优化OCR解析

rag/flow/parser/parser.py

@@ -29,7 +29,8 @@ from api.db.services.llm_service import LLMBundle
 from api.utils import get_uuid
 from api.utils.base64_image import image2id
 from deepdoc.parser import ExcelParser
-from deepdoc.parser.pdf_parser import PlainParser, RAGFlowPdfParser, VisionParser
+from deepdoc.parser.pdf_parser import PlainParser, VisionParser
+from ocr.service import get_ocr_service
 from rag.app.naive import Docx
 from rag.flow.base import ProcessBase, ProcessParamBase
 from rag.flow.parser.schema import ParserFromUpstream
@@ -204,7 +205,9 @@ class Parser(ProcessBase):
         self.set_output("output_format", conf["output_format"])
 
         if conf.get("parse_method").lower() == "deepdoc":
-            bboxes = RAGFlowPdfParser().parse_into_bboxes(blob, callback=self.callback)
+            # 注意：HTTP 调用中无法传递 callback，callback 将被忽略
+            ocr_service = get_ocr_service()
+            bboxes = ocr_service.parse_into_bboxes_sync(blob, callback=self.callback, filename=name)
         elif conf.get("parse_method").lower() == "plain_text":
             lines, _ = PlainParser()(blob)
             bboxes = [{"text": t} for t, _ in lines]