || props.fileName.split('.').pop()?.toLowerCase()
+  || ''
+))
+const isPdfSource = computed(() => normalizedFileFormat.value === 'pdf')
+const isOfficeSource = computed(() => ['docx', 'xlsx'].includes(normalizedFileFormat.value))
 
 const filteredItems = computed(() => props.items.filter((item, index) => {
   const matchesSearch = !search.value.trim()
@@ -71,6 +85,12 @@ function saveEditor() {
   closeEditor()
 }
 
+function restoreItem() {
+  if (!editingItem.value) return
+  emit('restore:item', editingItem.value.id)
+  closeEditor()
+}
+
 function removeItem(item: PreviewItem) {
   selectItem(item.id)
   emit('remove:item', item.id)
@@ -94,7 +114,7 @@ watch(selectedItem, async (item) => {
     currentPage.value = Math.floor(visibleIndex / PREVIEW_PAGE_SIZE) + 1
   }
 
-  if (item.sourceStart == null) return
+  if (isPdfSource.value || isOfficeSource.value || item.sourceStart == null) return
   await nextTick()
   const target = sourceViewerRef.value?.querySelector<HTMLElement>(`[data-source-start="${item.sourceStart}"]`)
     ?? sourceViewerRef.value?.querySelector<HTMLElement>('.source-line.is-highlighted')
@@ -106,6 +126,11 @@ function itemNumber(item: PreviewItem) {
 }
 
 function lineRange(item: PreviewItem) {
+  if (item.sourcePages?.length) {
+    const first = item.sourcePages[0]
+    const last = item.sourcePages[item.sourcePages.length - 1]
+    return first === last ? `来源：第 ${first} 页` : `来源：第 ${first}–${last} 页`
+  }
   if (item.sourceStartLine == null || item.sourceEndLine == null) return '手动新增，无源文件定位'
   return item.sourceStartLine === item.sourceEndLine
     ? `来源：第 ${item.sourceStartLine} 行`
@@ -152,7 +177,22 @@ function lineRange(item: PreviewItem) {
           </div>
         </div>
 
-        <div ref="sourceViewerRef" class="source-viewer" tabindex="0" aria-label="源文件内容">
+        <PdfSourceViewer
+          v-if="isPdfSource"
+          :task-id="taskId"
+          :source-file-id="sourceFileId"
+          :file-name="fileName"
+          :selected-item="selectedItem ?? null"
+        />
+        <OfficeSourceViewer
+          v-else-if="isOfficeSource"
+          :task-id="taskId"
+          :source-file-id="sourceFileId"
+          :file-name="fileName"
+          :file-format="normalizedFileFormat"
+          :selected-item="selectedItem ?? null"
+        />
+        <div v-else ref="sourceViewerRef" class="source-viewer" tabindex="0" aria-label="源文件内容">
           <div
             v-for="line in lines"
             :key="line.number"
@@ -169,7 +209,12 @@ function lineRange(item: PreviewItem) {
       <div class="preview-pane">
         <div class="pane-header">
           <strong>{{ processType === 'unstructured' ? '切片内容' : '记录内容' }}</strong>
-          <span>共 {{ items.length.toLocaleString() }} 条</span>
+          <div class="pane-header-actions">
+            <span>共 {{ items.length.toLocaleString() }} 条</span>
+            <el-button link type="primary" @click="emit('add:item')">
+              <i class="fa fa-plus" /> 手动新增
+            </el-button>
+          </div>
         </div>
 
         <template v-if="!editingItem">
@@ -185,6 +230,8 @@ function lineRange(item: PreviewItem) {
               :key="item.id"
               class="preview-item"
               :class="{ 'is-active': item.id === selectedItem?.id }"
+              :data-preview-id="item.id"
+              :aria-selected="item.id === selectedItem?.id"
               role="button"
               tabindex="0"
               @click="selectItem(item.id)"
@@ -234,6 +281,13 @@ function lineRange(item: PreviewItem) {
             resize="none"
           />
           <div class="editor-actions">
+            <el-button
+              v-if="editingItem.sourceStart != null"
+              link
+              @click="restoreItem"
+            >
+              <i class="fa fa-undo" /> 恢复原始内容
+            </el-button>
             <div>
               <el-button @click="closeEditor">取消</el-button>
               <el-button type="primary" @click="saveEditor">保存修改</el-button>
@@ -288,6 +342,17 @@ function lineRange(item: PreviewItem) {
   white-space: nowrap;
 }
 
+.pane-header-actions {
+  display: flex;
+  align-items: center;
+  gap: 10px;
+
+  > span {
+    color: #8a93a3;
+    font-size: 11px;
+  }
+}
+
 .file-option {
   display: grid;
   grid-template-columns: minmax(0, 1fr) auto auto;

frontend/src/views/data-process/create/ResultEditorStep.vue

@@ -1,22 +1,74 @@
 <script setup lang="ts">
 import { computed, ref } from 'vue'
-import type { ResultItem } from './types'
+import type { BulkResultRegenerationState, PreviewItem, ResultItem } from './types'
 
 const props = defineProps<{
   items: ResultItem[]
+  previewItems: PreviewItem[]
   selectedId: string | null
+  regeneratingResultId: string | null
+  bulkRegeneration: BulkResultRegenerationState
 }>()
 
 const emit = defineEmits<{
   'update:selectedId': [value: string]
   'update:field': [id: string, field: 'instruction' | 'input' | 'output', value: string]
-  'restore:item': [id: string]
+  'regenerate:item': [id: string]
+  'regenerate:all': []
 }>()
 
 const search = ref('')
 const invalidOnly = ref(false)
 const selectedItem = computed(() => props.items.find((item) => item.id === props.selectedId) ?? props.items[0])
 const selectedIndex = computed(() => props.items.findIndex((item) => item.id === selectedItem.value?.id))
+const invalidCount = computed(() => (
+  props.items.filter((item) => item.savedStatus === 'invalid').length
+))
+const bulkRegenerationActive = computed(() => props.bulkRegeneration.status === 'running')
+const bulkRegenerationVisible = computed(() => (
+  props.bulkRegeneration.status !== 'idle' && props.bulkRegeneration.total > 0
+))
+const bulkRegenerationPercentage = computed(() => (
+  props.bulkRegeneration.total > 0
+    ? Math.round((props.bulkRegeneration.completed / props.bulkRegeneration.total) * 100)
+    : 0
+))
+const itemRegenerating = (id: string) => (
+  props.regeneratingResultId === id
+  || props.bulkRegeneration.targetIds.includes(id)
+)
+const selectedItemRegenerating = computed(() => (
+  selectedItem.value ? itemRegenerating(selectedItem.value.id) : false
+))
+const selectedSource = computed(() => {
+  const previewItemId = selectedItem.value?.previewItemId
+  if (!previewItemId) return null
+  return props.previewItems.find((item) => item.id === previewItemId) ?? null
+})
+const selectedSourceContent = computed(() => (
+  selectedSource.value?.editedContent.trim()
+  || selectedSource.value?.originalContent.trim()
+  || ''
+))
+const selectedSourceWasPreprocessed = computed(() => Boolean(
+  selectedSource.value
+  && selectedSource.value.editedContent.trim()
+  && selectedSource.value.editedContent !== selectedSource.value.originalContent
+))
+const selectedSourceMeta = computed(() => {
+  const source = selectedSource.value
+  if (!source) return ''
+  const parts: string[] = []
+  if (source.sourceStartLine != null) {
+    parts.push(
+      source.sourceEndLine != null && source.sourceEndLine !== source.sourceStartLine
+        ? `第 ${source.sourceStartLine}–${source.sourceEndLine} 行`
+        : `第 ${source.sourceStartLine} 行`,
+    )
+  }
+  if (source.tokenCount > 0) parts.push(`${source.tokenCount} Token`)
+  return parts.join(' · ')
+})
 
 const filteredItems = computed(() => props.items.filter((item, index) => {
   const keyword = search.value.trim().toLowerCase()
@@ -37,8 +89,34 @@ function selectRelative(offset: number) {
 <template>
   <section class="result-step">
     <div class="result-workspace">
-      <aside class="result-list-pane">
-        <div class="pane-header"><strong>生成结果</strong><span>共 {{ items.length }} 条</span></div>
+      <aside class="result-list-pane" :class="{ 'has-bulk-progress': bulkRegenerationVisible }">
+        <div class="pane-header result-list-header">
+          <div class="result-list-title"><strong>生成结果</strong><span>共 {{ items.length }} 条</span></div>
+          <el-button
+            v-if="invalidCount > 0"
+            size="small"
+            plain
+            type="primary"
+            :loading="bulkRegenerationActive"
+            :disabled="Boolean(regeneratingResultId) || bulkRegenerationActive"
+            @click="emit('regenerate:all')"
+          >
+            <i v-if="!bulkRegenerationActive" class="fa fa-refresh" style="margin-right: 4px;" />
+            {{ bulkRegenerationActive ? '重新生成中' : `全部重新生成（${invalidCount}）` }}
+          </el-button>
+        </div>
+        <div v-if="bulkRegenerationVisible" class="bulk-regeneration-progress">
+          <div>
+            <span>已处理 {{ bulkRegeneration.completed }} / {{ bulkRegeneration.total }}</span>
+            <span>成功 {{ bulkRegeneration.succeeded }} · 失败 {{ bulkRegeneration.failed }}</span>
+          </div>
+          <el-progress
+            :percentage="bulkRegenerationPercentage"
+            :show-text="false"
+            :stroke-width="5"
+            :color="bulkRegeneration.failed > 0 ? '#d97706' : '#5b50f2'"
+          />
+        </div>
         <div class="result-toolbar">
           <el-input v-model="search" clearable size="small" placeholder="搜索结果">
             <template #prefix><i class="fa fa-search" /></template>
@@ -59,7 +137,12 @@ function selectRelative(offset: number) {
               <strong>{{ item.instruction || '未填写指令' }}</strong>
               <small>{{ item.output || '未填写输出' }}</small>
             </span>
-            <i class="fa" :class="item.status === 'invalid' ? 'fa-exclamation-circle is-error' : 'fa-check-circle is-valid'" />
+            <i v-if="itemRegenerating(item.id)" class="css-spinner" />
+            <i
+              v-else
+              class="fa"
+              :class="item.status === 'invalid' ? 'fa-exclamation-circle is-error' : 'fa-check-circle is-valid'"
+            />
           </button>
         </div>
       </aside>
@@ -69,14 +152,46 @@ function selectRelative(offset: number) {
           <div>
             <strong>结果 #{{ String(selectedIndex + 1).padStart(3, '0') }}</strong>
             <span v-if="selectedItem.status === 'modified'" class="modified-label">已修改</span>
+            <el-tag v-if="selectedItem.split" size="small" effect="plain">{{ selectedItem.split }}</el-tag>
+            <el-tag
+              v-if="selectedItem.qualityScore != null"
+              size="small"
+              :type="selectedItem.qualityScore >= 80 ? 'success' : selectedItem.qualityScore >= 60 ? 'warning' : 'danger'"
+            >质量 {{ selectedItem.qualityScore.toFixed(1) }}</el-tag>
+          </div>
+          <div class="result-header-actions">
+            <el-button
+              v-if="selectedItem.savedStatus === 'invalid'"
+              size="small"
+              plain
+              type="primary"
+              :loading="selectedItemRegenerating"
+              :disabled="bulkRegenerationActive || (Boolean(regeneratingResultId) && !selectedItemRegenerating)"
+              @click="emit('regenerate:item', selectedItem.id)"
+            >
+              <i v-if="!selectedItemRegenerating" class="fa fa-refresh" style="margin-right: 4px;" /> 重新生成
+            </el-button>
           </div>
-          <el-button link @click="emit('restore:item', selectedItem.id)"><i class="fa fa-undo" /> 恢复生成结果</el-button>
         </div>
 
+        <article class="source-reference" aria-labelledby="source-reference-title">
+          <div class="source-reference-heading">
+            <div>
+              <strong id="source-reference-title">原文参照</strong>
+              <span v-if="selectedSourceMeta">{{ selectedSourceMeta }}</span>
+            </div>
+            <el-tag v-if="selectedSourceWasPreprocessed" size="small" effect="plain">已智能预处理</el-tag>
+          </div>
+          <pre v-if="selectedSourceContent">{{ selectedSourceContent }}</pre>
+          <p v-else>当前结果没有关联到可用的原文切片。</p>
+          <small v-if="selectedSourceWasPreprocessed">这里展示的是实际送入模型的预处理后原文，便于核对问题和答案是否有依据。</small>
+        </article>
+
         <div class="field-editor">
           <label>Instruction <em>必填</em></label>
           <el-input
             :model-value="selectedItem.instruction"
+            :disabled="selectedItemRegenerating"
             type="textarea"
             :rows="3"
             @update:model-value="emit('update:field', selectedItem.id, 'instruction', $event)"
@@ -86,6 +201,7 @@ function selectRelative(offset: number) {
           <label>Input <span>选填</span></label>
           <el-input
             :model-value="selectedItem.input"
+            :disabled="selectedItemRegenerating"
             type="textarea"
             :rows="2"
             @update:model-value="emit('update:field', selectedItem.id, 'input', $event)"
@@ -95,6 +211,7 @@ function selectRelative(offset: number) {
           <label>Output <em>必填</em></label>
           <el-input
             :model-value="selectedItem.output"
+            :disabled="selectedItemRegenerating"
             type="textarea"
             :rows="7"
             @update:model-value="emit('update:field', selectedItem.id, 'output', $event)"
@@ -136,6 +253,25 @@ function selectRelative(offset: number) {
   border-right: 1px solid #e5e8ee;
 }
 
+.result-list-title {
+  min-width: 0;
+
+  span {
+    margin-left: 8px;
+    color: #8a93a3;
+    font-size: 11px;
+    font-weight: 400;
+  }
+}
+
+.result-list-header {
+  gap: 10px;
+
+  :deep(.el-button) {
+    flex: none;
+  }
+}
+
 .pane-header {
   display: flex;
   align-items: center;
@@ -169,12 +305,37 @@ function selectRelative(offset: number) {
   border-bottom: 1px solid #edf0f5;
 }
 
+.bulk-regeneration-progress {
+  padding: 9px 12px 10px;
+  background: #fafaff;
+  border-bottom: 1px solid #e8e7ff;
+
+  > div {
+    display: flex;
+    justify-content: space-between;
+    gap: 10px;
+    margin-bottom: 7px;
+    color: #667085;
+    font-size: 10px;
+  }
+}
+
 .result-list {
   height: 476px;
   padding: 7px;
   overflow: auto;
 }
 
+.has-bulk-progress .result-list {
+  height: 420px;
+}
+
+.result-header-actions {
+  display: flex;
+  align-items: center;
+  gap: 8px;
+}
+
 .result-item {
   display: grid;
   grid-template-columns: 46px minmax(0, 1fr) 18px;
@@ -236,10 +397,80 @@ function selectRelative(offset: number) {
   color: #d97706;
 }
 
+.css-spinner {
+  width: 14px;
+  height: 14px;
+  border: 2px solid rgba(91, 80, 242, 0.2);
+  border-top-color: #5b50f2;
+  border-radius: 50%;
+  animation: css-spin 0.8s linear infinite;
+  display: inline-block;
+}
+
+@keyframes css-spin {
+  to { transform: rotate(360deg); }
+}
+
 .result-editor-pane {
   min-width: 0;
 }
 
