feat(server): 新增文档智能识别服务，扩展OCR接口支持 Azure Document Intelligence

server/tests/test_ocr_service.py

@@ -26,15 +26,15 @@ for index, arg in enumerate(sys.argv):
                 "input_path": input_path,
                 "engine": "paddleocr_mobile",
                 "model": "PP-OCRv5_mobile",
-                "text": "发票金额 100 元",
-                "summary": "发票金额 100 元",
+                "text": "增值税电子发票 发票号码12345678 金额 100 元 2026-05-13",
+                "summary": "增值税电子发票，金额 100 元。",
                 "avg_score": 0.98,
                 "line_count": 1,
                 "page_count": 1,
                 "warnings": [],
                 "lines": [
                     {
-                        "text": "发票金额 100 元",
+                        "text": "增值税电子发票 发票号码12345678 金额 100 元 2026-05-13",
                         "score": 0.98,
                         "box": [[1, 2], [10, 2], [10, 8], [1, 8]],
                         "page_index": 0,
@@ -74,10 +74,106 @@ print("__OCR_JSON__=" + json.dumps(payload, ensure_ascii=False))
     assert len(result.documents) == 2
 
     recognized = next(item for item in result.documents if item.filename == "invoice.png")
-    assert recognized.summary == "发票金额 100 元"
+    assert recognized.summary == "增值税电子发票，金额 100 元。"
     assert recognized.line_count == 1
-    assert recognized.lines[0].text == "发票金额 100 元"
+    assert recognized.document_type == "vat_invoice"
+    assert recognized.document_type_label == "增值税发票"
+    assert any(field.label == "金额" and field.value == "100元" for field in recognized.document_fields)
+    assert any(field.label == "票据号码" and field.value == "12345678" for field in recognized.document_fields)
+    assert any(field.label == "日期" and field.value == "2026-05-13" for field in recognized.document_fields)
+    assert recognized.lines[0].text == "增值税电子发票 发票号码12345678 金额 100 元 2026-05-13"
 
     skipped = next(item for item in result.documents if item.filename == "notes.txt")
     assert skipped.line_count == 0
     assert skipped.warnings == ["当前仅支持图片和 PDF 文件进行 OCR。"]
+
+
+def test_ocr_service_converts_pdf_to_images_and_returns_image_preview(
+    monkeypatch,
+    tmp_path: Path,
+) -> None:
+    def fake_convert_pdf_to_images(self, *, pdf_path: Path, output_dir: Path) -> list[Path]:
+        first = output_dir / "page-1.png"
+        second = output_dir / "page-2.png"
+        first.write_bytes(b"fake-page-1")
+        second.write_bytes(b"fake-page-2")
+        return [first, second]
+
+    def fake_invoke_worker(
+        self,
+        *,
+        python_bin: str,
+        worker_path: str,
+        input_paths: list[Path],
+    ) -> dict:
+        assert [path.name for path in input_paths] == ["page-1.png", "page-2.png"]
+        return {
+            "engine": "paddleocr_mobile",
+            "model": "PP-OCRv5_mobile",
+            "documents": [
+                {
+                    "input_path": str(input_paths[0]),
+                    "engine": "paddleocr_mobile",
+                    "model": "PP-OCRv5_mobile",
+                    "text": "高铁票 深圳北-广州南 车次 G1234 2026-05-13 金额 188 元",
+                    "summary": "高铁票第一页",
+                    "avg_score": 0.97,
+                    "line_count": 1,
+                    "page_count": 1,
+                    "warnings": [],
+                    "lines": [
+                        {
+                            "text": "高铁票 深圳北-广州南 车次 G1234 2026-05-13 金额 188 元",
+                            "score": 0.97,
+                            "box": [[1, 2], [10, 2], [10, 8], [1, 8]],
+                        }
+                    ],
+                },
+                {
+                    "input_path": str(input_paths[1]),
+                    "engine": "paddleocr_mobile",
+                    "model": "PP-OCRv5_mobile",
+                    "text": "乘车人 张三",
+                    "summary": "高铁票第二页",
+                    "avg_score": 0.94,
+                    "line_count": 1,
+                    "page_count": 1,
+                    "warnings": [],
+                    "lines": [
+                        {
+                            "text": "乘车人 张三",
+                            "score": 0.94,
+                            "box": [[1, 2], [10, 2], [10, 8], [1, 8]],
+                        }
+                    ],
+                },
+            ],
+        }
+
+    monkeypatch.setenv("STORAGE_ROOT_DIR", str(tmp_path / "storage"))
+    monkeypatch.setattr(OcrService, "_resolve_python_bin", lambda self: "python")
+    monkeypatch.setattr(OcrService, "_resolve_worker_path", lambda self: "worker.py")
+    monkeypatch.setattr(OcrService, "_convert_pdf_to_images", fake_convert_pdf_to_images)
+    monkeypatch.setattr(OcrService, "_invoke_worker", fake_invoke_worker)
+    get_settings.cache_clear()
+    try:
+        result = OcrService().recognize_files(
+            [
+                ("train-ticket.pdf", b"%PDF-1.4 fake", "application/pdf"),
+            ]
+        )
+    finally:
+        get_settings.cache_clear()
+
+    assert result.success_count == 1
+    assert len(result.documents) == 1
+    recognized = result.documents[0]
+    assert recognized.filename == "train-ticket.pdf"
+    assert recognized.page_count == 2
+    assert recognized.preview_kind == "image"
+    assert recognized.preview_data_url.startswith("data:image/png;base64,")
+    assert recognized.document_type == "train_ticket"
+    assert any(field.label == "金额" and field.value == "188元" for field in recognized.document_fields)
+    assert any(field.label == "车次/航班" and field.value == "G1234" for field in recognized.document_fields)
+    assert recognized.lines[0].page_index == 0
+    assert recognized.lines[1].page_index == 1