feat(data-process): 增加 Office 原文件预览接口

backend/app/api/v1/endpoints/data_process.py

@@ -53,6 +53,11 @@ from app.modules.data_process.document_chunking import (
     merge_short_chunks,
 )
 from app.modules.data_process.generation import generate_model_records
+from app.modules.data_process.office_preview import (
+    MAX_XLSX_PREVIEW_ROWS,
+    build_docx_preview,
+    build_xlsx_preview,
+)
 from app.modules.data_process.storage import (
     LocalDataProcessStorage,
     StagedSourceObject,
@@ -111,6 +116,12 @@ LEGACY_OFFICE_CONVERSIONS = {
     ".xls": ".xlsx",
     ".ppt": ".pptx",
 }
+RAW_INLINE_PREVIEW_MEDIA_TYPES = {
+    "pdf": "application/pdf",
+    "docx": "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
+    "xlsx": "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
+    "pptx": "application/vnd.openxmlformats-officedocument.presentationml.presentation",
+}
 
 
 def ok(data: Any = None, message: str = "ok") -> dict[str, Any]:
@@ -897,8 +908,13 @@ def source_file_raw(
 ) -> StreamingResponse:
     with api_errors():
         source = store.get_source_file(task_id, file_id, include_content=False)
-        if str(source.get("file_format") or "").lower() != "pdf":
-            raise fail(415, "raw inline preview is only available for PDF source files")
+        file_format = str(source.get("file_format") or "").lower()
+        media_type = RAW_INLINE_PREVIEW_MEDIA_TYPES.get(file_format)
+        if media_type is None:
+            raise fail(
+                415,
+                "raw inline preview is only available for PDF and modern Office source files",
+            )
         storage_object_id = str(source.get("storage_object_id") or "")
         actual_size = storage.file_size(
             storage_object_id,
@@ -906,14 +922,15 @@ def source_file_raw(
             expected_source_file_id=file_id,
         )
         if actual_size is None:
-            raise fail(410, "the original PDF is unavailable for this legacy source file")
+            raise fail(410, "the original file is unavailable for this legacy source file")
         expected_size = int(source.get("size_bytes") or 0)
         if actual_size != expected_size:
             raise ValueError("source object size does not match metadata")
         selected_range = _source_byte_range(range_header, actual_size)
         start, end = selected_range or (0, actual_size - 1)
         length = end - start + 1
-        name = _safe_file_name(str(source.get("name") or "source.pdf"), "source.pdf")
+        default_name = f"source.{file_format}"
+        name = _safe_file_name(str(source.get("name") or default_name), default_name)
         headers = {
             "Accept-Ranges": "bytes",
             "Cache-Control": "private, no-store",
@@ -938,11 +955,61 @@ def source_file_raw(
         return StreamingResponse(
             body,
             status_code=206 if selected_range is not None else 200,
-            media_type="application/pdf",
+            media_type=media_type,
             headers=headers,
         )
 
 
+@router.get("/{task_id}/source-files/{file_id}/office-preview")
+def source_file_office_preview(
+    task_id: str,
+    file_id: str,
+    sheet_index: int = Query(default=0, ge=0),
+    offset: int = Query(default=0, ge=0),
+    limit: int = Query(default=100, ge=1, le=MAX_XLSX_PREVIEW_ROWS),
+    store: DataProcessStore = Depends(get_data_process_store),
+    storage: LocalDataProcessStorage = Depends(get_data_process_storage),
+) -> dict[str, Any]:
+    """返回 Word 版式块或 Excel 工作表网格，不把二进制内容下发给组件解析。"""
+
+    with api_errors():
+        source = store.get_source_file(task_id, file_id, include_content=False)
+        file_format = str(source.get("file_format") or "").lower()
+        if file_format not in {"docx", "xlsx"}:
+            raise fail(415, "Office preview is only available for DOCX and XLSX source files")
+        storage_object_id = str(source.get("storage_object_id") or "")
+        actual_size = storage.file_size(
+            storage_object_id,
+            expected_task_id=task_id,
+            expected_source_file_id=file_id,
+        )
+        if actual_size is None:
+            raise fail(410, "the original Office file is unavailable for this legacy source file")
+        expected_size = int(source.get("size_bytes") or 0)
+        if actual_size != expected_size:
+            raise ValueError("source object size does not match metadata")
+        raw = b"".join(
+            storage.iter_bytes(
+                storage_object_id,
+                expected_task_id=task_id,
+                expected_source_file_id=file_id,
+                expected_size=actual_size,
+            )
+        )
+        preview = (
+            build_docx_preview(raw)
+            if file_format == "docx"
+            else build_xlsx_preview(
+                raw,
+                sheet_index=sheet_index,
+                offset=offset,
+                limit=limit,
+            )
+        )
+        preview["file_name"] = str(source.get("name") or f"source.{file_format}")
+        return ok(preview)
+
+
 @router.get("/{task_id}/source-files/{file_id}/pdf-pages")
 def source_file_pdf_pages(
     task_id: str,