+.source-reference {
+  margin: 14px 18px 2px;
+  padding: 12px 14px;
+  border: 1px solid #dfe4ec;
+  border-radius: 7px;
+  background: #f8fafc;
+
+  pre {
+    max-height: 168px;
+    margin: 10px 0 0;
+    overflow: auto;
+    color: #344054;
+    font-family: inherit;
+    font-size: 12px;
+    line-height: 1.7;
+    white-space: pre-wrap;
+    word-break: break-word;
+  }
+
+  p {
+    margin: 10px 0 0;
+    color: #98a2b3;
+    font-size: 12px;
+  }
+
+  > small {
+    display: block;
+    margin-top: 8px;
+    color: #8a93a3;
+    font-size: 10px;
+  }
+}
+
+.source-reference-heading {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  gap: 12px;
+
+  > div {
+    display: flex;
+    align-items: baseline;
+    gap: 8px;
+  }
+
+  strong {
+    color: #344054;
+    font-size: 12px;
+  }
+
+  span {
+    color: #8a93a3;
+    font-size: 10px;
+  }
+}
+
 .field-editor {
   padding: 13px 18px 0;
 

frontend/src/views/data-process/create/SourceUploadStep.vue

@@ -9,6 +9,8 @@ const props = defineProps<{
   externalSource: ExternalDataSource
   externalPulling: boolean
   externalConnected: boolean
+  previewBuilding: boolean
+  sourceUploading: boolean
 }>()
 
 const emit = defineEmits<{
@@ -21,26 +23,33 @@ const emit = defineEmits<{
 }>()
 
 const DATA_SOURCE_TYPES = [
-  { value: 'mysql', label: 'MySQL' },
   { value: 'postgresql', label: 'PostgreSQL' },
-  { value: 'mongodb', label: 'MongoDB' },
-  { value: 'api', label: 'REST API' },
 ]
 
 const AUTH_MODES = [
   { value: 'none', label: '免鉴权' },
   { value: 'basic', label: '账号密码' },
-  { value: 'token', label: 'Token' },
 ]
 
 const FILE_PAGE_SIZE = 10
 const currentFilePage = ref(1)
+type FileStage = 'queued' | 'uploading' | 'waiting' | 'processing' | 'success' | 'upload-failed' | 'preview-failed'
+
+const FILE_STAGE_META: Record<FileStage, { label: string; icon: string }> = {
+  queued: { label: '等待上传', icon: 'fa-clock-o' },
+  uploading: { label: '正在上传', icon: 'fa-cloud-upload' },
+  waiting: { label: '等待切分', icon: 'fa-clock-o' },
+  processing: { label: '正在切分', icon: 'fa-spinner fa-spin' },
+  success: { label: '切分完成', icon: 'fa-check-circle' },
+  'upload-failed': { label: '上传失败', icon: 'fa-exclamation-circle' },
+  'preview-failed': { label: '切分失败', icon: 'fa-exclamation-circle' },
+}
 
 const isExternal = computed(() => props.processType === 'external')
 
 const uploadAccept = computed(() => props.processType === 'unstructured'
-  ? '.txt,.md,.pdf,.docx,.doc,.json,.jsonl'
-  : '.json,.jsonl,.csv,.xlsx,.xls')
+  ? '.txt,.md,.markdown,.pdf,.docx,.pptx,.json,.jsonl,.ndjson'
+  : '.json,.jsonl,.ndjson,.csv,.tsv,.xlsx')
 
 const pagedUploadedFiles = computed(() => {
   const start = (currentFilePage.value - 1) * FILE_PAGE_SIZE
@@ -69,6 +78,48 @@ function formatSize(size: number) {
   if (size >= 1024 * 1024) return `${(size / 1024 / 1024).toFixed(1)} MB`
   return `${(size / 1024).toFixed(1)} KB`
 }
+
+function getFileStage(file: UploadedDataFile): FileStage {
+  if (file.status === 'queued' || file.status === 'uploading') return file.status
+  if (file.status === 'failed') return 'upload-failed'
+  if (file.previewStatus === 'processing' || file.previewStatus === 'success') return file.previewStatus
+  if (file.previewStatus === 'failed') return 'preview-failed'
+  return 'waiting'
+}
+
+function getFileProgress(file: UploadedDataFile) {
+  const stage = getFileStage(file)
+  const progress = stage === 'queued' || stage === 'uploading' || stage === 'upload-failed'
+    ? file.uploadProgress
+    : stage === 'waiting' ? 100 : file.previewProgress ?? 0
+  return Math.min(100, Math.max(0, progress))
+}
+
+function isFileProcessing(file: UploadedDataFile) {
+  return getFileStage(file) === 'processing'
+}
+
+function getFileBarPercentage(file: UploadedDataFile) {
+  // Element Plus 的不定进度动画需要非零宽度；这里不作为完成百分比展示。
+  return isFileProcessing(file) ? 100 : getFileProgress(file)
+}
+
+function getFileProgressStatus(file: UploadedDataFile): 'success' | 'exception' | undefined {
+  const stage = getFileStage(file)
+  if (stage === 'waiting' || stage === 'success') return 'success'
+  if (stage === 'upload-failed' || stage === 'preview-failed') return 'exception'
+  return undefined
+}
+
+function getFileProgressText(file: UploadedDataFile) {
+  if (isFileProcessing(file)) return '处理中'
+  if (getFileStage(file) === 'waiting') return '已上传'
+  return `${getFileProgress(file)}%`
+}
+
+function getFileError(file: UploadedDataFile) {
+  return file.uploadError || file.previewError
+}
 </script>
 
 <template>
@@ -101,7 +152,7 @@ function formatSize(size: number) {
           <el-form-item label="地址 / URL">
             <el-input
               :model-value="externalSource.url"
-              placeholder="例如：mysql://host:3306/db 或 https://api.example.com/data"
+              placeholder="例如：postgresql://db.example.com:5432/my_database"
               aria-label="数据源地址或 URL"
               @update:model-value="updateExternalField('url', $event)"
             />
@@ -140,17 +191,6 @@ function formatSize(size: number) {
               @update:model-value="updateExternalField('password', $event)"
             />
           </el-form-item>
-          <el-form-item v-if="externalSource.authMode === 'token'" label="Token">
-            <el-input
-              :model-value="externalSource.token"
-              type="password"
-              show-password
-              autocomplete="off"
-              placeholder="请输入访问 Token"
-              aria-label="数据源访问 Token"
-              @update:model-value="updateExternalField('token', $event)"
-            />
-          </el-form-item>
           <el-form-item label="拉取条数">
             <el-input-number
               :model-value="externalSource.limit"
@@ -162,12 +202,25 @@ function formatSize(size: number) {
               @update:model-value="updateExternalField('limit', Number($event) || 0)"
             />
           </el-form-item>
+          <el-form-item label="只读查询语句" class="external-query-field">
+            <el-input
+              :model-value="externalSource.query"
+              type="textarea"
+              :rows="4"
+              maxlength="20000"
+              show-word-limit
+              placeholder="例如：SELECT question, answer FROM qa_data ORDER BY id"
+              aria-label="外部数据源只读查询语句"
+              @update:model-value="updateExternalField('query', $event)"
+            />
+            <small>只允许单条 SELECT 或 WITH 查询；后端会拒绝写入、DDL 和多语句。</small>
+          </el-form-item>
         </el-form>
 
         <div class="external-actions">
           <el-button
             :loading="externalPulling && !externalConnected"
-            :disabled="externalPulling"
+            :disabled="externalPulling || previewBuilding"
             plain
             @click="emit('test-connection')"
           >
@@ -176,7 +229,7 @@ function formatSize(size: number) {
           <el-button
             type="primary"
             :loading="externalPulling"
-            :disabled="externalPulling"
+            :disabled="externalPulling || previewBuilding"
             @click="emit('pull-data')"
           >
             拉取数据
@@ -200,10 +253,34 @@ function formatSize(size: number) {
                   <template v-if="file.count"> · {{ file.count.toLocaleString() }} 条</template>
                 </span>
               </div>
-              <span class="file-status"><i class="fa fa-check-circle" aria-hidden="true" /> 拉取成功</span>
+              <div
+                class="file-preview-progress"
+                :class="`is-${getFileStage(file)}`"
+                :title="getFileError(file)"
+              >
+                <div class="file-status" role="status" aria-live="polite">
+                  <span>
+                    <i class="fa" :class="FILE_STAGE_META[getFileStage(file)].icon" aria-hidden="true" />
+                    {{ FILE_STAGE_META[getFileStage(file)].label }}
+                  </span>
+                  <span class="file-progress-value">{{ getFileProgressText(file) }}</span>
+                </div>
+                <el-progress
+                  :percentage="getFileBarPercentage(file)"
+                  :indeterminate="isFileProcessing(file)"
+                  :duration="1.5"
+                  :stroke-width="5"
+                  :show-text="false"
+                  :status="getFileProgressStatus(file)"
+                  :aria-valuenow="isFileProcessing(file) ? undefined : getFileProgress(file)"
+                  :aria-valuetext="isFileProcessing(file) ? '正在切分，进度未知' : getFileProgressText(file)"
+                />
+                <small v-if="getFileError(file)" class="file-preview-error">{{ getFileError(file) }}</small>
+              </div>
               <el-button
                 link
                 type="danger"
+                :disabled="previewBuilding || file.status === 'uploading'"
                 :aria-label="`删除数据集 ${file.name}`"
                 @click="emit('remove-file', file.uid)"
               >
@@ -233,7 +310,13 @@ function formatSize(size: number) {
           <h3 id="source-upload-title">源数据上传</h3>
           <p>上传后可在下一步检查内容和切分效果，支持同时添加多个文件</p>
         </div>
-        <el-button v-if="uploadedFiles.length === 0" link type="primary" @click="emit('use-sample')">
+        <el-button
+          v-if="uploadedFiles.length === 0"
+          link
+          type="primary"
+          :disabled="previewBuilding"
+          @click="emit('use-sample')"
+        >
           使用示例数据
         </el-button>
       </div>
@@ -243,6 +326,7 @@ function formatSize(size: number) {
         drag
         multiple
         :accept="uploadAccept"
+        :disabled="previewBuilding"
         :auto-upload="false"
         :show-file-list="false"
         :on-change="(file: UploadFile) => emit('file-change', file)"
@@ -253,25 +337,29 @@ function formatSize(size: number) {
         <template #tip>
           <div class="el-upload__tip">
             {{ processType === 'unstructured'
-              ? '支持 TXT、Markdown、PDF、Word、JSON、JSONL，单文件不超过 200MB'
-              : '支持 JSON、JSONL、CSV、Excel，单文件不超过 200MB' }}
+              ? '支持 TXT、MD、MARKDOWN、PDF、DOCX、PPTX、JSON、JSONL、NDJSON；旧版 DOC/PPT 请先转换，单文件不超过 200MB'
+              : '支持 JSON、JSONL、NDJSON、CSV、TSV、XLSX；旧版 XLS 请先转换，单文件不超过 200MB' }}
           </div>
         </template>
       </el-upload>
 
       <section v-else class="uploaded-file-list" aria-label="已上传文件列表">
         <div class="uploaded-file-list-header">
-          <span>已添加 {{ uploadedFiles.length }} 个文件</span>
+          <span>
+            已选择 {{ uploadedFiles.length }} 个文件
+            <small v-if="sourceUploading" class="upload-queue-status"> · 正在逐个上传</small>
+          </span>
           <div class="continue-upload">
             <el-upload
               multiple
               :accept="uploadAccept"
+              :disabled="previewBuilding"
               :auto-upload="false"
               :show-file-list="false"
               :on-change="(file: UploadFile) => emit('file-change', file)"
               aria-label="继续添加源数据文件"
             >
-              <el-button size="small" type="primary">继续上传</el-button>
+              <el-button size="small" type="primary" :disabled="previewBuilding">继续上传</el-button>
             </el-upload>
           </div>
         </div>
@@ -285,10 +373,34 @@ function formatSize(size: number) {
                 <template v-if="file.count"> · {{ file.count.toLocaleString() }} 条</template>
               </span>
             </div>
-            <span class="file-status"><i class="fa fa-check-circle" aria-hidden="true" /> 校验通过</span>
+            <div
+              class="file-preview-progress"
+              :class="`is-${getFileStage(file)}`"
+              :title="getFileError(file)"
+            >
+              <div class="file-status" role="status" aria-live="polite">
+                <span>
+                  <i class="fa" :class="FILE_STAGE_META[getFileStage(file)].icon" aria-hidden="true" />
+                  {{ FILE_STAGE_META[getFileStage(file)].label }}
+                </span>
+                <span class="file-progress-value">{{ getFileProgressText(file) }}</span>
+              </div>
+              <el-progress
+                :percentage="getFileBarPercentage(file)"
+                :indeterminate="isFileProcessing(file)"
+                :duration="1.5"
+                :stroke-width="5"
+                :show-text="false"
+                :status="getFileProgressStatus(file)"
+                :aria-valuenow="isFileProcessing(file) ? undefined : getFileProgress(file)"
+                :aria-valuetext="isFileProcessing(file) ? '正在切分，进度未知' : getFileProgressText(file)"
+              />
+              <small v-if="getFileError(file)" class="file-preview-error">{{ getFileError(file) }}</small>
+            </div>
             <el-button
               link
               type="danger"
+              :disabled="previewBuilding || file.status === 'uploading'"
               :aria-label="`删除文件 ${file.name}`"
               @click="emit('remove-file', file.uid)"
             >
@@ -441,6 +553,11 @@ function formatSize(size: number) {
   margin-top: 0;
 }
 
+.upload-queue-status {
+  color: #5b50f2;
+  font-size: inherit;
+}
+
 .uploaded-file-pagination {
   display: flex;
   justify-content: flex-end;
@@ -495,9 +612,59 @@ function formatSize(size: number) {
   }
 }
 
+.file-preview-progress {
+  display: flex;
+  flex: 0 1 220px;
+  flex-direction: column;
+  gap: 5px;
+  min-width: 150px;
+
+  &.is-queued .file-status,
+  &.is-waiting .file-status {
+    color: #8a93a3;
+  }
+
+  &.is-uploading .file-status,
+  &.is-processing .file-status {
+    color: #5b50f2;
+  }
+
+  &.is-success .file-status {
+    color: #2ca66a;
+  }
+
+  &.is-upload-failed .file-status,
+  &.is-preview-failed .file-status,
+  .file-preview-error {
+    color: #d94b4b;
+  }
+}
+
 .file-status {
-  color: #2ca66a;
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  gap: 8px;
   font-size: 12px;
+
+  > span:first-child {
+    overflow: hidden;
+    text-overflow: ellipsis;
+    white-space: nowrap;
+  }
+}
+
+.file-progress-value {
+  flex: 0 0 auto;
+  font-variant-numeric: tabular-nums;
+}
+
+.file-preview-error {
+  overflow: hidden;
+  font-size: 11px;
+  line-height: 1.4;
+  text-overflow: ellipsis;
+  white-space: nowrap;
 }
 
 .uploaded-file :deep(.el-button) {
@@ -514,8 +681,12 @@ function formatSize(size: number) {
     padding: 8px 10px;
   }
 
-  .file-status {
+  .file-preview-progress {
     flex: 0 1 auto;
+    min-width: 130px;
+  }
+
+  .file-status {
     line-height: 1.4;
     white-space: normal;
   }
@@ -545,7 +716,8 @@ function formatSize(size: number) {
     min-width: calc(100% - 40px);
   }
 
-  .file-status {
+  .file-preview-progress {
+    flex: 1 0 calc(100% - 38px);
     margin-left: 38px;
   }
 }

frontend/src/views/data-process/create/StructuredOptionsPanel.vue

@@ -4,6 +4,10 @@ import type {
   PreprocessOption,
   StructuredProcessOptions,
 } from './types'
+import {
+  normalizeQaPairsGenerationCount,
+  QA_PAIRS_GENERATION_LIMITS,
+} from './types'
 import DatasetSplitEditor from './DatasetSplitEditor.vue'
 import GenerationOptionsPanel from './GenerationOptionsPanel.vue'
 
@@ -22,12 +26,20 @@ const PREPROCESS_OPTIONS: Array<{
   label: string
   description: string
 }> = [
-  { value: 'clean_invalid', label: '清理无效数据', description: '处理空行、空列和残缺行' },
-  { value: 'detect_structure', label: '识别表格结构', description: '识别表头、多级表头和合并单元格' },
-  { value: 'deduplicate', label: '重复数据去重', description: '删除完全重复或关键字段重复的数据' },
-  { value: 'normalize_format', label: '数据格式标准化', description: '统一日期、数字、单位和枚举值格式' },
-  { value: 'filter_anomaly', label: '异常数据过滤', description: '过滤乱码、无效内容和异常记录' },
-  { value: 'desensitize', label: '敏感信息脱敏', description: '处理姓名、手机号、邮箱等敏感信息' },
+  { value: 'clean_invalid', label: '清理无效数据', description: '清理全空列，并剔除关键字段残缺的数据行' },
+  {
+    value: 'detect_structure',
+    label: '嵌套结构展平',
+    description: '展平嵌套对象和可解析的 JSON 字段；Excel 表头与合并单元格在上传时自动解析',
+  },
+  {
+    value: 'deduplicate',
+    label: '重复记录去重',
+    description: '按整行内容或 id、uuid、key、code、*_id 等身份字段去重，暂不支持自定义组合字段',
+  },
+  { value: 'normalize_format', label: '数据格式标准化', description: '按所选规则统一编码、空白、字段名及 JSON 序列化格式' },
+  { value: 'filter_anomaly', label: '异常数据过滤', description: '使用 IQR 识别数值离群值，并过滤乱码等异常记录' },
+  { value: 'desensitize', label: '敏感信息脱敏', description: '识别并脱敏姓名、手机号、邮箱和身份证号' },
 ]
 
 function updateField<K extends keyof StructuredProcessOptions>(
@@ -41,11 +53,17 @@ function updateGenerationOptions(value: GenerationControlOptions) {
   emit('update:options', { ...props.options, ...value })
 }
 
