feat(data-process): 增加 Office 原文件预览接口

backend/tests/test_data_process_api.py

@@ -6,6 +6,7 @@ from pathlib import Path
 from typing import Any
 
 import pytest
+from docx import Document as WordDocument
 from fastapi import FastAPI
 from fastapi.testclient import TestClient
 from openpyxl import Workbook
@@ -1546,11 +1547,112 @@ def test_xlsx_upload_is_accepted_as_structured_records(tmp_path: Path) -> None:
     )
     assert content.status_code == 200
     assert '"answer":"答案二"' in content.json()["data"]["content"]
+    office_preview_url = (
+        f"/modelTF/data-process/{task_id}/source-files/{source['id']}/office-preview"
+    )
+    first_page = client.get(office_preview_url, params={"offset": 0, "limit": 1})
+    assert first_page.status_code == 200
+    preview_data = first_page.json()["data"]
+    assert preview_data["format"] == "xlsx"
+    assert preview_data["sheets"] == [{"index": 0, "name": "Sheet", "state": "visible"}]
+    assert preview_data["active_sheet"]["columns"] == ["question", "answer"]
+    assert preview_data["active_sheet"]["rows"][0]["record"] == {
+        "question": "问题一",
+        "answer": "答案一",
+    }
+    assert preview_data["active_sheet"]["has_more"] is True
+
+    second_page = client.get(office_preview_url, params={"offset": 1, "limit": 1})
+    assert second_page.status_code == 200
+    assert second_page.json()["data"]["active_sheet"]["rows"][0]["record"] == {
+        "question": "问题二",
+        "answer": "答案二",
+    }
+    assert second_page.json()["data"]["active_sheet"]["has_more"] is False
+
+    raw_preview = client.get(
+        f"/modelTF/data-process/{task_id}/source-files/{source['id']}/raw"
+    )
+    assert raw_preview.status_code == 200
+    assert raw_preview.content == original_bytes
+    assert raw_preview.headers["content-type"].startswith(
+        "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"
+    )
     preview = client.post(f"/modelTF/data-process/{task_id}/preview/build")
     assert preview.status_code == 200
     assert preview.json()["data"]["total"] == 2
 
 
+def test_docx_preview_preserves_document_block_order_and_source_offsets(
+    tmp_path: Path,
+) -> None:
+    client, store, _ = make_client(tmp_path)
+    task_id = client.post(
+        "/modelTF/data-process",
+        json={"name": "Word 原件预览", "process_type": "unstructured", "config": {}},
+    ).json()["data"]["id"]
+    document = WordDocument()
+    document.add_heading("费用管理办法", level=1)
+    document.add_paragraph("第一条 本办法用于规范费用报销。")
+    table = document.add_table(rows=2, cols=2)
+    table.cell(0, 0).text = "费用类型"
+    table.cell(0, 1).text = "审批人"
+    table.cell(1, 0).text = "差旅费"
+    table.cell(1, 1).text = "部门负责人"
+    output = BytesIO()
+    document.save(output)
+    original_bytes = output.getvalue()
+
+    uploaded = client.post(
+        f"/modelTF/data-process/{task_id}/source-files",
+        files={
+            "files": (
+                "费用 管理.docx",
+                original_bytes,
+                "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
+            )
+        },
+    ).json()["data"]["files"][0]
+    response = client.get(
+        f"/modelTF/data-process/{task_id}/source-files/{uploaded['id']}/office-preview"
+    )
+
+    assert response.status_code == 200
+    data = response.json()["data"]
+    assert data["format"] == "docx"
+    assert data["file_name"] == "费用 管理.docx"
+    assert data["truncated"] is False
+    assert [block["type"] for block in data["blocks"]] == [
+        "paragraph",
+        "paragraph",
+        "table",
+    ]
+    assert data["blocks"][0]["heading_level"] == 1
+    assert data["blocks"][0]["text"] == "费用管理办法"
+    assert data["blocks"][2]["rows"][1]["cells"] == ["差旅费", "部门负责人"]
+
+    source_text = store.get_source_file(task_id, uploaded["id"])["content"]
+    first_paragraph = data["blocks"][0]
+    assert (
+        source_text[first_paragraph["source_start"] : first_paragraph["source_end"]]
+        == first_paragraph["text"]
+    )
+    table_row = data["blocks"][2]["rows"][1]
+    assert (
+        source_text[table_row["source_start"] : table_row["source_end"]]
+        == "\t".join(table_row["cells"])
+    )
+
+    raw_preview = client.get(
+        f"/modelTF/data-process/{task_id}/source-files/{uploaded['id']}/raw"
+    )
+    assert raw_preview.status_code == 200
+    assert raw_preview.content == original_bytes
+    assert raw_preview.headers["content-type"].startswith(
+        "application/vnd.openxmlformats-officedocument.wordprocessingml.document"
+    )
+
+
 def test_pdf_raw_preview_streams_original_file_and_supports_ranges(tmp_path: Path) -> None:
     client, store, _ = make_client(tmp_path)
     task_id = client.post(