支持mineru 2.7.1

docutranslate/app.py

@@ -337,7 +337,7 @@ class TranslateServiceRequest(BaseModel):
         ],
     )
     file_content: str = Field(
-        ..., description="Base64编码的文件内容。", examples=["JVBERi0xLjQK..."]
+        ..., description="Base64编码的文件内容。", examples=["JVBERi0xLjcKJeLjz9MKMSAwIG9iago8PC9..."]
     )
     payload: TranslatePayload = Field(
         ..., description="包含工作流类型和相应参数的载荷。"
@@ -666,7 +666,7 @@ async def _perform_translation(
                     "force_json",
                     "rpm",
                     "tpm",
-                    "provider",  # Added provider
+                    "provider",
                 },
                 exclude_none=True,
             )
@@ -688,7 +688,9 @@ async def _perform_translation(
                 converter_config = ConverterMineruDeployConfig(
                     base_url=payload.mineru_deploy_base_url,
                     backend=payload.mineru_deploy_backend,
+                    parse_method=payload.mineru_deploy_parse_method,
                     formula_enable=payload.mineru_deploy_formula_enable,
+                    table_enable=payload.mineru_deploy_table_enable,
                     start_page_id=payload.mineru_deploy_start_page_id,
                     end_page_id=payload.mineru_deploy_end_page_id,
                     lang_list=payload.mineru_deploy_lang_list,
@@ -2214,8 +2216,10 @@ async def service_flat_translate(
         formula_ocr: bool = Form(True, description="[PDF] 是否启用公式识别"),
         code_ocr: bool = Form(True, description="[Docling] 是否启用代码块识别"),
         mineru_deploy_base_url: str = Form("http://127.0.0.1:8000", description="[MinerU Local] 服务地址"),
-        mineru_deploy_backend: str = Form("VLM", description="[MinerU Local] 后端类型"),
+        mineru_deploy_backend: str = Form("hybrid-auto-engine", description="[MinerU Local] 后端类型: hybrid-auto-engine, pipeline 等"),
+        mineru_deploy_parse_method: str = Form("auto", description="[MinerU Local] 解析方法: auto, txt, ocr"),
         mineru_deploy_formula_enable: bool = Form(True, description="[MinerU Local] 是否启用公式"),
+        mineru_deploy_table_enable: bool = Form(True, description="[MinerU Local] 是否启用表格"),
         mineru_deploy_start_page_id: int = Form(0, description="[MinerU Local] 起始页码"),
         mineru_deploy_end_page_id: int = Form(99999, description="[MinerU Local] 结束页码"),
         mineru_deploy_lang_list: Optional[List[str]] = Form(None, description="[MinerU Local] 语言列表"),
@@ -2317,7 +2321,9 @@ async def service_flat_translate(
         # --- MinerU 本地部署参数 ---
         "mineru_deploy_base_url": mineru_deploy_base_url,
         "mineru_deploy_backend": mineru_deploy_backend,
+        "mineru_deploy_parse_method": mineru_deploy_parse_method,
         "mineru_deploy_formula_enable": mineru_deploy_formula_enable,
+        "mineru_deploy_table_enable": mineru_deploy_table_enable,
         "mineru_deploy_start_page_id": mineru_deploy_start_page_id,
         "mineru_deploy_end_page_id": mineru_deploy_end_page_id,
         "mineru_deploy_lang_list": mineru_deploy_lang_list,
@@ -2531,4 +2537,4 @@ def run_app(host=None, port: int | None = None, enable_CORS=False,
 
 
 if __name__ == "__main__":
-    run_app()
+    run_app()

docutranslate/converter/x2md/converter_mineru_deploy.py

@@ -2,7 +2,7 @@
 # SPDX-License-Identifier: MPL-2.0
 import asyncio
 from dataclasses import dataclass
-from typing import Literal, Hashable
+from typing import Literal, Hashable, List
 