+function updateQaPairsPerRow(value: number | undefined) {
+  updateField('qaPairsPerRow', normalizeQaPairsGenerationCount(value))
+}
+
 function updatePreprocessOptions(value: Array<string | number | boolean>) {
   const allowedValues = new Set(PREPROCESS_OPTIONS.map((option) => option.value))
-  const preprocessOptions = value.filter(
-    (option): option is PreprocessOption => typeof option === 'string' && allowedValues.has(option as PreprocessOption),
-  )
+  const preprocessOptions = Array.from(new Set(value.filter(
+    (option): option is PreprocessOption => (
+      typeof option === 'string' && allowedValues.has(option as PreprocessOption)
+    ),
+  )))
   updateField('preprocessOptions', preprocessOptions)
 }
 </script>
@@ -95,15 +113,16 @@ function updatePreprocessOptions(value: Array<string | number | boolean>) {
       <div class="generation-option-row">
         <div class="generation-option-copy">
           <strong>每行生成数量</strong>
-          <small>每行结构化数据生成的问答对数量</small>
+          <small>支持 1～50 条；数量越大，处理耗时和 Token 消耗越高</small>
         </div>
         <el-input-number
           :model-value="options.qaPairsPerRow"
-          :min="1"
-          :max="5"
+          :min="QA_PAIRS_GENERATION_LIMITS.min"
+          :max="QA_PAIRS_GENERATION_LIMITS.max"
           :step="1"
+          :precision="0"
           controls-position="right"
-          @update:model-value="updateField('qaPairsPerRow', Number($event) || 1)"
+          @update:model-value="updateQaPairsPerRow"
         />
       </div>
       <DatasetSplitEditor

frontend/src/views/data-process/create/TaskSetupStep.vue

@@ -16,6 +16,7 @@ const props = defineProps<{
   processType: ProcessType
   structuredOptions: StructuredProcessOptions
   unstructuredOptions: UnstructuredProcessOptions
+  processTypeLocked?: boolean
 }>()
 
 const emit = defineEmits<{
@@ -75,9 +76,6 @@ const chunkValidationMessage = computed(() => {
   ) {
     return '重叠长度与最小切片长度之和不能大于切片长度'
   }
-  if (props.unstructuredOptions.chunkMethod === 'custom' && !props.unstructuredOptions.customDelimiter.trim()) {
-    return '请输入自定义分隔符'
-  }
   return ''
 })
 
@@ -156,7 +154,8 @@ defineExpose({ validate })
         <div class="section-title-row">
           <div>
             <h3>处理类型</h3>
-            <p>类型会影响下一步支持的数据源格式和后续预览方式</p>
+            <p v-if="processTypeLocked">重新生成沿用原任务处理类型，不可修改</p>
+            <p v-else>类型会影响下一步支持的数据源格式和后续预览方式</p>
           </div>
         </div>
         <el-form-item prop="processType" class="type-form-item">
@@ -165,6 +164,7 @@ defineExpose({ validate })
               type="button"
               class="type-option"
               :class="{ 'is-active': processType === 'structured' }"
+              :disabled="processTypeLocked"
               @click="emit('update:processType', 'structured')"
             >
               <span class="type-icon"><i class="fa fa-table" /></span>
@@ -178,6 +178,7 @@ defineExpose({ validate })
               type="button"
               class="type-option"
               :class="{ 'is-active': processType === 'unstructured' }"
+              :disabled="processTypeLocked"
               @click="emit('update:processType', 'unstructured')"
             >
               <span class="type-icon"><i class="fa fa-file-text-o" /></span>
@@ -191,6 +192,7 @@ defineExpose({ validate })
               type="button"
               class="type-option"
               :class="{ 'is-active': processType === 'external' }"
+              :disabled="processTypeLocked"
               @click="emit('update:processType', 'external')"
             >
               <span class="type-icon"><i class="fa fa-cloud-download" /></span>
@@ -296,12 +298,25 @@ defineExpose({ validate })
     border-color: #a8a3ff;
   }
 
+  &:disabled {
+    cursor: not-allowed;
+    opacity: 0.58;
+  }
+
+  &:disabled:hover {
+    border-color: #dfe3ea;
+  }
+
   &.is-active {
     background: #fafaff;
     border-color: #5b50f2;
     box-shadow: 0 0 0 1px rgba(91, 80, 242, 0.08);
   }
 
+  &.is-active:disabled {
+    opacity: 1;
+  }
+
   strong,
   small {
     display: block;

frontend/src/views/data-process/create/UnstructuredOptionsPanel.vue

@@ -6,6 +6,7 @@ import type {
   UnstructuredPreprocessOption,
   UnstructuredProcessOptions,
 } from './types'
+import { QA_PAIRS_GENERATION_LIMITS } from './types'
 import DatasetSplitEditor from './DatasetSplitEditor.vue'
 import GenerationOptionsPanel from './GenerationOptionsPanel.vue'
 
@@ -30,17 +31,17 @@ const SMART_PREPROCESS_OPTIONS: UnstructuredPreprocessOption[] = [
 ]
 
 const CHUNK_METHODS: Array<{ value: ChunkMethod; label: string }> = [
-  { value: 'semantic', label: '自动语义切分' },
-  { value: 'heading', label: '按标题和段落' },
-  { value: 'fixed', label: '按固定长度' },
-  { value: 'custom', label: '自定义分隔符' },
+  { value: 'layout_hybrid', label: '版面结构混合切分' },
+  { value: 'semantic', label: '语义切分' },
+  { value: 'fixed', label: '固定 Token 切分' },
 ]
 
 const UNSTRUCTURED_NUMBER_LIMITS = {
   chunkSize: { min: 200, max: 2000 },
   chunkOverlap: { min: 0, max: 500 },
   minChunkSize: { min: 20, max: 500 },
-  qaPairsPerChunk: { min: 1, max: 3 },
+  semanticBreakpointPercentile: { min: 1, max: 99 },
+  qaPairsPerChunk: QA_PAIRS_GENERATION_LIMITS,
 } as const
 
 type UnstructuredNumberField = keyof typeof UNSTRUCTURED_NUMBER_LIMITS
@@ -70,9 +71,9 @@ function updateGenerationOptions(value: GenerationControlOptions) {
 
 function updateSmartPreprocess(value: string | number | boolean) {
   const enabled = Boolean(value)
-  const remainingOptions = props.options.preprocessOptions.filter(
+  const remainingOptions = Array.from(new Set(props.options.preprocessOptions.filter(
     (option) => !SMART_PREPROCESS_OPTIONS.includes(option),
-  )
+  )))
   updateField(
     'preprocessOptions',
     enabled ? [...SMART_PREPROCESS_OPTIONS, ...remainingOptions] : remainingOptions,
@@ -80,9 +81,11 @@ function updateSmartPreprocess(value: string | number | boolean) {
 }
 
 function updateDesensitize(value: string | number | boolean) {
-  const preprocessOptions: UnstructuredPreprocessOption[] = props.options.preprocessOptions.filter(
-    (option) => option !== 'desensitize',
-  )
+  const preprocessOptions: UnstructuredPreprocessOption[] = Array.from(new Set(
+    props.options.preprocessOptions.filter(
+      (option) => option !== 'desensitize',
+    ),
+  ))
   if (Boolean(value)) preprocessOptions.push('desensitize')
   updateField('preprocessOptions', preprocessOptions)
 }
@@ -116,7 +119,7 @@ defineExpose({ revealValidation })
     <div class="section-title-row">
       <div>
         <h3>预处理选项</h3>
-        <p>默认使用推荐策略，只需决定是否需要脱敏</p>
+        <p>默认启用结构感知的推荐策略，只需决定是否需要脱敏</p>
       </div>
     </div>
     <div class="preprocess-option-grid">
@@ -127,7 +130,7 @@ defineExpose({ revealValidation })
         />
         <span class="preprocess-option-copy">
           <strong>智能预处理</strong>
-          <small>自动清理、解析、去重及保留上下文</small>
+          <small>清理页眉页脚、页码、目录和无效内容，感知文档结构、合并短块、预过滤低质量内容、近重复去重，并通过重叠保护上下文</small>
         </span>
       </label>
 
@@ -138,7 +141,7 @@ defineExpose({ revealValidation })
         />
         <span class="preprocess-option-copy">
           <strong>敏感信息脱敏</strong>
-          <small>处理姓名、手机号等隐私信息</small>
+          <small>识别并脱敏姓名、手机号、邮箱和身份证号</small>
         </span>
       </label>
 
@@ -159,7 +162,7 @@ defineExpose({ revealValidation })
     <div class="section-title-row">
       <div>
         <h3>切分选项</h3>
-        <p>以语义完整为优先，将长文档拆成可独立生成问答的内容块</p>
+        <p>优先保留文档版面结构，也可按主题变化或固定 Token 切分</p>
       </div>
     </div>
 
@@ -178,7 +181,9 @@ defineExpose({ revealValidation })
             :value="method.value"
           />
         </el-select>
-        <small>推荐使用自动语义切分，在长度限制内优先保留完整句段</small>
+        <small v-if="options.chunkMethod === 'layout_hybrid'">默认推荐，按标题、段落、列表和表格结构切分，并合并过短内容块</small>
+        <small v-else-if="options.chunkMethod === 'semantic'">根据相邻内容的语义变化寻找主题边界，适合长文章和知识材料</small>
+        <small v-else>按句子边界控制固定 Token 长度，速度快且结果稳定</small>
       </label>
 
       <label class="config-field">
@@ -234,17 +239,22 @@ defineExpose({ revealValidation })
           <small>过短的尾部内容会尽量并入前一个切片</small>
         </label>
 
-        <label v-if="options.chunkMethod === 'custom'" class="config-field">
-          <span class="config-field-label">自定义分隔符</span>
-          <el-input
-            :model-value="options.customDelimiter"
-            maxlength="40"
-            show-word-limit
-            placeholder="例如：--- 或 ###"
-            aria-label="自定义分隔符"
-            @update:model-value="updateField('customDelimiter', $event)"
-          />
-          <small>系统会优先在分隔符位置结束当前切片</small>
+        <label v-if="options.chunkMethod === 'semantic'" class="config-field">
+          <span class="config-field-label">语义断点百分位</span>
+          <span class="unit-input">
+            <el-input-number
+              :model-value="options.semanticBreakpointPercentile"
+              :min="1"
+              :max="99"
+              :step="1"
+              :precision="0"
+              controls-position="right"
+              aria-label="语义断点百分位"
+              @update:model-value="updateUnstructuredNumber('semanticBreakpointPercentile', $event)"
+            />
+            <span>%</span>
+          </span>
+          <small>百分位越低切片越多，推荐保持 95%</small>
         </label>
     </div>
 
@@ -276,12 +286,12 @@ defineExpose({ revealValidation })
       <div class="generation-option-row">
         <div class="generation-option-copy">
           <strong>每个切片生成数量</strong>
-          <small>每个内容切片最多生成 3 个不同角度的问答对</small>
+          <small>支持 1～50 条；数量越大，处理耗时和 Token 消耗越高</small>
         </div>
         <el-input-number
           :model-value="options.qaPairsPerChunk"
-          :min="1"
-          :max="3"
+          :min="QA_PAIRS_GENERATION_LIMITS.min"
+          :max="QA_PAIRS_GENERATION_LIMITS.max"
           :step="1"
           :precision="0"
           controls-position="right"

frontend/src/views/data-process/create/data-process-create.scss

@@ -117,6 +117,12 @@
   min-height: 400px;
 }
 
+.initialization-error {
+  display: grid;
+  justify-items: start;
+  gap: 16px;
+}
+
 .wizard-footer {
   display: grid;
   flex-shrink: 0;

frontend/src/views/data-process/create/dataProcessCreateState.ts

@@ -1,6 +1,99 @@
-import type { StructuredProcessOptions, UnstructuredProcessOptions } from './types'
+import type { DataProcessConfig, DataProcessDatasetSplit } from '@/types/dataProcess'
+import type {
+  GenerationControlOptions,
+  PreprocessOption,
+  StructuredProcessOptions,
+  UnstructuredPreprocessOption,
+  UnstructuredProcessOptions,
+  ProcessType,
+} from './types'
+import { normalizeQaPairsGenerationCount } from './types'
 
-export const DEFAULT_GENERATION_PROMPT = '你是一名专业的数据生成助手。请根据输入内容生成准确、完整、可直接用于模型训练的问答数据。仅输出符合目标格式的内容，答案应事实清晰、语言自然，不要添加分析过程、说明或无关内容。'
+const LEGACY_DEFAULT_GENERATION_PROMPT = '你是一名专业的数据生成助手。请根据输入内容生成准确、完整、可直接用于模型训练的问答数据。仅输出符合所选输出类型和目标格式的内容，答案应事实清晰、语言自然，不要添加无关说明。'
+
+const PREVIOUS_DEFAULT_STANDARD_GENERATION_PROMPT = '你是一名专业的数据生成助手。请严格依据输入内容生成准确、完整、可直接用于监督微调的问答数据。只生成问题和最终答案，不输出分析、推理过程或来源中不存在的信息；答案应事实清晰、语言自然。'
+
+const PREVIOUS_DEFAULT_REASONING_GENERATION_PROMPT = '你是一名专业的推理数据生成助手。请严格依据输入内容生成问题、可核验的推理过程和最终答案。推理需要说明关键依据与必要步骤，不得引入来源中不存在的事实；最终答案应准确、完整且语言自然。'
+
+export const PREVIOUS_DEFAULT_STANDARD_GENERATION_PROMPT_2 = `你是一名专业的数据生成助手。请根据下方来源内容生成可用于监督微调的问答数据。
+来源内容：
+{{ content }}
+要求：严格依据来源生成问题和最终答案，不输出分析、中间思考过程或来源中不存在的信息；答案应准确、完整、语言自然。`
+
+export const PREVIOUS_DEFAULT_REASONING_GENERATION_PROMPT_2 = `你是一名专业的推理数据生成助手。请根据下方来源内容生成可用于推理模型训练的问答数据。
+来源内容：
+{{ content }}
+要求：每条数据必须包含问题、可核验的中间思考过程和最终答案。中间思考过程需说明关键依据、必要步骤和中间计算，不得跳过推理只给结论，也不得引入来源中不存在的事实。`
+
+export const PREVIOUS_DEFAULT_STANDARD_GENERATION_PROMPT_3 = `你是一名专业的数据生成专家。你的任务是根据下方的来源内容，提取核心信息并生成高质量的问答数据，用于大模型的监督微调（SFT）。
+
+来源内容：
+{{ content }}
+
+任务要求：
+1. 问题构建：基于来源内容设计问题，问题应当具有实际意义，不仅限于表面的事实问答，可以包括对内容的主旨概括、具体细节的询问或应用场景的探讨。
+2. 答案生成：严格依据来源内容生成最终答案，答案需准确无误、逻辑连贯、完整且表达自然流畅。
+3. 忠于原文：不输出分析、中间思考过程或任何多余的解释，切勿引入来源内容中未提及的事实或主观臆造的信息。`
+
+export const PREVIOUS_DEFAULT_REASONING_GENERATION_PROMPT_3 = `你是一名具有深厚专业知识的推理数据生成专家。你的任务是根据给定的来源内容，生成高质量、逻辑严密且可用于训练推理大语言模型（如 O1 或 DeepSeek-R1）的问答数据。
+
+来源内容：
+{{ content }}
+
+任务要求：
+1. 多维度提问：从来源内容中提取核心概念或关键逻辑，构建具有挑战性的问题，避免简单的信息抽取，提问可以涉及因果分析、方案设计、逻辑推断或跨段落信息整合。
+2. 深度思维链（CoT）：
+   - 必须提供详尽的中间思考过程。
+   - 在思考过程中，需首先分析问题的核心需求，然后逐步拆解来源内容中的相关事实或数据。
+   - 对于涉及计算或多步推导的问题，必须一步步展示计算或推导过程，并说明每一步的理论依据或事实来源。
+   - 如果遇到矛盾或缺失的信息，需在思考过程中指出，并合理假设或推理，但最终答案不得编造来源中完全不相关的事实。
+3. 最终答案：在充分推理后给出明确的最终结论。结论应当是对问题的直接回应，语言自然、结构清晰，并且结论必须能从前面的思考过程中自然得出。
+
+数据格式要求：
+请确保每条生成的数据清晰包含“问题”、“中间思考过程”和“最终答案”三个部分，且内容充实、专业。`
+
+export const DEFAULT_STANDARD_GENERATION_PROMPT = `你是一名专业的数据生成专家。请基于下方来源内容，生成高质量、可直接用于大模型监督微调（SFT）的问答数据。
+
+来源内容：
+{{ content }}
+
+任务要求：
+1. 提问有价值：问题应聚焦来源内容的核心信息点，避免无意义的字面抄录或过于宽泛的提问；优先设计需要理解、归纳或推理才能回答的问题，例如主旨概括、细节追问、因果解释或场景应用。
+2. 答案忠于来源：严格依据来源内容作答，不得引入来源中不存在的事实、观点或数据；若来源信息不足以完整回答该问题，请如实说明不足，不要编造或臆测。
+3. 答案完整且自然：回答应直接回应问题，要点完整、逻辑清晰、语言流畅；涉及多项内容时建议分点陈述，但不必照搬原文措辞。`
+
+export const DEFAULT_REASONING_GENERATION_PROMPT = `你是一名具有深厚专业知识的推理数据生成专家。请基于下方来源内容，生成包含完整思维链、可用于训练推理大语言模型（如 O1、DeepSeek-R1）的问答数据。
+
+来源内容：
+{{ content }}
+
+任务要求：
+1. 提问有挑战性：从来源内容中提炼需要多步思考才能回答的问题，避免可直接定位答案的简单事实抽取；可涉及因果分析、逻辑推断或跨信息整合。
+2. 推理可核验：逐步展开思考过程，每个关键结论都应标注其在来源中的依据或必要的中间推导；不得跳过关键步骤只给结论，也不得引入来源中完全不存在的事实。
+3. 信息不足时如实处理：遇到来源信息缺失或相互矛盾时，需在推理中明确指出，并基于来源做出合理假设，但不得编造与来源无关的内容。
+4. 最终答案须能从前面的推理过程中自然得出，语言简洁、结构清晰。`
+
+export function defaultGenerationPrompt(outputType: 'standard' | 'reasoning') {
+  return outputType === 'reasoning'
+    ? DEFAULT_REASONING_GENERATION_PROMPT
+    : DEFAULT_STANDARD_GENERATION_PROMPT
+}
+
+export function isBuiltInGenerationPrompt(value: string) {
+  const normalize = (s: string) => s.replace(/\r\n/g, '\n').trim()
+  const normalizedValue = normalize(value)
+  return [
+    LEGACY_DEFAULT_GENERATION_PROMPT,
+    PREVIOUS_DEFAULT_STANDARD_GENERATION_PROMPT,
+    PREVIOUS_DEFAULT_STANDARD_GENERATION_PROMPT_2,
+    PREVIOUS_DEFAULT_STANDARD_GENERATION_PROMPT_3,
+    PREVIOUS_DEFAULT_REASONING_GENERATION_PROMPT,
+    PREVIOUS_DEFAULT_REASONING_GENERATION_PROMPT_2,
+    PREVIOUS_DEFAULT_REASONING_GENERATION_PROMPT_3,
+    DEFAULT_STANDARD_GENERATION_PROMPT,
+    DEFAULT_REASONING_GENERATION_PROMPT,
+  ].map(normalize).includes(normalizedValue)
+}
 
 export function createDefaultStructuredOptions(): StructuredProcessOptions {
   return {
@@ -9,7 +102,9 @@ export function createDefaultStructuredOptions(): StructuredProcessOptions {
     qaPairsPerRow: 1,
     datasetSplit: { train: 80, validation: 10, test: 10 },
     generationModelId: '',
-    generationPrompt: DEFAULT_GENERATION_PROMPT,
+    generationPrompt: DEFAULT_STANDARD_GENERATION_PROMPT,
+    outputType: 'standard',
+    reasoningDetail: 'normal',
     temperature: 0.7,
     maxTokens: 1024,
     jsonMode: false,
@@ -30,11 +125,11 @@ export function createDefaultUnstructuredOptions(): UnstructuredProcessOptions {
       'deduplicate_content',
       'preserve_context',
     ],
-    chunkMethod: 'semantic',
+    chunkMethod: 'layout_hybrid',
     chunkSize: 800,
     chunkOverlap: 100,
     minChunkSize: 100,
-    customDelimiter: '',
+    semanticBreakpointPercentile: 95,
     preserveTables: true,
     preserveCodeBlocks: true,
     preserveLists: true,
@@ -42,7 +137,9 @@ export function createDefaultUnstructuredOptions(): UnstructuredProcessOptions {
     qaPairsPerChunk: 1,
     datasetSplit: { train: 80, validation: 10, test: 10 },
     generationModelId: '',
-    generationPrompt: DEFAULT_GENERATION_PROMPT,
+    generationPrompt: DEFAULT_STANDARD_GENERATION_PROMPT,
+    outputType: 'standard',
+    reasoningDetail: 'normal',
     temperature: 0.7,
     maxTokens: 1024,
     jsonMode: false,
@@ -52,3 +149,186 @@ export function createDefaultUnstructuredOptions(): UnstructuredProcessOptions {
     minOutputLength: 20,
   }
 }
+
+function configValue<T>(config: DataProcessConfig, key: string, fallback: T): T {
+  return Object.prototype.hasOwnProperty.call(config, key) ? config[key] as T : fallback
+}
+
+function numberValue(config: DataProcessConfig, key: string, fallback: number): number {
+  const value = Number(configValue(config, key, fallback))
+  return Number.isFinite(value) ? value : fallback
+}
+
+function datasetSplitValue(config: DataProcessConfig, fallback: DataProcessDatasetSplit) {
+  const value = config.dataset_split
+  if (!value || typeof value !== 'object') return { ...fallback }
+  const split = value as unknown as Record<string, unknown>
+  const splitNumber = (key: keyof DataProcessDatasetSplit) => {
+    const parsed = Number(split[key])
+    return Number.isFinite(parsed) ? parsed : fallback[key]
+  }
+  return {
+    train: splitNumber('train'),
+    validation: splitNumber('validation'),
+    test: splitNumber('test'),
+  }
+}
+
+function generationOptionsFromConfig(
+  config: DataProcessConfig,
+  defaults: GenerationControlOptions,
+): GenerationControlOptions {
+  const outputType = configValue(config, 'output_type', defaults.outputType) === 'reasoning'
+    ? 'reasoning'
+    : 'standard'
+  const configuredPrompt = String(
+    configValue(config, 'generation_prompt', defaults.generationPrompt),
+  )
+  return {
+    generationModelId: configValue(config, 'generation_model_id', defaults.generationModelId),
+    generationPrompt: isBuiltInGenerationPrompt(configuredPrompt)
+      ? defaultGenerationPrompt(outputType)
+      : configuredPrompt,
+    outputType,
+    reasoningDetail: configValue(config, 'reasoning_detail', defaults.reasoningDetail) === 'detailed'
+      ? 'detailed'
+      : 'normal',
+    temperature: numberValue(config, 'temperature', defaults.temperature),
+    maxTokens: numberValue(config, 'max_tokens', defaults.maxTokens),
+    jsonMode: Boolean(configValue(config, 'json_mode', defaults.jsonMode)),
+    qualityFilterEnabled: Boolean(configValue(
+      config,
+      'quality_filter_enabled',
+      defaults.qualityFilterEnabled,
+    )),
+    filterLowQuality: Boolean(configValue(
+      config,
+      'filter_low_quality',
+      defaults.filterLowQuality,
+    )),
+    filterShortContent: Boolean(configValue(
+      config,
+      'filter_short_content',
+      defaults.filterShortContent,
+    )),
+    minOutputLength: numberValue(config, 'min_output_length', defaults.minOutputLength),
+  }
+}
+
+export function createStructuredOptionsFromConfig(config: DataProcessConfig): StructuredProcessOptions {
+  const defaults = createDefaultStructuredOptions()
+  const preprocessOptions = configValue<unknown>(config, 'preprocess_options', [])
+  return {
+    ...defaults,
+    ...generationOptionsFromConfig(config, defaults),
+    preprocessOptions: Array.isArray(preprocessOptions)
+      ? preprocessOptions.map(String) as PreprocessOption[]
+      : defaults.preprocessOptions,
+    semanticEnrichment: Boolean(configValue(
+      config,
+      'semantic_enrichment',
+      defaults.semanticEnrichment,
+    )),
+    qaPairsPerRow: normalizeQaPairsGenerationCount(
+      numberValue(config, 'qa_pairs_per_row', defaults.qaPairsPerRow),
+      defaults.qaPairsPerRow,
+    ),
+    datasetSplit: datasetSplitValue(config, defaults.datasetSplit),
+  }
+}
+
+export function createUnstructuredOptionsFromConfig(config: DataProcessConfig): UnstructuredProcessOptions {
+  const defaults = createDefaultUnstructuredOptions()
+  const preprocessOptions = configValue<unknown>(config, 'preprocess_options', [])
+  return {
+    ...defaults,
+    ...generationOptionsFromConfig(config, defaults),
+    preprocessOptions: Array.isArray(preprocessOptions)
+      ? preprocessOptions.map(String) as UnstructuredPreprocessOption[]
+      : defaults.preprocessOptions,
+    chunkMethod: configValue(config, 'chunk_method', defaults.chunkMethod),
+    chunkSize: numberValue(config, 'chunk_size', defaults.chunkSize),
+    chunkOverlap: numberValue(config, 'chunk_overlap', defaults.chunkOverlap),
+    minChunkSize: numberValue(config, 'min_chunk_size', defaults.minChunkSize),
+    semanticBreakpointPercentile: numberValue(
+      config,
+      'semantic_breakpoint_percentile',
+      defaults.semanticBreakpointPercentile,
+    ),
+    preserveTables: Boolean(configValue(config, 'preserve_tables', defaults.preserveTables)),
+    preserveCodeBlocks: Boolean(configValue(
+      config,
+      'preserve_code_blocks',
+      defaults.preserveCodeBlocks,
+    )),
+    preserveLists: Boolean(configValue(config, 'preserve_lists', defaults.preserveLists)),
+    semanticEnrichment: Boolean(configValue(
+      config,
+      'semantic_enrichment',
+      defaults.semanticEnrichment,
+    )),
+    qaPairsPerChunk: normalizeQaPairsGenerationCount(
+      numberValue(config, 'qa_pairs_per_chunk', defaults.qaPairsPerChunk),
+      defaults.qaPairsPerChunk,
+    ),
+    datasetSplit: datasetSplitValue(config, defaults.datasetSplit),
+  }
+}
+
+export function previewAffectingOptionsFor(
+  processType: ProcessType,
+  structured: StructuredProcessOptions,
+  unstructured: UnstructuredProcessOptions,
+) {
+  if (processType === 'structured') return { preprocessOptions: structured.preprocessOptions }
+  if (processType !== 'unstructured') return null
+  const {
+    preprocessOptions,
+    chunkMethod,
+    chunkSize,
+    chunkOverlap,
+    minChunkSize,
+    semanticBreakpointPercentile,
+    preserveTables,
+    preserveCodeBlocks,
+    preserveLists,
+  } = unstructured
+  return {
+    preprocessOptions,
+    chunkMethod,
+    chunkSize,
+    chunkOverlap,
+    minChunkSize,
+    semanticBreakpointPercentile,
+    preserveTables,
+    preserveCodeBlocks,
+    preserveLists,
+  }
+}
+
+export function generationAffectingOptionsFor(
+  processType: ProcessType,
+  structured: StructuredProcessOptions,
+  unstructured: UnstructuredProcessOptions,
+) {
+  const options = processType === 'unstructured' ? unstructured : structured
+  if (processType === 'external') return null
+  const common = {
+    semanticEnrichment: options.semanticEnrichment,
+    datasetSplit: options.datasetSplit,
+    generationModelId: options.generationModelId,
+    generationPrompt: options.generationPrompt,
+    outputType: options.outputType,
+    reasoningDetail: options.reasoningDetail,
+    temperature: options.temperature,
+    maxTokens: options.maxTokens,
+    jsonMode: options.jsonMode,
+    qualityFilterEnabled: options.qualityFilterEnabled,
+    filterLowQuality: options.filterLowQuality,
+    filterShortContent: options.filterShortContent,
+    minOutputLength: options.minOutputLength,
+  }
+  return processType === 'unstructured'
+    ? { ...common, qaPairsPerChunk: unstructured.qaPairsPerChunk }
+    : { ...common, qaPairsPerRow: structured.qaPairsPerRow }
+}

frontend/src/views/data-process/create/previewModel.ts

@@ -1,12 +1,6 @@
-import type {
-  PreviewItem,
-  ProcessType,
-  ResultItem,
-  SourceLine,
-  StructuredProcessOptions,
-  UnstructuredProcessOptions,
-} from './types'
+import type { SourceLine } from './types'
 
+/** 仅用于“使用示例”上传；正式预览和切片全部由后端生成。 */
 export const DEFAULT_SOURCE_TEXT = [
   '问：如何看待当前的通货膨胀风险？',
   '答：当前通胀水平总体可控，但仍需关注能源价格与供给扰动。',
@@ -14,26 +8,13 @@ export const DEFAULT_SOURCE_TEXT = [
   '答：会议时间以美联储官方日历为准，市场会重点关注利率路径指引。',
   '问：人民币汇率未来走势如何？',
   '答：人民币汇率取决于中美利差、经济基本面与政策预期。',
-  '问：银行理财产品收益率为何持续走低？',
-  '答：主要与市场利率下行、资产端收益下降以及风险偏好变化有关。',
   '问：什么是复利？',
-  '答：复利是指在计算利息时，将上一期利息加入本金，再计算下一期利息。',
-  '问：如何评估股票的投资价值？',
-  '答：评估股票投资价值可以从以下几个方面进行：',
-  '1. 公司基本面：分析公司的财务状况、盈利能力、成长性等。',
-  '2. 行业前景：考察公司所处行业的发展趋势和竞争格局。',
-  '3. 估值水平：通过市盈率、市净率等指标判断估值是否合理。',
-  '4. 财务健康：关注公司的负债情况、现金流状况等。',
-  '5. 管理团队：评估管理层的能力和过往业绩。',
-  '此外，还需要关注宏观经济环境、政策变化等因素对股票市场的影响。',
-  '问：债券和股票的主要区别是什么？',
-  '答：债券收益相对稳定但上行有限，股票波动更大且承担更高风险。',
-  '问：什么是市盈率？',
-  '答：市盈率是股票价格与每股收益的比值，常用于衡量估值水平。',
-  '问：如何进行资产配置？',
-  '答：应根据投资目标、风险承受能力和市场环境合理分配资产。',
+  '答：复利是将上一期利息加入本金，再计算下一期利息。',
 ].join('\n')
 
+/**
+ * 把后端返回的字符偏移映射为源文件行，仅负责界面高亮，不参与切片。
+ */
 export function sourceLines(sourceText: string): SourceLine[] {
   const rawLines = sourceText.split('\n')
   let cursor = 0
@@ -46,495 +27,7 @@ export function sourceLines(sourceText: string): SourceLine[] {
   })
 }
 