 import httpx
 
@@ -16,24 +16,42 @@ from docutranslate.utils.markdown_utils import embed_inline_image_from_zip
 @dataclass(kw_only=True)
 class ConverterMineruDeployConfig(X2MarkdownConverterConfig):
     base_url: str = "http://127.0.0.1:8000"
-    output_dir: str = "./output"  # 覆盖默认值 ./output
+    output_dir: str = "./output"
-    lang_list: list[Literal["ch", "ch_server", "ch_lite", "en", "korean", "japan", "chinese_cht", "ta", "te", "ka", "th", "el", "latin", "arabic", "east_slavic", "cyrillic", "devanagari"]] | None = None
+    # 支持的语言列表 (来自 MinerU API)
-    backend: Literal["pipeline", "vlm-transformers", "vlm-mlx-engine", "vlm-vllm-async-engine", "vlm-lmdeploy-engine","vlm-http-client"] = "pipeline"
+    lang_list: List[str] | None = None  # 默认值在 API 侧处理，这里 None 即可
-    # parse_method: str = "auto"
+
+    # 后端引擎选项 (更新适配最新的 MinerU API)
+    backend: Literal[
+        "pipeline",
+        "vlm-auto-engine",
+        "vlm-http-client",
+        "hybrid-auto-engine",
+        "hybrid-http-client"
+    ] = "hybrid-auto-engine"
+
+    parse_method: Literal["auto", "txt", "ocr"] = "auto"
     formula_enable: bool = True
-    # table_enable: bool = True
+    table_enable: bool = True
-    server_url: str | None = None #(Adapted only for vlm-http-client backend)openai compatible server url, e.g., http://127.0.0.1:30000
+
-    # return_md: bool = True
+    # 用于 vlm-http-client 或 hybrid-http-client 后端
-    # return_middle_json: bool = True
+    server_url: str | None = None
-    # return_model_output: bool = False
+
-    # return_content_list: bool = False
+    # 返回选项
-    # return_images: bool = True
+    return_md: bool = True
-    # response_format_zip: bool = True
+    return_middle_json: bool = False
+    return_model_output: bool = False
+    return_content_list: bool = False
+    return_images: bool = True
+    response_format_zip: bool = True
+
+    # 页面范围
     start_page_id: int = 0
     end_page_id: int = 99999
 
-    def gethash(self) ->Hashable:
+    def gethash(self) -> Hashable:
-        return (self.backend,self.formula_enable,self.start_page_id,self.end_page_id)
+        return (self.backend, self.formula_enable, self.table_enable,
+                self.parse_method, self.start_page_id, self.end_page_id)
+
 
 # 配置HTTP客户端
 timeout = httpx.Timeout(
@@ -57,27 +75,37 @@ class ConverterMineruDeploy(X2MarkdownConverter):
 
         self._api_url = f"{self.base_url}/file_parse"
 
-    def _build_form_data(self)->dict:
+    def _build_form_data(self) -> dict:
+        # httpx 在处理 data 参数时，如果值为 list，会自动展开为多个同名 key (例如 lang_list=ch&lang_list=en)
+        # 这符合 FastAPI/Starlette 对 List 字段的解析要求
         data = {
             "output_dir": self.config.output_dir,
             "backend": self.config.backend,
-            "parse_method": "auto",
+            "parse_method": self.config.parse_method,
-            "formula_enable": self.config.formula_enable,
+            # bool 类型在 multipart/form-data 中通常需要转为字符串 'true'/'false'，但 httpx 会处理 python bool
-            "table_enable": True,
+            "formula_enable": str(self.config.formula_enable).lower(),
-            "server_url": None,
+            "table_enable": str(self.config.table_enable).lower(),
-            "return_md": True,
+            "return_md": str(self.config.return_md).lower(),
-            "return_middle_json": True,
+            "return_middle_json": str(self.config.return_middle_json).lower(),
-            "return_model_output": False,
+            "return_model_output": str(self.config.return_model_output).lower(),
-            "return_content_list": False,
+            "return_content_list": str(self.config.return_content_list).lower(),
-            "return_images": True,
+            "return_images": str(self.config.return_images).lower(),
-            "response_format_zip": True,
+            "response_format_zip": str(self.config.response_format_zip).lower(),
             "start_page_id": self.config.start_page_id,
             "end_page_id": self.config.end_page_id
         }
+
+        if self.config.lang_list:
+            data["lang_list"] = self.config.lang_list
+        else:
+            data["lang_list"] = ["ch"]  # 默认值
+
+        if self.config.server_url:
+            data["server_url"] = self.config.server_url
+
         return data
 