-interface SourceRange {
-  start: number
-  end: number
-}
-
-interface ProtectedRange extends SourceRange {
-  kind: 'code' | 'table' | 'list'
-}
-
-const DEFAULT_CHUNK_SIZE = 800
-const DEFAULT_CHUNK_OVERLAP = 100
-const DEFAULT_MIN_CHUNK_SIZE = 100
-
-function finiteInteger(value: number | undefined, fallback: number, min: number): number {
-  return Number.isFinite(value) ? Math.max(min, Math.round(value as number)) : fallback
-}
-
-function trimSourceRange(sourceText: string, start: number, end: number): SourceRange {
-  let nextStart = Math.max(0, start)
-  let nextEnd = Math.min(sourceText.length, end)
-
-  while (nextStart < nextEnd && /\s/.test(sourceText[nextStart])) nextStart += 1
-  while (nextEnd > nextStart && /\s/.test(sourceText[nextEnd - 1])) nextEnd -= 1
-
-  return { start: nextStart, end: nextEnd }
-}
-
-function normalizeDelimiter(delimiter: string | undefined): string {
-  return (delimiter ?? '').replace(/\\n/g, '\n').replace(/\\t/g, '\t')
-}
-
-function overlapsRange(line: SourceLine, range: SourceRange): boolean {
-  return line.start < range.end && line.end > range.start
-}
-
-function isLineProtected(line: SourceLine, ranges: SourceRange[]): boolean {
-  return ranges.some((range) => overlapsRange(line, range))
-}
-
-function detectCodeBlockRanges(sourceText: string, lines: SourceLine[]): ProtectedRange[] {
-  const ranges: ProtectedRange[] = []
-  let openFence: { start: number; marker: string; length: number } | null = null
-
-  for (const line of lines) {
-    const fence = line.content.match(/^\s*(`{3,}|~{3,})/)
-    if (!fence) continue
-
-    const marker = fence[1][0]
-    if (!openFence) {
-      openFence = { start: line.start, marker, length: fence[1].length }
-      continue
-    }
-
-    if (marker === openFence.marker && fence[1].length >= openFence.length) {
-      ranges.push({ start: openFence.start, end: line.end, kind: 'code' })
-      openFence = null
-    }
-  }
-
-  if (openFence) ranges.push({ start: openFence.start, end: sourceText.length, kind: 'code' })
-  return ranges
-}
-
-function isTableSeparator(content: string): boolean {
-  const normalized = content.trim().replace(/^\|/, '').replace(/\|$/, '')
-  const cells = normalized.split('|').map((cell) => cell.trim())
-  return cells.length >= 2 && cells.every((cell) => /^:?-{3,}:?$/.test(cell))
-}
-
-function detectTableRanges(lines: SourceLine[], codeRanges: SourceRange[]): ProtectedRange[] {
-  const ranges: ProtectedRange[] = []
-
-  for (let index = 0; index < lines.length - 1; index += 1) {
-    const header = lines[index]
-    const separator = lines[index + 1]
-    if (
-      isLineProtected(header, codeRanges)
-      || isLineProtected(separator, codeRanges)
-      || !header.content.includes('|')
-      || !isTableSeparator(separator.content)
-    ) {
-      continue
-    }
-
-    let endIndex = index + 1
-    while (
-      endIndex + 1 < lines.length
-      && !isLineProtected(lines[endIndex + 1], codeRanges)
-      && lines[endIndex + 1].content.trim()
-      && lines[endIndex + 1].content.includes('|')
-    ) {
-      endIndex += 1
-    }
-
-    ranges.push({ start: header.start, end: lines[endIndex].end, kind: 'table' })
-    index = endIndex
-  }
-
-  return ranges
-}
-
-function isListItem(content: string): boolean {
-  return /^\s*(?:[-+*]|\d+[.)])\s+\S/.test(content)
-}
-
-function isListContinuation(content: string): boolean {
-  return /^\s{2,}\S/.test(content)
-}
-
-function detectListRanges(
-  lines: SourceLine[],
-  excludedRanges: SourceRange[],
-): ProtectedRange[] {
-  const ranges: ProtectedRange[] = []
-
-  for (let index = 0; index < lines.length; index += 1) {
-    if (isLineProtected(lines[index], excludedRanges) || !isListItem(lines[index].content)) continue
-
-    let endIndex = index
-    let itemCount = 1
-    while (endIndex + 1 < lines.length && !isLineProtected(lines[endIndex + 1], excludedRanges)) {
-      const nextContent = lines[endIndex + 1].content
-      if (isListItem(nextContent)) {
-        itemCount += 1
-        endIndex += 1
-        continue
-      }
-      if (isListContinuation(nextContent)) {
-        endIndex += 1
-        continue
-      }
-      break
-    }
-
-    if (itemCount >= 2) {
-      ranges.push({ start: lines[index].start, end: lines[endIndex].end, kind: 'list' })
-      index = endIndex
-    }
-  }
-
-  return ranges
-}
-
-function mergeProtectedRanges(ranges: ProtectedRange[]): ProtectedRange[] {
-  return ranges
-    .sort((left, right) => left.start - right.start || left.end - right.end)
-    .reduce<ProtectedRange[]>((merged, range) => {
-      const previous = merged[merged.length - 1]
-      if (previous && range.start < previous.end) {
-        previous.end = Math.max(previous.end, range.end)
-        return merged
-      }
-      merged.push({ ...range })
-      return merged
-    }, [])
-}
-
-function protectedRangesForOptions(
-  sourceText: string,
-  options?: UnstructuredProcessOptions,
-): ProtectedRange[] {
-  if (!options?.preserveCodeBlocks && !options?.preserveTables && !options?.preserveLists) return []
-
-  const lines = sourceLines(sourceText)
-  const codeRanges = detectCodeBlockRanges(sourceText, lines)
-  const tableRanges = detectTableRanges(lines, codeRanges)
-  const listRanges = detectListRanges(lines, [...codeRanges, ...tableRanges])
-  const enabledRanges = [
-    ...(options?.preserveCodeBlocks ? codeRanges : []),
-    ...(options?.preserveTables ? tableRanges : []),
-    ...(options?.preserveLists ? listRanges : []),
-  ]
-
-  return mergeProtectedRanges(enabledRanges)
-}
-
-function protectedRangeContaining(
-  ranges: ProtectedRange[],
-  offset: number,
-): ProtectedRange | undefined {
-  return ranges.find((range) => range.start < offset && offset < range.end)
-}
-
-function normalizeChunkStart(
-  sourceText: string,
-  cursor: number,
-  protectedRanges: ProtectedRange[],
-): number {
-  let start = Math.max(0, Math.min(cursor, sourceText.length))
-  const overlapBlock = protectedRangeContaining(protectedRanges, start)
-  if (overlapBlock) start = overlapBlock.end
-
-  while (start < sourceText.length && /\s/.test(sourceText[start])) start += 1
-
-  // 去除块前空白时可能进入缩进代码块/列表；此时恢复到完整块起点。
-  const blockAfterTrim = protectedRangeContaining(protectedRanges, start)
-  if (blockAfterTrim) return cursor <= blockAfterTrim.start ? blockAfterTrim.start : blockAfterTrim.end
-
-  return start
-}
-
-function protectChunkEnd(
-  proposedEnd: number,
-  start: number,
-  minimumEnd: number,
-  protectedRanges: ProtectedRange[],
-): number {
-  const splitBlock = protectedRangeContaining(protectedRanges, proposedEnd)
-  if (!splitBlock) return proposedEnd
-
-  // 优先在块前结束；块前不足最小切片长度时，将整个块收入当前切片。
-  return splitBlock.start > start && splitBlock.start >= minimumEnd
-    ? splitBlock.start
-    : splitBlock.end
-}
-
-function restoreProtectedEdges(
-  range: SourceRange,
-  rawStart: number,
-  rawEnd: number,
-  protectedRanges: ProtectedRange[],
-): SourceRange {
-  const nextRange = { ...range }
-  const startBlock = protectedRangeContaining(protectedRanges, nextRange.start)
-  if (startBlock && rawStart <= startBlock.start) nextRange.start = startBlock.start
-
-  const endBlock = protectedRangeContaining(protectedRanges, nextRange.end)
-  if (endBlock && rawEnd >= endBlock.end) nextRange.end = endBlock.end
-  return nextRange
-}
-
-function lastBoundaryInRange(
-  sourceText: string,
-  idealEnd: number,
-  minimumEnd: number,
-): number | null {
-  const candidates: number[] = []
-  const boundaryTokens = ['\n\n', '\n', '。', '！', '？', '；', '.', '!', '?', ';']
-
-  boundaryTokens.forEach((token) => {
-    const tokenStart = sourceText.lastIndexOf(token, idealEnd - token.length)
-    const boundary = tokenStart === -1 ? -1 : tokenStart + token.length
-    if (boundary >= minimumEnd && boundary <= idealEnd) candidates.push(boundary)
-  })
-
-  return candidates.length ? Math.max(...candidates) : null
-}
-
-function lastHeadingBoundary(
-  sourceText: string,
-  start: number,
-  idealEnd: number,
-  minimumEnd: number,
-): number | null {
-  const section = sourceText.slice(start, idealEnd)
-  const headingPattern = /^(?:#{1,6}\s+|第[一二三四五六七八九十百]+[章节篇部分]|\d+(?:\.\d+)*[、.\s])/gm
-  let boundary: number | null = null
-  let match: RegExpExecArray | null
-
-  while ((match = headingPattern.exec(section))) {
-    const absoluteStart = start + match.index
-    if (absoluteStart >= minimumEnd) boundary = absoluteStart
-  }
-
-  return boundary
-}
-
-function resolveChunkEnd(
-  sourceText: string,
-  start: number,
-  idealEnd: number,
-  minimumEnd: number,
-  options: UnstructuredProcessOptions | undefined,
-): number {
-  const method = options?.chunkMethod ?? 'semantic'
-
-  if (method === 'fixed') return idealEnd
-
-  if (method === 'custom') {
-    const delimiter = normalizeDelimiter(options?.customDelimiter)
-    if (!delimiter) return idealEnd
-
-    const delimiterStart = sourceText.lastIndexOf(delimiter, idealEnd - delimiter.length)
-    const boundary = delimiterStart === -1 ? -1 : delimiterStart + delimiter.length
-    return boundary >= minimumEnd ? boundary : idealEnd
-  }
-
-  if (method === 'heading') {
-    const headingBoundary = lastHeadingBoundary(sourceText, start, idealEnd, minimumEnd)
-    if (headingBoundary !== null) return headingBoundary
-  }
-
-  return lastBoundaryInRange(sourceText, idealEnd, minimumEnd) ?? idealEnd
-}
-
-function buildUnstructuredRanges(
-  sourceText: string,
-  options?: UnstructuredProcessOptions,
-): SourceRange[] {
-  // 预览统一沿用“约 2 个字符 = 1 token”的轻量估算，避免引入分词器依赖。
-  const targetCharacters = finiteInteger(options?.chunkSize, DEFAULT_CHUNK_SIZE, 1) * 2
-  const minimumCharacters = Math.min(
-    targetCharacters,
-    finiteInteger(options?.minChunkSize, DEFAULT_MIN_CHUNK_SIZE, 1) * 2,
-  )
-  const requestedOverlap = finiteInteger(options?.chunkOverlap, DEFAULT_CHUNK_OVERLAP, 0) * 2
-  const protectedRanges = protectedRangesForOptions(sourceText, options)
-  const ranges: SourceRange[] = []
-  let cursor = 0
-
-  while (cursor < sourceText.length) {
-    const start = normalizeChunkStart(sourceText, cursor, protectedRanges)
-    if (start >= sourceText.length) break
-
-    const idealEnd = Math.min(sourceText.length, start + targetCharacters)
-    const minimumEnd = Math.min(idealEnd, start + minimumCharacters)
-    let end = idealEnd === sourceText.length
-      ? idealEnd
-      : resolveChunkEnd(sourceText, start, idealEnd, minimumEnd, options)
-    end = protectChunkEnd(end, start, minimumEnd, protectedRanges)
-
-    // 所有自定义边界都必须向前推进；异常配置回退到固定长度切分。
-    if (end <= start) end = Math.min(sourceText.length, start + targetCharacters)
-
-    let range = trimSourceRange(sourceText, start, end)
-    range = restoreProtectedEdges(range, start, end, protectedRanges)
-    if (end < sourceText.length && range.end - range.start < minimumCharacters) {
-      range.end = Math.min(end, range.start + minimumCharacters)
-    }
-    if (range.end <= range.start) {
-      cursor = Math.max(cursor + 1, end)
-      continue
-    }
-
-    const isLastRange = end >= sourceText.length
-    if (isLastRange && range.end - range.start < minimumCharacters && ranges.length) {
-      ranges[ranges.length - 1].end = range.end
-      break
-    }
-
-    ranges.push(range)
-    if (isLastRange) break
-
-    // overlap 是允许的最大重叠量；按当前切片动态收缩，保证每轮至少推进最小切片长度。
-    const maximumOverlap = Math.max(0, range.end - range.start - minimumCharacters)
-    const actualOverlap = Math.min(requestedOverlap, maximumOverlap)
-    const nextCursor = range.end - actualOverlap
-    cursor = nextCursor > start ? nextCursor : range.end
-  }
-
-  return ranges
-}
-
-function lineNumberAtOffset(lines: SourceLine[], offset: number): number | null {
-  if (!lines.length) return null
-
-  let low = 0
-  let high = lines.length - 1
-  let result = 0
-
-  while (low <= high) {
-    const middle = Math.floor((low + high) / 2)
-    if (lines[middle].start <= offset) {
-      result = middle
-      low = middle + 1
-    } else {
-      high = middle - 1
-    }
-  }
-
-  return lines[result].number
-}
-
-function previewItemFromRange(
-  sourceText: string,
-  lines: SourceLine[],
-  range: SourceRange,
-  sourceFileId: string,
-  index: number,
-): PreviewItem {
-  const content = sourceText.slice(range.start, range.end)
-
-  return {
-    id: `preview-${sourceFileId}-${index + 1}`,
-    sourceFileId,
-    originalContent: content,
-    editedContent: content,
-    sourceStart: range.start,
-    sourceEnd: range.end,
-    sourceStartLine: lineNumberAtOffset(lines, range.start),
-    sourceEndLine: lineNumberAtOffset(lines, Math.max(range.start, range.end - 1)),
-    tokenCount: Math.max(1, Math.ceil(content.length / 2)),
-    status: 'original',
-  }
-}
-
-export function buildPreviewItems(
-  sourceText: string,
-  processType: ProcessType,
-  sourceFileId = 'default-source',
-  unstructuredOptions?: UnstructuredProcessOptions,
-): PreviewItem[] {
-  const lines = sourceLines(sourceText)
-
-  if (processType === 'unstructured') {
-    return buildUnstructuredRanges(sourceText, unstructuredOptions).map((range, index) => (
-      previewItemFromRange(sourceText, lines, range, sourceFileId, index)
-    ))
-  }
-
-  const meaningfulLines = lines.filter((line) => line.content.trim())
-  const groupSize = processType === 'structured' ? 1 : 3
-  const items: PreviewItem[] = []
-
-  for (let index = 0; index < meaningfulLines.length; index += groupSize) {
-    const group = meaningfulLines.slice(index, index + groupSize)
-    if (!group.length) continue
-
-    const sourceStart = group[0].start
-    const sourceEnd = group[group.length - 1].end
-    const content = sourceText.slice(sourceStart, sourceEnd)
-
-    items.push({
-      id: `preview-${sourceFileId}-${items.length + 1}`,
-      sourceFileId,
-      originalContent: content,
-      editedContent: content,
-      sourceStart,
-      sourceEnd,
-      sourceStartLine: group[0].number,
-      sourceEndLine: group[group.length - 1].number,
-      tokenCount: Math.max(1, Math.ceil(content.length / 2)),
-      status: 'original',
-    })
-  }
-
-  return items
-}
-
-const SEMANTIC_PREFIXES = [
-  '请结合实际情况，说明一下：',
-  '如果方便的话，请详细解答：',
-  '请用通俗易懂的方式说明：',
-  '请从实际应用角度说明：',
-  '请简洁、自然地说明：',
-]
-
-export function createResults(
-  items: PreviewItem[],
-  options?: StructuredProcessOptions | UnstructuredProcessOptions,
-): ResultItem[] {
-  const resultCount = options && 'qaPairsPerChunk' in options
-    ? Math.min(3, finiteInteger(options.qaPairsPerChunk, 1, 1))
-    : Math.min(5, finiteInteger(options?.qaPairsPerRow, 1, 1))
-
-  const generatedResults = items.flatMap((item, index) => {
-    if (options?.qualityFilterEnabled && options.filterLowQuality) {
-      if (item.status === 'invalid' || !item.editedContent.trim()) return []
-    }
-
-    const [firstLine = '', ...rest] = item.editedContent.split('\n')
-    const output = rest.join('\n').trim() || item.editedContent.trim()
-    const baseInstruction = firstLine.replace(/^问[:：]\s*/, '').trim() || `数据条目 ${index + 1}`
-
-    return Array.from({ length: resultCount }, (_, variantIndex) => {
-      const instruction = options?.semanticEnrichment
-        ? `${SEMANTIC_PREFIXES[variantIndex]}${baseInstruction}`
-        : variantIndex === 0
-          ? baseInstruction
-          : `${baseInstruction}（问法 ${variantIndex + 1}）`
-
-      return {
-        id: resultCount === 1 ? `result-${index + 1}` : `result-${index + 1}-${variantIndex + 1}`,
-        instruction,
-        input: '',
-        output,
-        originalInstruction: instruction,
-        originalInput: '',
-        originalOutput: output,
-        status: 'valid' as const,
-      }
-    })
-  })
-
-  if (!options?.qualityFilterEnabled) return generatedResults
-
-  return generatedResults.filter((result) => {
-    if (options.filterLowQuality && (!result.instruction.trim() || !result.output.trim())) return false
-    if (options.filterShortContent && result.output.trim().length < options.minOutputLength) return false
-    return true
-  })
+/** 与后端预览 token 估算规则一致，仅用于编辑中的即时计数。 */
+export function estimateTokenCount(text: string): number {
+  return text.match(/[\u3400-\u4dbf\u4e00-\u9fff]|[A-Za-z0-9_]+|[^\s]/gu)?.length ?? 0
 }

frontend/src/views/data-process/create/types.ts

@@ -1,7 +1,24 @@
+import type {
+  DataProcessOutputType,
+  DataProcessPreviewFileStatus,
+  DataProcessReasoningDetail,
+} from '@/types/dataProcess'
+
 export type ProcessType = 'structured' | 'unstructured' | 'external'
 
 export type StepId = 'create' | 'model' | 'upload' | 'preview' | 'generate' | 'results'
 
+export const QA_PAIRS_GENERATION_LIMITS = { min: 1, max: 50 } as const
+
+export function normalizeQaPairsGenerationCount(value: unknown, fallback = 1): number {
+  const parsed = Number(value)
+  const normalized = Number.isFinite(parsed) ? Math.trunc(parsed) : fallback
+  return Math.min(
+    QA_PAIRS_GENERATION_LIMITS.max,
+    Math.max(QA_PAIRS_GENERATION_LIMITS.min, normalized),
+  )
+}
+
 export type PreprocessOption =
   | 'clean_invalid'
   | 'detect_structure'
@@ -19,6 +36,8 @@ export interface DatasetSplitOptions {
 export interface GenerationControlOptions {
   generationModelId: string | number | ''
   generationPrompt: string
+  outputType: DataProcessOutputType
+  reasoningDetail: DataProcessReasoningDetail
   temperature: number
   maxTokens: number
   jsonMode: boolean
@@ -44,7 +63,7 @@ export type UnstructuredPreprocessOption =
   | 'desensitize'
   | 'preserve_context'
 
-export type ChunkMethod = 'semantic' | 'heading' | 'fixed' | 'custom'
+export type ChunkMethod = 'layout_hybrid' | 'semantic' | 'fixed'
 
 export interface UnstructuredProcessOptions extends GenerationControlOptions {
   preprocessOptions: UnstructuredPreprocessOption[]
@@ -52,7 +71,7 @@ export interface UnstructuredProcessOptions extends GenerationControlOptions {
   chunkSize: number
   chunkOverlap: number
   minChunkSize: number
-  customDelimiter: string
+  semanticBreakpointPercentile: number
   preserveTables: boolean
   preserveCodeBlocks: boolean
   preserveLists: boolean
@@ -62,21 +81,34 @@ export interface UnstructuredProcessOptions extends GenerationControlOptions {
 }
 
 export interface ExternalDataSource {
-  type: string
+  type: 'postgresql'
   url: string
-  authMode: string
+  authMode: 'none' | 'basic'
   username?: string
   password?: string
-  token?: string
   limit: number
+  query?: string
+  fileName?: string
 }
 
 export interface UploadedDataFile {
   uid: string | number
+  sourceFileId?: string
+  rawFile?: File
   name: string
   size: number
   count: number
   content: string
+  fileFormat?: string
+  checksumSha256?: string
+  status: 'queued' | 'uploading' | 'ready' | 'failed'
+  uploadProgress: number
+  uploadError?: string
+  previewStatus?: DataProcessPreviewFileStatus
+  previewProgress?: number
+  previewError?: string
+  previewCount?: number
+  previewConfigSignature?: string
 }
 
 export interface SourceLine {
@@ -91,12 +123,18 @@ export interface PreviewItem {
   sourceFileId: string
   originalContent: string
   editedContent: string
+  savedEditedContent: string
   sourceStart: number | null
   sourceEnd: number | null
   sourceStartLine: number | null
   sourceEndLine: number | null
+  sourcePages?: number[]
   tokenCount: number
   status: 'original' | 'modified' | 'manual' | 'invalid'
+  qualityScore?: number
+  qualityDetails?: Record<string, number>
+  piiStats?: Record<string, number>
+  updatedAt?: string
 }
 
 export interface GenerationState {
@@ -107,12 +145,31 @@ export interface GenerationState {
 
 export interface ResultItem {
   id: string
+  previewItemId: string | null
   instruction: string
   input: string
   output: string
   originalInstruction: string
   originalInput: string
   originalOutput: string
+  savedInstruction: string
+  savedInput: string
+  savedOutput: string
+  savedStatus: 'valid' | 'modified' | 'invalid'
   status: 'valid' | 'modified' | 'invalid'
   error?: string
+  split?: 'train' | 'validation' | 'test'
+  qualityScore?: number
+  qualityDetails?: Record<string, number>
+  updatedAt?: string
+}
+
+export interface BulkResultRegenerationState {
+  status: 'idle' | 'running' | 'completed' | 'partial' | 'failed'
+  total: number
+  completed: number
+  succeeded: number
+  failed: number
+  targetIds: string[]
+  failedIds: string[]
 }

frontend/src/views/data-process/create/useDataProcessGeneration.ts

@@ -1,35 +1,81 @@
-import { reactive, ref, type Ref } from 'vue'
+import { computed, reactive, ref, type Ref } from 'vue'
 import { ElMessage } from 'element-plus'
-import { createResults } from './previewModel'
-import type {
-  GenerationState,
-  PreviewItem,
-  ProcessType,
-  ResultItem,
-  StructuredProcessOptions,
-  UnstructuredProcessOptions,
-} from './types'
+import {
+  generateDataProcess,
+  getDataProcessProgress,
+  getDataProcessResults,
+  regenerateDataProcessResult,
+  regenerateDataProcessResults,
+  restoreDataProcessResult,
+  updateDataProcessResult,
+  type DataProcessProgress,
+  type DataProcessResult,
+} from '@/api/modules/dataProcess'
+import type { BulkResultRegenerationState, GenerationState, ResultItem } from './types'
 
 interface GenerationBindings {
-  previewItems: Ref<PreviewItem[]>
-  processType: Ref<ProcessType>
-  structuredOptions: Ref<StructuredProcessOptions>
-  unstructuredOptions: Ref<UnstructuredProcessOptions>
+  taskId: Ref<string | null>
   dirty: Ref<boolean>
+  beforeGenerate?: () => Promise<boolean | void>
+}
+
+const RESULT_PAGE_SIZE = 500
+const POLL_INTERVAL_MS = 1500
+// 4 个后端 worker 可连续消费三轮，减少每 4 条等待最慢项造成的空闲；
+// 单条重生成最长 60 秒，因此 12 条仍处于批量接口 240 秒超时预算内。
+const BULK_REGENERATION_CHUNK_SIZE = 12
+
+function mapResult(item: DataProcessResult): ResultItem {
+  return {
+    id: String(item.id),
+    previewItemId: item.preview_item_id == null ? null : String(item.preview_item_id),
+    instruction: item.instruction,
+    input: item.input || '',
+    output: item.output,
+    originalInstruction: item.original_instruction ?? item.instruction,
+    originalInput: item.original_input ?? item.input ?? '',
+    originalOutput: item.original_output ?? item.output,
+    savedInstruction: item.instruction,
+    savedInput: item.input || '',
+    savedOutput: item.output,
+    savedStatus: item.status,
+    status: item.status,
+    error: item.error || undefined,
+    split: item.split || undefined,
+    qualityScore: item.quality_score?.overall,
+    updatedAt: item.updated_at,
+  }
 }
 
 export function useDataProcessGeneration(bindings: GenerationBindings) {
   const results = ref<ResultItem[]>([])
   const selectedResultId = ref<string | null>(null)
+  const regeneratingResultId = ref<string | null>(null)
+  const bulkRegeneration = reactive<BulkResultRegenerationState>({
+    status: 'idle',
+    total: 0,
+    completed: 0,
+    succeeded: 0,
+    failed: 0,
+    targetIds: [],
+    failedIds: [],
+  })
+  const resultRegenerationBusy = computed(() => (
+    Boolean(regeneratingResultId.value) || bulkRegeneration.status === 'running'
+  ))
   const generation = reactive<GenerationState>({
     status: 'idle',
     progress: 0,
     message: '确认摘要后即可开始生成，过程中可查看实时进度。',
   })
-  let generationTimer: ReturnType<typeof setInterval> | null = null
+  let generationTimer: ReturnType<typeof setTimeout> | null = null
+  let generationRun = 0
+  let pollFailureCount = 0
+  let generationStarting = false
 
   function stopGenerationTimer() {
-    if (generationTimer) clearInterval(generationTimer)
+    generationRun += 1
+    if (generationTimer) clearTimeout(generationTimer)
     generationTimer = null
   }
 
@@ -40,39 +86,157 @@ export function useDataProcessGeneration(bindings: GenerationBindings) {
     generation.message = '确认摘要后即可开始生成，过程中可查看实时进度。'
     results.value = []
     selectedResultId.value = null
+    regeneratingResultId.value = null
+    Object.assign(bulkRegeneration, {
+      status: 'idle',
+      total: 0,
+      completed: 0,
+      succeeded: 0,
+      failed: 0,
+      targetIds: [],
+      failedIds: [],
+    })
   }
 
-  function startGeneration() {
-    stopGenerationTimer()
-    generation.status = 'running'
-    generation.progress = 0
-    generation.message = '正在应用预览修改并生成标准化结果，请稍候。'
+  function applyProgress(progress: DataProcessProgress) {
+    generation.progress = Math.max(0, Math.min(100, Number(progress.progress) || 0))
+    generation.message = progress.message || (
+      progress.status === 'running'
+        ? '后端正在生成标准化结果并进行质量评分。'
+        : progress.status === 'completed'
+          ? '数据处理已完成。'
+          : progress.failure_reason || '任务已停止。'
+    )
+  }
 
-    generationTimer = setInterval(() => {
-      generation.progress = Math.min(100, generation.progress + 8)
-      if (generation.progress < 100) return
+  async function loadAllResults(taskId: string) {
+    const first = await getDataProcessResults(taskId, { page: 1, page_size: RESULT_PAGE_SIZE })
+    const items = [...first.items]
+    const pages = Math.ceil(first.total / first.page_size)
+    for (let page = 2; page <= pages; page += 1) {
+      const next = await getDataProcessResults(taskId, { page, page_size: RESULT_PAGE_SIZE })
+      items.push(...next.items)
+    }
+    results.value = items.map(mapResult)
+    selectedResultId.value = results.value[0]?.id ?? null
+  }
 
-      stopGenerationTimer()
+  async function finishFromProgress(progress: DataProcessProgress, notify = true) {
+    pollFailureCount = 0
+    applyProgress(progress)
+    if (progress.status === 'completed') {
+      const taskId = bindings.taskId.value
+      if (!taskId) return
+      await loadAllResults(taskId)
       generation.status = 'success'
-      results.value = createResults(
-        bindings.previewItems.value,
-        bindings.processType.value === 'structured'
-          ? bindings.structuredOptions.value
-          : bindings.processType.value === 'unstructured'
-            ? bindings.unstructuredOptions.value
-            : undefined,
-      )
+      generation.progress = 100
       generation.message = `已完成 ${results.value.length.toLocaleString()} 条数据处理，可进入结果页检查。`
-      selectedResultId.value = results.value[0]?.id ?? null
-      bindings.dirty.value = true
-      ElMessage.success('数据处理完成')
-    }, 180)
+      if (notify) ElMessage.success('数据处理完成')
+      return
+    }
+    if (progress.status === 'failed' || progress.status === 'stopped') {
+      generation.status = 'failed'
+      generation.message = progress.failure_reason || progress.message || (
+        progress.status === 'stopped' ? '任务已停止，可以重新生成。' : '数据处理失败，请检查配置后重试。'
+      )
+    }
   }
 
-  function stopGeneration() {
+  async function pollGeneration(runId: number) {
+    const taskId = bindings.taskId.value
+    if (!taskId || runId !== generationRun || generation.status !== 'running') return
+
+    try {
+      const progress = await getDataProcessProgress(taskId)
+      if (runId !== generationRun) return
+      pollFailureCount = 0
+      if (progress.status === 'running' || progress.status === 'pending') {
+        applyProgress(progress)
+        generationTimer = setTimeout(() => void pollGeneration(runId), POLL_INTERVAL_MS)
+        return
+      }
+      await finishFromProgress(progress)
+    } catch (error) {
+      if (runId !== generationRun) return
+      pollFailureCount += 1
+      if (pollFailureCount <= 3) {
+        generation.message = `进度查询暂时失败，正在重试（${pollFailureCount}/3）…`
+        generationTimer = setTimeout(() => void pollGeneration(runId), POLL_INTERVAL_MS)
+        return
+      }
+      generation.status = 'failed'
+      generation.message = error instanceof Error ? error.message : '查询任务进度失败，请重试。'
+    }
+  }
+
+  async function startGeneration() {
+    if (generationStarting || generation.status === 'running') return false
+    const taskId = bindings.taskId.value
+    if (!taskId) {
+      ElMessage.error('任务尚未创建，请返回上一步重试')
+      return false
+    }
+
+    generationStarting = true
+    let runId: number | null = null
+    try {
+      const canStart = await bindings.beforeGenerate?.()
+      if (canStart === false) return false
+      stopGenerationTimer()
+      const activeRunId = generationRun
+      runId = activeRunId
+      generation.status = 'running'
+      pollFailureCount = 0
+      generation.progress = 0
+      generation.message = '正在同步预览修改并启动后端处理，请稍候。'
+      const progress = await generateDataProcess(taskId)
+      if (activeRunId !== generationRun) return false
+      if (progress.status === 'completed' || progress.status === 'failed' || progress.status === 'stopped') {
+        await finishFromProgress(progress)
+        return true
+      }
+      applyProgress(progress)
+      generationTimer = setTimeout(() => void pollGeneration(activeRunId), POLL_INTERVAL_MS)
+      return true
+    } catch (error) {
+      if (runId !== null && runId !== generationRun) return
+      generation.status = 'failed'
+      generation.message = error instanceof Error ? error.message : '启动数据处理失败，请重试。'
+      return false
+    } finally {
+      generationStarting = false
+    }
+  }
+
+  async function resumeGeneration() {
+    const taskId = bindings.taskId.value
+    if (!taskId) return
     stopGenerationTimer()
-    generation.status = 'failed'
-    generation.message = '任务已停止，预览修改仍然保留，可以重新生成。'
+    const activeRunId = generationRun
+    pollFailureCount = 0
+    try {
+      const progress = await getDataProcessProgress(taskId)
+      if (activeRunId !== generationRun) return
+      if (progress.status === 'running') {
+        generation.status = 'running'
+        applyProgress(progress)
+        generationTimer = setTimeout(() => void pollGeneration(activeRunId), POLL_INTERVAL_MS)
+        return
+      }
+      if (progress.status === 'completed') {
+        await finishFromProgress(progress, false)
+        return
+      }
+      if (progress.status === 'failed' || progress.status === 'stopped') {
+        await finishFromProgress(progress, false)
+        return
+      }
+      generation.status = 'idle'
+      applyProgress(progress)
+    } catch (error) {
+      generation.status = 'failed'
+      generation.message = error instanceof Error ? error.message : '查询任务进度失败，请重试。'
+    }
   }
 
   function updateResultField(id: string, field: 'instruction' | 'input' | 'output', value: string) {
@@ -88,22 +252,174 @@ export function useDataProcessGeneration(bindings: GenerationBindings) {
     bindings.dirty.value = true
   }
 
-  function restoreResult(id: string) {
+
+
+  async function regenerateResult(id: string) {
+    const taskId = bindings.taskId.value
     const item = results.value.find((entry) => entry.id === id)
-    if (!item) return
-    item.instruction = item.originalInstruction
-    item.input = item.originalInput
-    item.output = item.originalOutput
-    item.error = undefined
-    item.status = 'valid'
-    bindings.dirty.value = true
+    if (!taskId || !item) return false
+    if (resultRegenerationBusy.value) {
+      ElMessage.warning('请等待当前失败结果重新生成完成')
+      return false
+    }
+    if (item.savedStatus !== 'invalid') {
+      ElMessage.warning('只有生成失败的结果可以重新生成')
+      return false
+    }
+    if (!item.updatedAt) {
+      ElMessage.error('结果版本信息缺失，请刷新页面后重试')
+      return false
+    }
+
+    regeneratingResultId.value = id
+    try {
+      const regenerated = await regenerateDataProcessResult(taskId, id, {
+        expected_updated_at: item.updatedAt,
+      })
+      const index = results.value.findIndex((entry) => entry.id === id)
+      if (index >= 0) results.value[index] = mapResult(regenerated)
+      ElMessage.success('当前结果已重新生成')
+      return true
+    } catch {
+      return false
+    } finally {
+      regeneratingResultId.value = null
+    }
+  }
+
+  async function regenerateAllResults() {
+    const taskId = bindings.taskId.value
+    if (!taskId) return false
+    if (resultRegenerationBusy.value) {
+      ElMessage.warning('请等待当前失败结果重新生成完成')
+      return false
+    }
+
+    const candidates = results.value.filter((item) => item.savedStatus === 'invalid')
+    if (!candidates.length) {
+      ElMessage.info('当前没有需要重新生成的失败结果')
+      return false
+    }
+    const unsaved = candidates.find((item) => (
+      item.instruction !== item.savedInstruction
+      || item.input !== item.savedInput
+      || item.output !== item.savedOutput
+    ))
+    if (unsaved) {
+      selectedResultId.value = unsaved.id
+      ElMessage.warning('失败结果存在未保存修改，请先保存或恢复后再全部重新生成')
+      return false
+    }
+    const missingVersion = candidates.find((item) => !item.updatedAt)
+    if (missingVersion) {
+      selectedResultId.value = missingVersion.id
+      ElMessage.error('失败结果版本信息缺失，请刷新页面后重试')
+      return false
+    }
+
+    Object.assign(bulkRegeneration, {
+      status: 'running',
+      total: candidates.length,
+      completed: 0,
+      succeeded: 0,
+      failed: 0,
+      targetIds: candidates.map((item) => item.id),
+      failedIds: [],
+    })
+
+    try {
+      for (let offset = 0; offset < candidates.length; offset += BULK_REGENERATION_CHUNK_SIZE) {
+        const chunk = candidates.slice(offset, offset + BULK_REGENERATION_CHUNK_SIZE)
+        try {
+          const regenerated = await regenerateDataProcessResults(taskId, {
+            items: chunk.map((item) => ({
+              result_id: item.id,
+              expected_updated_at: item.updatedAt as string,
+            })),
+          })
+          for (const item of regenerated.items) {
+            const index = results.value.findIndex((entry) => entry.id === String(item.id))
+            if (index >= 0) results.value[index] = mapResult(item)
+          }
+          bulkRegeneration.completed += regenerated.total
+          bulkRegeneration.succeeded += regenerated.succeeded
+          bulkRegeneration.failed += regenerated.failed
+          bulkRegeneration.failedIds.push(
+            ...regenerated.failures.map((failure) => failure.result_id),
+          )
+        } catch {
+          const remaining = candidates.slice(offset)
+          bulkRegeneration.completed = bulkRegeneration.total
+          bulkRegeneration.failed += remaining.length
+          bulkRegeneration.failedIds.push(...remaining.map((item) => item.id))
+          break
+        }
+      }
+
+      if (bulkRegeneration.failed === 0) {
+        bulkRegeneration.status = 'completed'
+        ElMessage.success(`已重新生成 ${bulkRegeneration.succeeded} 条失败结果`)
+      } else if (bulkRegeneration.succeeded > 0) {
+        bulkRegeneration.status = 'partial'
+        ElMessage.warning(
+          `重新生成完成：成功 ${bulkRegeneration.succeeded} 条，仍失败 ${bulkRegeneration.failed} 条`,
+        )
+      } else {
+        bulkRegeneration.status = 'failed'
+        ElMessage.error(`重新生成失败：${bulkRegeneration.failed} 条结果仍需重试`)
+      }
+      const firstFailed = bulkRegeneration.failedIds.find((id) => (
+        results.value.find((item) => item.id === id)?.savedStatus === 'invalid'
+      ))
+      if (firstFailed) selectedResultId.value = firstFailed
+      return bulkRegeneration.failed === 0
+    } catch {
+      const remaining = Math.max(0, bulkRegeneration.total - bulkRegeneration.completed)
+      const remainingIds = candidates
+        .filter((candidate) => (
+          results.value.find((item) => item.id === candidate.id)?.savedStatus === 'invalid'
+        ))
+        .map((item) => item.id)
+      bulkRegeneration.completed = bulkRegeneration.total
+      bulkRegeneration.failed += remaining
+      bulkRegeneration.failedIds = [...new Set([
+        ...bulkRegeneration.failedIds,
+        ...remainingIds,
+      ])]
+      bulkRegeneration.status = 'failed'
+      if (bulkRegeneration.failedIds[0]) selectedResultId.value = bulkRegeneration.failedIds[0]
+      ElMessage.error('批量重新生成意外中断，尚未成功的结果保持原状')
+      return false
+    } finally {
+      bulkRegeneration.targetIds = []
+    }
+  }
+
+  async function persistResultChanges() {
+    const taskId = bindings.taskId.value
+    if (!taskId) throw new Error('任务尚未创建')
+    const changed = results.value.filter((item) => (
+      item.instruction !== item.savedInstruction
+      || item.input !== item.savedInput
+      || item.output !== item.savedOutput
+    ))
+    for (const item of changed) {
+      const saved = await updateDataProcessResult(taskId, item.id, {
+        instruction: item.instruction,
+        input: item.input,
+        output: item.output,
+        expected_updated_at: item.updatedAt,
+      })
+      const index = results.value.findIndex((entry) => entry.id === item.id)
+      if (index >= 0) results.value[index] = mapResult(saved)
+    }
   }
 