-
+    def convert(self, d: Document) -> MarkdownDocument:
-    def convert(self,d:Document)->MarkdownDocument:
         self.logger.info("开始解析文件")
         files = [("files", (d.name, d.content, "application/octet-stream"))]
         response = client.post(
@@ -88,15 +116,15 @@ class ConverterMineruDeploy(X2MarkdownConverter):
         )
 
         response.raise_for_status()  # 检查是否有错误
-        md=embed_inline_image_from_zip(response.content,None)
+        # Mineru API 返回 zip 时包含图片和 md
+        md = embed_inline_image_from_zip(response.content, None)
         self.logger.info("已转化为markdown")
-        return MarkdownDocument.from_bytes(md.encode(),suffix=".md",stem=d.stem)
+        return MarkdownDocument.from_bytes(md.encode(), suffix=".md", stem=d.stem)
-
 
     async def convert_async(self, d: Document) -> MarkdownDocument:
         self.logger.info("开始解析文件")
         files = [("files", (d.name, d.content, "application/octet-stream"))]
-        response =await client_async.post(
+        response = await client_async.post(
             self._api_url,
             files=files,
             data=self._build_form_data(),
@@ -104,15 +132,9 @@ class ConverterMineruDeploy(X2MarkdownConverter):
         )
 
         response.raise_for_status()
-        md = await asyncio.to_thread(embed_inline_image_from_zip,response.content, None)
+        md = await asyncio.to_thread(embed_inline_image_from_zip, response.content, None)
         self.logger.info("已转化为markdown")
         return MarkdownDocument.from_bytes(md.encode(), suffix=".md", stem=d.stem)
 
     def support_format(self) -> list[str]:
-        return [".pdf", ".doc", ".docx", ".ppt", ".pptx", ".png", ".jpg", ".jpeg"]
+        return [".pdf", ".doc", ".docx", ".ppt", ".pptx", ".png", ".jpg", ".jpeg"]
-
-if __name__ == '__main__':
-    d = Document.from_path(r"C:\Users\jxgm\Desktop\testfiles\table.pdf")
-    config=ConverterMineruDeployConfig()
-    converter = ConverterMineruDeploy(config=config)
-    converter.convert(d)

docutranslate/core/schemas.py

@@ -252,17 +252,27 @@ class MarkdownWorkflowParams(BaseWorkflowParams):
         "http://127.0.0.1:8000",
         description="[仅当 convert_engine='mineru_deploy'] 本地部署的 MinerU 服务地址。",
     )
+    # --- UPDATED BACKEND LIST ---
     mineru_deploy_backend: Literal[
         "pipeline",
-        "vlm-transformers",
+        "vlm-auto-engine",
-        "vlm-mlx-engine",
-        "vlm-vllm-async-engine",
-        "vlm-lmdeploy-engine",
         "vlm-http-client",
+        "hybrid-auto-engine",
+        "hybrid-http-client"
     ] = Field(
-        "pipeline",
+        "hybrid-auto-engine",
         description="[仅当 convert_engine='mineru_deploy'] 本地部署的 MinerU 服务使用的后端。",
     )
+    # --- NEW PARAMETERS START ---
+    mineru_deploy_parse_method: Literal["auto", "txt", "ocr"] = Field(
+        "auto",
+        description="[仅当 convert_engine='mineru_deploy'] 解析方法: auto, txt, ocr"
+    )
+    mineru_deploy_table_enable: bool = Field(
+        True,
+        description="[仅当 convert_engine='mineru_deploy'] 本地部署的服务是否启用表格解析。",
+    )
+    # --- NEW PARAMETERS END ---
     mineru_deploy_formula_enable: bool = Field(
         True,
         description="[仅当 convert_engine='mineru_deploy'] 本地部署的服务是否启用公式解析。",
@@ -275,13 +285,13 @@ class MarkdownWorkflowParams(BaseWorkflowParams):
     )
     mineru_deploy_lang_list: Optional[List[str]] = Field(
         None,
-        description="[仅当 convert_engine='mineru_deploy' 且 backend='pipeline'] 语言列表。",
+        description="[仅当 convert_engine='mineru_deploy'] 语言列表, 默认 ['ch']。",
-        examples=[None],
+        examples=[["ch", "en"]],
     )
     # 修改: 默认值改为 ""
     mineru_deploy_server_url: Optional[str] = Field(
         default="",
-        description="[仅当 convert_engine='mineru_deploy' 且 backend='vlm-http-client'] Server URL.",
+        description="[仅当 convert_engine='mineru_deploy' 且 backend为http-client相关时] Server URL.",
     )
 
     @model_validator(mode="after")
@@ -312,10 +322,6 @@ class TextWorkflowParams(BaseWorkflowParams):
         "\n",
         description="当 insert_mode 为 'append' 或 'prepend' 时，用于分隔原文和译文的分隔符。",
     )
-    segment_mode: Literal["line", "paragraph", "none"] = Field(
-        "line",
-        description="分段模式。'line'：按行分段（每行独立翻译），'paragraph'：按段落分段（连续非空行合并为段落），'none'：不分段（全文视为一个段落）。",
-    )
 
 
 class JsonWorkflowParams(BaseWorkflowParams):