   function validateResults() {
     let firstInvalidId: string | null = null
     for (const item of results.value) {
-      if (!item.instruction.trim() || !item.output.trim()) {
-        item.error = 'Instruction 和 Output 不能为空'
+      if (!item.instruction.trim() || !item.output.trim() || item.status === 'invalid') {
+        item.error ||= '结果未通过后端质量校验，请修改后重新保存'
         item.status = 'invalid'
         firstInvalidId ??= item.id
       }
@@ -113,13 +429,18 @@ export function useDataProcessGeneration(bindings: GenerationBindings) {
   }
 
   return {
+    bulkRegeneration,
     generation,
+    regeneratingResultId,
+    resultRegenerationBusy,
     results,
     selectedResultId,
+    persistResultChanges,
     resetDownstream,
-    restoreResult,
+    regenerateAllResults,
+    regenerateResult,
+    resumeGeneration,
     startGeneration,
-    stopGeneration,
     stopGenerationTimer,
     updateResultField,
     validateResults,

frontend/src/views/data-process/create/useDataProcessPreviewBuild.ts

@@ -0,0 +1,139 @@
+import { onBeforeUnmount } from 'vue'
+import {
+  getDataProcessPreviewProgress,
+  startDataProcessPreview,
+} from '@/api/modules/dataProcess'
+import type {
+  DataProcessPreviewFileBuildProgress,
+  DataProcessPreviewProgress,
+} from '@/types/dataProcess'
+
+const POLL_INTERVAL_MS = 1500
+const MAX_CONSECUTIVE_FAILURES = 3
+
+type PreviewProgressCallback = (progress: DataProcessPreviewProgress) => void
+
+function isActive(progress: DataProcessPreviewProgress) {
+  return progress.preview_status === 'queued' || progress.preview_status === 'running'
+}
+
+function waitForNextPoll() {
+  return new Promise<void>((resolve) => setTimeout(resolve, POLL_INTERVAL_MS))
+}
+
+/**
+ * 后台切分控制器。
+ *
+ * startPreviewBuild 只发起一次 POST，随后查询任务级进度；resumePreviewBuild
+ * 不重复启动任务，仅接管已有切分。两者都在 completed/failed 时返回最终进度。
+ * 响应不依赖逐文件明细，完成后由调用方重新读取预览及源文件统计。
+ */
+export function useDataProcessPreviewBuild() {
+  let activePollRun = 0
+
+  function stopPreviewPolling() {
+    activePollRun += 1
+  }
+
+  async function pollUntilSettled(
+    taskId: string | number,
+    initialProgress: DataProcessPreviewProgress,
+    pollRun: number,
+    onProgress?: PreviewProgressCallback,
+  ) {
+    let progress = initialProgress
+    let consecutiveFailures = 0
+    if (pollRun !== activePollRun) return progress
+    onProgress?.(progress)
+
+    while (isActive(progress) && pollRun === activePollRun) {
+      await waitForNextPoll()
+      if (pollRun !== activePollRun) break
+      try {
+        progress = await getDataProcessPreviewProgress(taskId)
+        consecutiveFailures = 0
+        onProgress?.(progress)
+      } catch (error) {
+        consecutiveFailures += 1
+        if (consecutiveFailures >= MAX_CONSECUTIVE_FAILURES) throw error
+      }
+    }
+    return progress
+  }
+
+  async function startPreviewBuild(
+    taskId: string | number,
+    sourceFileIds: Array<string | number>,
+    onProgress?: PreviewProgressCallback,
+  ): Promise<DataProcessPreviewProgress> {
+    stopPreviewPolling()
+    const pollRun = activePollRun
+    const progress = await startDataProcessPreview(taskId, {
+      replace_existing: true,
+      source_file_ids: sourceFileIds,
+    })
+    return pollUntilSettled(taskId, progress, pollRun, onProgress)
+  }
+
+  async function resumePreviewBuild(
+    taskId: string | number,
+    onProgress?: PreviewProgressCallback,
+  ): Promise<DataProcessPreviewProgress> {
+    stopPreviewPolling()
+    const pollRun = activePollRun
+    const progress = await getDataProcessPreviewProgress(taskId)
+    return pollUntilSettled(taskId, progress, pollRun, onProgress)
+  }
+
+  /**
+   * 兼容现有上传步骤的逐文件展示契约；底层仍只启动一个后台切分任务。
+   * 后端不返回 files[]，因此运行中按任务总进度同步展示，完成后调用方
+   * 必须重新加载预览，不能把这里的可视状态当作文件统计来源。
+   */
+  async function buildPreviewsByFile(
+    taskId: string | number,
+    sourceFileIds: Array<string | number>,
+    onProgress: (progress: DataProcessPreviewFileBuildProgress) => void,
+  ): Promise<DataProcessPreviewFileBuildProgress[]> {
+    const emitTaskProgress = (taskProgress: DataProcessPreviewProgress) => {
+      for (const sourceFileId of sourceFileIds) {
+        if (taskProgress.preview_status === 'completed') {
+          onProgress({ source_file_id: sourceFileId, status: 'success', progress: 100 })
+        } else if (taskProgress.preview_status === 'failed' || taskProgress.preview_status === 'cancelled') {
+          onProgress({
+            source_file_id: sourceFileId,
+            status: 'failed',
+            progress: 0,
+            error: taskProgress.preview_failure_reason || '切分失败，请重试',
+          })
+        } else {
+          onProgress({
+            source_file_id: sourceFileId,
+            status: 'processing',
+            progress: taskProgress.preview_progress,
+          })
+        }
+      }
+    }
+    const finalProgress = await startPreviewBuild(taskId, sourceFileIds, emitTaskProgress)
+    return sourceFileIds.map((sourceFileId) => (
+      finalProgress.preview_status === 'completed'
+        ? { source_file_id: sourceFileId, status: 'success', progress: 100 }
+        : {
+            source_file_id: sourceFileId,
+            status: 'failed',
+            progress: 0,
+            error: finalProgress.preview_failure_reason || '切分失败，请重试',
+          }
+    ))
+  }
+
+  onBeforeUnmount(stopPreviewPolling)
+
+  return {
+    startPreviewBuild,
+    resumePreviewBuild,
+    buildPreviewsByFile,
+    stopPreviewPolling,
+  }
+}

frontend/src/views/data-process/create/useDataProcessRegeneration.ts

@@ -0,0 +1,254 @@
+import { computed, nextTick, ref, type Reactive, type Ref } from 'vue'
+import { useRoute } from 'vue-router'
+import {
+  getDataProcessPreview,
+  getDataProcessSourceContent,
+  getDataProcessTask,
+  regenerateDataProcessTask,
+} from '@/api/modules/dataProcess'
+import type {
+  DataProcessPreviewItem,
+  DataProcessRegeneratePayload,
+  DataProcessTask,
+} from '@/types/dataProcess'
+import {
+  createStructuredOptionsFromConfig,
+  createUnstructuredOptionsFromConfig,
+} from './dataProcessCreateState'
+import { mapDataProcessSourceFile } from './useDataProcessSourceUpload'
+import type {
+  PreviewItem,
+  ProcessType,
+  StructuredProcessOptions,
+  UnstructuredProcessOptions,
+  UploadedDataFile,
+} from './types'
+
+interface ConfirmOptions {
+  title: string
+  message: string
+  confirmText: string
+  cancelText: string
+  tone: 'warning'
+}
+
+interface RegenerationBindings {
+  task: Reactive<{ name: string; description: string }>
+  processType: Ref<ProcessType>
+  structuredOptions: Ref<StructuredProcessOptions>
+  unstructuredOptions: Ref<UnstructuredProcessOptions>
+  uploadedFiles: Ref<UploadedDataFile[]>
+  previewItems: Ref<PreviewItem[]>
+  selectedPreviewFileId: Ref<string | null>
+  selectedPreviewId: Ref<string | null>
+  selectedPreviewIdsByFile: Ref<Record<string, string>>
+  previewSignature: Ref<string>
+  dirty: Ref<boolean>
+  buildPreviewConfigSignature: () => string
+  buildPreviewSignature: () => string
+  mapPreviewItem: (item: DataProcessPreviewItem) => PreviewItem
+  resetDownstream: () => void
+}
+
+async function loadSourceContent(taskId: string, fileId: string | number) {
+  const chunks: string[] = []
+  let startLine = 1
+  while (true) {
+    const source = await getDataProcessSourceContent(taskId, fileId, {
+      start_line: startLine,
+      line_count: 10_000,
+    })
+    chunks.push(source.content || '')
+    if (!source.has_more) break
+    const nextLine = Number(source.end_line || startLine) + 1
+    if (nextLine <= startLine) break
+    startLine = nextLine
+  }
+  // source_content_lines 已保留原始换行；分页之间直接拼接，避免凭空增加空行并破坏偏移。
+  return chunks.join('')
+}
+
+async function loadAllPreviews(taskId: string, mapPreviewItem: RegenerationBindings['mapPreviewItem']) {
+  const first = await getDataProcessPreview(taskId, { page: 1, page_size: 500 })
+  const items = [...first.items]
+  const pages = Math.ceil(first.total / first.page_size)
+  for (let page = 2; page <= pages; page += 1) {
+    const next = await getDataProcessPreview(taskId, { page, page_size: 500 })
+    items.push(...next.items)
+  }
+  return items.map(mapPreviewItem)
+}
+
+export function useDataProcessRegeneration(bindings: RegenerationBindings) {
+  const route = useRoute()
+  const isRegeneration = computed(() => route.name === 'data-process-regenerate')
+  const isWorkflowResume = computed(() => route.name === 'data-process-workflow')
+  const sourceTaskId = computed(() => (
+    isRegeneration.value || isWorkflowResume.value ? String(route.params.id || '') : ''
+  ))
+  const originalProcessType = ref<ProcessType | null>(null)
+  const originalTaskUpdatedAt = ref('')
+  const regenerationPrepared = ref(false)
+  const originalPreviewConfigSignature = ref('')
+  const confirmedPreviewConfigSignature = ref('')
+  const hydrating = ref(false)
+  const initializationError = ref('')
+
+  async function hydrateWorkspace(task: DataProcessTask, preservePreviews: boolean) {
+    const taskId = String(task.id)
+    bindings.uploadedFiles.value = await Promise.all((task.source_files || []).map(async (file) => (
+      mapDataProcessSourceFile(file, await loadSourceContent(taskId, file.id))
+    )))
+    bindings.previewItems.value = preservePreviews
+      ? await loadAllPreviews(taskId, bindings.mapPreviewItem)
+      : []
+
+    const configSignature = bindings.buildPreviewConfigSignature()
+    const previewCounts = new Map<string, number>()
+    for (const item of bindings.previewItems.value) {
+      previewCounts.set(item.sourceFileId, (previewCounts.get(item.sourceFileId) || 0) + 1)
+    }
+    for (const file of bindings.uploadedFiles.value) {
+      const count = previewCounts.get(String(file.sourceFileId)) || 0
+      file.previewCount = count
+      file.previewStatus = preservePreviews && count > 0 ? 'success' : 'waiting'
+      file.previewProgress = preservePreviews && count > 0 ? 100 : 0
+      file.previewConfigSignature = preservePreviews && count > 0 ? configSignature : undefined
+    }
+
+    bindings.selectedPreviewFileId.value = String(bindings.uploadedFiles.value[0]?.uid ?? '') || null
+    bindings.selectedPreviewId.value = bindings.selectedPreviewFileId.value
+      ? bindings.previewItems.value.find((item) => (
+          item.sourceFileId === bindings.selectedPreviewFileId.value
+        ))?.id ?? null
+      : null
+    bindings.selectedPreviewIdsByFile.value = (
+      bindings.selectedPreviewFileId.value && bindings.selectedPreviewId.value
+    ) ? { [bindings.selectedPreviewFileId.value]: bindings.selectedPreviewId.value } : {}
+    bindings.previewSignature.value = preservePreviews && bindings.previewItems.value.length
+      ? bindings.buildPreviewSignature()
+      : ''
+    bindings.resetDownstream()
+  }
+
+  async function loadSource() {
+    if (!sourceTaskId.value) return
+    hydrating.value = true
+    initializationError.value = ''
+    try {
+      const sourceTask = await getDataProcessTask(sourceTaskId.value)
+      const sourceType = sourceTask.process_type as ProcessType
+      originalProcessType.value = isRegeneration.value ? sourceType : null
+      originalTaskUpdatedAt.value = sourceTask.updated_at
+      bindings.task.name = sourceTask.name
+      bindings.task.description = sourceTask.description || ''
+      bindings.processType.value = sourceType
+      const config = sourceTask.config || {}
+      bindings.structuredOptions.value = createStructuredOptionsFromConfig(config)
+      bindings.unstructuredOptions.value = createUnstructuredOptionsFromConfig(config)
+      originalPreviewConfigSignature.value = bindings.buildPreviewConfigSignature()
+      confirmedPreviewConfigSignature.value = ''
+      await hydrateWorkspace(sourceTask, true)
+      await nextTick()
+      bindings.dirty.value = false
+      return sourceTask
+    } catch (error) {
+      initializationError.value = error instanceof Error
+        ? error.message
+        : '原数据处理任务加载失败，请返回详情页后重试'
+    } finally {
+      hydrating.value = false
+    }
+  }
+
+  async function confirmPreviewConfigChange(
+    openConfirm: (options: ConfirmOptions) => Promise<boolean | undefined>,
+  ) {
+    if (!isRegeneration.value) return true
+    if (!bindings.previewItems.value.length) return true
+    const currentSignature = bindings.buildPreviewConfigSignature()
+    const changed = currentSignature !== originalPreviewConfigSignature.value
+    if (!changed || currentSignature === confirmedPreviewConfigSignature.value) return true
+    const confirmed = await openConfirm({
+      title: '确认修改切分配置？',
+      message: '修改预处理或切分配置后，将按新配置重新切分。在点击“开始生成”前，原生成结果和已发布数据会继续保留。',
+      confirmText: '确认并继续',
+      cancelText: '返回检查',
+      tone: 'warning',
+    })
+    if (confirmed) confirmedPreviewConfigSignature.value = currentSignature
+    return Boolean(confirmed)
+  }
+
+  async function prepareRegeneration(
+    payload: Omit<DataProcessRegeneratePayload, 'expected_updated_at'>,
+  ) {
+    // 第一次 prepare 后，上传、切分、编辑预览和生成都会推进任务版本。
+    // 再次提交前读取同一任务的最新版本，仍由后端事务处理读取后的并发竞争。
+    if (regenerationPrepared.value) {
+      const latestTask = await getDataProcessTask(sourceTaskId.value)
+      originalTaskUpdatedAt.value = latestTask.updated_at || originalTaskUpdatedAt.value
+    }
+    const regenerated = await regenerateDataProcessTask(sourceTaskId.value, {
+      ...payload,
+      expected_updated_at: originalTaskUpdatedAt.value,
+    })
+    originalTaskUpdatedAt.value = regenerated.task.updated_at || originalTaskUpdatedAt.value
+    originalPreviewConfigSignature.value = bindings.buildPreviewConfigSignature()
+    confirmedPreviewConfigSignature.value = ''
+    regenerationPrepared.value = true
+    bindings.dirty.value = true
+    try {
+      const regeneratedTask = regenerated.task.source_files
+        ? regenerated.task
+        : await getDataProcessTask(regenerated.task.id)
+      originalTaskUpdatedAt.value = regeneratedTask.updated_at || originalTaskUpdatedAt.value
+      hydrating.value = true
+      try {
+        await hydrateWorkspace(regeneratedTask, !regenerated.preview_invalidated)
+      } finally {
+        hydrating.value = false
+      }
+    } catch (error) {
+      initializationError.value = error instanceof Error
+        ? `重新生成配置已保存，但工作区恢复失败：${error.message}`
+        : '重新生成配置已保存，但工作区恢复失败，请重试加载原任务'
+      throw error
+    }
+    return regenerated
+  }
+
+  async function confirmStartGeneration(
+    openConfirm: (options: ConfirmOptions) => Promise<boolean | undefined>,
+    syncPreviewChanges: () => Promise<void>,
+  ) {
+    if (isRegeneration.value) {
+      const confirmed = await openConfirm({
+        title: '开始重新生成？',
+        message: '点击开始后，当前生成结果将被替换。已发布数据集会继续保留，直到重新发布。',
+        confirmText: '开始生成',
+        cancelText: '继续检查',
+        tone: 'warning',
+      })
+      if (!confirmed) return false
+    }
+    await syncPreviewChanges()
+    return true
+  }
+
+  return {
+    sourceTaskId,
+    isRegeneration,
+    isWorkflowResume,
+    originalProcessType,
+    originalTaskUpdatedAt,
+    regenerationPrepared,
+    hydrating,
+    initializationError,
+    hydrateWorkspace,
+    loadSource,
+    confirmPreviewConfigChange,
+    prepareRegeneration,
+    confirmStartGeneration,
+  }
+}

frontend/src/views/data-process/create/useDataProcessSourceUpload.ts

@@ -0,0 +1,198 @@
+import { ref, type Ref } from 'vue'
+import {
+  getDataProcessSourceContent,
+  uploadDataProcessSourceFiles,
+  type DataProcessSourceFile,
+} from '@/api/modules/dataProcess'
+import type { ProcessType, UploadedDataFile } from './types'
+
+const BINARY_FILE_EXTENSIONS = new Set(['xlsx', 'pdf', 'docx', 'pptx'])
+const STRUCTURED_FILE_EXTENSIONS = new Set(['json', 'jsonl', 'ndjson', 'csv', 'tsv', 'xlsx'])
+const UNSTRUCTURED_FILE_EXTENSIONS = new Set([
+  'txt', 'md', 'markdown', 'pdf', 'docx', 'pptx', 'json', 'jsonl', 'ndjson',
+])
+const LEGACY_OFFICE_EXTENSIONS = new Set(['doc', 'xls', 'ppt'])
+const MAX_SOURCE_FILE_BYTES = 200 * 1024 * 1024
+const MAX_SOURCE_FILE_COUNT = 20
+const MAX_SOURCE_BATCH_BYTES = 500 * 1024 * 1024
+
+interface SourceUploadJob {
+  uid: string
+  file: File
+  extension: string
+}
+
+interface SourceUploadOptions {
+  taskId: Ref<string | null>
+  uploadedFiles: Ref<UploadedDataFile[]>
+  onUploaded: (file: UploadedDataFile) => void
+}
+
+type SourceFileValidation =
+  | { valid: true; extension: string }
+  | { valid: false; severity: 'error' | 'warning'; message: string }
+
+export function validateSourceFileSelection(
+  raw: File,
+  processType: ProcessType,
+  selectedFiles: UploadedDataFile[],
+): SourceFileValidation {
+  if (raw.size > MAX_SOURCE_FILE_BYTES) {
+    return { valid: false, severity: 'warning', message: '单文件不能超过 200MB' }
+  }
+  const extension = raw.name.split('.').pop()?.toLowerCase() ?? ''
+  if (LEGACY_OFFICE_EXTENSIONS.has(extension)) {
+    return {
+      valid: false,
+      severity: 'error',
+      message: '旧版 DOC、XLS、PPT 文件暂不支持，请分别转换为 DOCX、XLSX、PPTX 后上传',
+    }
+  }
+  const supportedExtensions = processType === 'unstructured'
+    ? UNSTRUCTURED_FILE_EXTENSIONS
+    : STRUCTURED_FILE_EXTENSIONS
+  if (!supportedExtensions.has(extension)) {
+    return {
+      valid: false,
+      severity: 'error',
+      message: processType === 'unstructured'
+        ? '非结构化数据支持 TXT、MD、MARKDOWN、PDF、DOCX、PPTX、JSON、JSONL、NDJSON'
+        : '结构化数据支持 JSON、JSONL、NDJSON、CSV、TSV、XLSX',
+    }
+  }
+  if (selectedFiles.some((file) => file.name === raw.name && file.size === raw.size)) {
+    return { valid: false, severity: 'warning', message: '同名且同大小的文件已经选择' }
+  }
+  if (selectedFiles.length >= MAX_SOURCE_FILE_COUNT) {
+    return { valid: false, severity: 'warning', message: `每个任务最多选择 ${MAX_SOURCE_FILE_COUNT} 个文件` }
+  }
+  const selectedBytes = selectedFiles.reduce((total, file) => total + file.size, 0)
+  if (selectedBytes + raw.size > MAX_SOURCE_BATCH_BYTES) {
+    return { valid: false, severity: 'warning', message: '当前任务选择的文件总大小不能超过 500MB' }
+  }
+  return { valid: true, extension }
+}
+
+export function mapDataProcessSourceFile(
+  file: DataProcessSourceFile,
+  content = '',
+): UploadedDataFile {
+  return {
+    uid: String(file.id),
+    sourceFileId: String(file.id),
+    name: file.name,
+    size: file.size_bytes,
+    count: file.record_count,
+    content,
+    fileFormat: file.file_format,
+    checksumSha256: file.checksum_sha256,
+    status: 'ready',
+    uploadProgress: 100,
+    previewStatus: 'waiting',
+    previewProgress: 0,
+  }
+}
+
+function getUploadErrorMessage(error: unknown) {
+  const responseData = (error as {
+    response?: { data?: { detail?: string | { message?: string }; message?: string } }
+  })?.response?.data
+  const detail = responseData?.detail
+  if (typeof detail === 'string') return detail
+  if (detail?.message) return detail.message
+  if (responseData?.message) return responseData.message
+  if (error instanceof Error && error.message) return error.message
+  return '上传失败，请删除该文件后重新选择'
+}
+
+export function useDataProcessSourceUpload(options: SourceUploadOptions) {
+  const sourceUploading = ref(false)
+  const queue: SourceUploadJob[] = []
+  let worker: Promise<void> | null = null
+
+  async function uploadOne(job: SourceUploadJob) {
+    const pending = options.uploadedFiles.value.find((file) => String(file.uid) === job.uid)
+    if (!pending) return
+
+    const currentTaskId = options.taskId.value
+    if (!currentTaskId) {
+      pending.status = 'failed'
+      pending.uploadError = '任务尚未创建，请返回模型选择步骤后重试'
+      return
+    }
+
+    pending.status = 'uploading'
+    pending.uploadProgress = 0
+    pending.uploadError = undefined
+
+    try {
+      let content = ''
+      if (!BINARY_FILE_EXTENSIONS.has(job.extension)) {
+        try {
+          content = new TextDecoder('utf-8', { fatal: true }).decode(await job.file.arrayBuffer())
+        } catch {
+          throw new Error('文本文件不是有效的 UTF-8 编码，请转换编码后重试')
+        }
+        if (!content.trim()) throw new Error('不能上传空文件')
+      }
+
+      const uploaded = await uploadDataProcessSourceFiles(currentTaskId, [job.file], (progress) => {
+        pending.uploadProgress = progress
+      })
+      const source = uploaded.files[0]
+      if (!source) throw new Error('后端未返回源文件记录')
+
+      // 先登记后端 ID，确保正文读取失败时仍可正确删除已落库的文件。
+      Object.assign(pending, mapDataProcessSourceFile(source), {
+        rawFile: job.file,
+        status: 'uploading',
+        uploadProgress: 99,
+      })
+      if (BINARY_FILE_EXTENSIONS.has(job.extension)) {
+        try {
+          const parsed = await getDataProcessSourceContent(currentTaskId, source.id, {
+            start_line: 1,
+            line_count: 10_000,
+          })
+          pending.content = parsed.content
+        } catch {
+          // 原文件已经成功落库，正文稍后仍可由预览构建接口读取，不重复上传。
+        }
+      } else {
+        pending.content = content
+      }
+
+      pending.status = 'ready'
+      pending.uploadProgress = 100
+      options.onUploaded(pending)
+    } catch (error) {
+      pending.status = 'failed'
+      pending.uploadError = getUploadErrorMessage(error)
+    }
+  }
+
+  async function drainQueue() {
+    sourceUploading.value = true
+    try {
+      while (queue.length) {
+        const job = queue.shift()
+        if (job) await uploadOne(job)
+      }
+    } finally {
+      sourceUploading.value = false
+    }
+  }
+
+  function enqueueSourceUpload(job: SourceUploadJob) {
+    queue.push(job)
+    if (worker) return
+    worker = drainQueue().finally(() => {
+      worker = null
+    })
+  }
+
+  return {
+    enqueueSourceUpload,
+    sourceUploading,
+  }
+}

frontend/vite.config.ts

@@ -44,8 +44,8 @@ export default defineConfig({
     port: 16801,
     proxy: {
       // Frontend uses /modelTF and proxies to the local five-digit backend port.
-      // 后端跑在 WSL2（networkingMode=mirrored）里，uvicorn 必须以
-      // --host 0.0.0.0 启动，Windows 的 localhost 才能到达它。详见 UI测试手册第 0 节。
+      // 后端跑在 WSL2 默认（NAT）网络下，Windows 的 localhost 经 localhostForwarding 转发到 WSL；
+      // 后端以 --host 0.0.0.0 启动即可。不要用 networkingMode=mirrored（本机镜像回环对 WSL 服务失效）。
       '/modelTF': {
         target: 'http://localhost:17861',
         changeOrigin: true,

前端功能失效问题排查.md

@@ -0,0 +1,134 @@
+# 前端功能失效问题排查报告
+
+> 排查时间：2026-07-30
+> 环境：前端（Windows）`localhost:16801` → Vite 代理 `/modelTF` → 后端（WSL）`localhost:17861`
+> 方法：抓取后端 openapi 路由表 + 实测 curl 比对前端真实请求路径
+
+## 一、现象
+
+页面可以打开（app 能正常加载、仪表盘等模块正常），但**大量功能点击无响应或报"请求失败"**。
+经排查，故障集中在"路由前缀叠加导致后端 404"，并非前端崩溃。
+
+## 二、已验证正常的部分（排除法）
+
+| 模块 | 接口 | 实测结果 |
+|------|------|----------|
+| 仪表盘 | `GET /modelTF/dashboard/stats` | ✅ 返回真实数据 |
+| 登录 | `/modelTF/login` | ✅ 正常 |
+| 健康检查 | `/modelTF/health` | ✅ 正常 |
+| 其余 83 个后端路由 | 各 `/modelTF/xxx` | ✅ 均为正确的单层前缀 |
+
+后端共 111 个路由，**仅 28 个 `data-process` 路由异常（双重前缀）**，其余均正确。
+
+## 三、问题清单
+
+### 问题 1（严重，根因）：数据处理模块整体 404
+
+**链路（前缀被叠加了两次 `/modelTF`）：**
+
+1. `backend/app/main.py:22`：`app.include_router(api_router, prefix=settings.route_prefix)`
+   - `route_prefix` 来自 `config.py`，默认值为 `"/modelTF"`
+2. `backend/app/api/v1/router.py`：`api_router.include_router(data_process_router, prefix="/modelTF", ...)`
+   - 这里又额外加了一次 `prefix="/modelTF"`
+3. `backend/app/api/v1/endpoints/data_process.py`：`router = APIRouter(prefix="/data-process")`
+
+**结果**：实际注册路径变成 `/modelTF/modelTF/data-process/...`（双层前缀）。
+
+**影响**：整个"数据处理"模块（列表 / 详情 / 创建任务 / 上传源文件 / 预览 / 生成 / 结果编辑 / 发布 / 重新生成 / 外部拉取测试 等）共 **28 个接口全部 404**。
+
+**前端请求路径**：`baseURL('/modelTF')` + `get('/data-process')` → 实际请求 `/modelTF/data-process/...` → 与后端真实路径不匹配 → 404。
+
+**实测证据**：
+
+```
+前端真实请求  GET /modelTF/data-process
+  -> {"detail":"Not Found"}                          (404)
+
+后端真实路径    GET /modelTF/modelTF/data-process
+  -> {"code":0,"message":"ok","data":{...有"测试"任务}}  (路径存在，有数据)
+```
+
+---
+
+### 问题 2：数据集下载 / 预览 双重前缀 404
+
+`frontend/src/api/request.ts:16` 的 `baseURL: '/modelTF'`，而
+`frontend/src/api/modules/dataset.ts:92` 与 `:98` 的下载地址直接写了**绝对前缀**：
+
+```ts
+const baseUrl = `/modelTF/dataset-manage/download/${datasetId}/${fileId}`
+```
+
+axios 会把 `baseURL('/modelTF')` 与以 `/` 开头的 url 拼接成 `/modelTF/modelTF/dataset-manage/download/...`，导致 404。
+
+**注意**：后端 `dataset-manage` 是**正确的单层** `/modelTF/dataset-manage/...`（实测真实路径返回 500 = 路径存在但资源参数无效，而非 404）。
+
+**实测证据**：
+
+```
+前端实际发出  GET /modelTF/modelTF/dataset-manage/download/x/y  -> 404 Not Found
+后端真实路径    GET /modelTF/dataset-manage/download/x/y          -> 500 (路径存在)
+```
+
+**影响**：数据集的下载、预览功能不可用。
+
+---
+
+### 问题 3：模型导出 双重前缀 + 后端无此端点
+
+`frontend/src/api/modules/model.ts:48` 导出地址同样写了绝对前缀：
+
+```ts
+`/modelTF/model-manage/trained-models/${encodeURIComponent(modelName)}/export`
+```
+
+- 经 axios 拼接 → `/modelTF/modelTF/model-manage/trained-models/.../export` → 404（双重前缀）。
+- **进一步**：openapi 中 `model-manage/trained-models` 仅有 `GET` 列表与 `GET {model_id}` 详情，**并不存在 `/export` 端点**。因此即使修掉双前缀，导出仍会 404，需后端补充该接口或确认正确路径。
+
+**影响**：模型导出功能不可用（两个独立原因叠加）。
+
+---
+
+### 问题 4（次生隐患，修复问题 1 后会暴露）：data-process 源文件 raw 链接
+
+`frontend/src/api/modules/dataProcess.ts:162`：
+
+```ts
+`/modelTF/data-process/${taskId}/source-files/${fileId}/raw`
+```
+
+当前被 axios 拼成 `/modelTF/modelTF/data-process/.../raw`，**恰好命中后端双前缀**，暂时可用。
+**一旦修复问题 1（后端改为单层 `/modelTF/data-process/...`）**，此处会反向变成 404，必须同步去掉多余的 `/modelTF`。
+
+## 四、前端其他使用绝对 `/modelTF` 前缀的位置（汇总）
+
+| 文件:行 | 接口 | 当前状态 |
+|---------|------|----------|
+| `dataProcess.ts:162` | 源文件 raw 下载 | 靠后端双前缀"侥幸"命中，修复问题 1 后失效 |
+| `dataset.ts:92` `:98` | 数据集下载/预览 | 双重前缀 → 404（问题 2） |
+| `model.ts:48` | 模型导出 | 双重前缀 + 后端无端点 → 404（问题 3） |
+
+其余模块（`compare / fineTune / project / eval / model(其他) / dataset(其他)`）均使用相对路径 `/xxx`，正常。
+
+## 五、修复建议（未实施，待确认）
+
+1. **后端** `backend/app/api/v1/router.py`：去掉 `data_process_router` 的 `prefix="/modelTF"`
+   （`api_router` 已挂 `/modelTF`，`data_process` 自身已有 `/data-process`，无需再叠加）。
+   改后 data-process 路径变为 `/modelTF/data-process/...`，与前端请求一致。
+
+2. **前端** `dataProcess.ts:162`：raw url 去掉 `/modelTF`，改为 `/data-process/.../raw`。
+
+3. **前端** `dataset.ts:92/:98`：download url 去掉 `/modelTF`，改为 `/dataset-manage/download/...`。
+
+4. **前端** `model.ts:48`：export url 去掉 `/modelTF`，改为 `/model-manage/trained-models/${name}/export`；
+   同时**与后端确认 `/export` 端点是否存在**（openapi 显示无），需后端补实现或给出正确路径。
+
+5. 重启后端 + 前端，对"数据处理 / 数据集下载 / 模型导出"三个模块做回归。
+
+## 六、根因归类
+
+上述问题是一次改动/合并引入的**路由前缀叠加 bug**：
+- 后端在 `api_router` 已统一挂 `/modelTF` 的前提下，又对 `data_process` 多挂了一次 `/modelTF`；
+- 前端部分下载/导出接口误用了绝对 `/modelTF` 前缀，与 `request.ts` 的 `baseURL` 再次叠加。
+
+修复核心是"前后端前缀只保留一层 `/modelTF`"。