From 0a748abb0c670ccec42798ea6ff550661760cf3e Mon Sep 17 00:00:00 2001 From: xunbu Date: Fri, 1 Aug 2025 15:27:53 +0800 Subject: [PATCH] =?UTF-8?q?=E5=AE=8C=E5=96=84thinking=E6=A8=A1=E5=BC=8F?= =?UTF-8?q?=E3=80=81=E4=B8=B4=E6=97=B6=E5=90=8C=E6=AD=A5=E6=8E=A5=E5=8F=A3?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- docutranslate/agents/agent.py | 2 +- docutranslate/agents/markdown_agent.py | 1 - docutranslate/agents/txt_agent.py | 1 - docutranslate/app.py | 96 ++++++------------- docutranslate/static/index.html | 2 +- .../ai_translator/txt_translator.py | 19 ++-- 6 files changed, 43 insertions(+), 78 deletions(-) diff --git a/docutranslate/agents/agent.py b/docutranslate/agents/agent.py index 76bb5b0..25eb738 100644 --- a/docutranslate/agents/agent.py +++ b/docutranslate/agents/agent.py @@ -68,7 +68,7 @@ TIMEOUT = 600 class Agent: _think_factory = { - "open.bigmodel.cn": ("thinking", "enable", "disabled") + "open.bigmodel.cn": ("thinking", {"type":"enabled"}, {"type":"disabled"}) } def __init__(self, config: AgentConfig): diff --git a/docutranslate/agents/markdown_agent.py b/docutranslate/agents/markdown_agent.py index 5a885c4..0e7ab9a 100644 --- a/docutranslate/agents/markdown_agent.py +++ b/docutranslate/agents/markdown_agent.py @@ -44,4 +44,3 @@ $$1+1=2$$ \\((c_0,c_1,c_2^2)\\)是一个坐标。""" if config.custom_prompt: self.system_prompt += "\n# 重要规则或背景【非常重要】\n" + config.custom_prompt + '\n' - self.system_prompt += r'\no_think' diff --git a/docutranslate/agents/txt_agent.py b/docutranslate/agents/txt_agent.py index 946396c..55a4ee4 100644 --- a/docutranslate/agents/txt_agent.py +++ b/docutranslate/agents/txt_agent.py @@ -27,4 +27,3 @@ class TXTTranslateAgent(Agent): """ if config.custom_prompt: self.system_prompt += "\n# 重要规则或背景【非常重要】\n" + config.custom_prompt + '\n' - self.system_prompt += r'\no_think' diff --git a/docutranslate/app.py b/docutranslate/app.py index 4d50196..85e9f16 100644 --- a/docutranslate/app.py +++ b/docutranslate/app.py @@ -23,6 +23,7 @@ from pydantic import BaseModel, Field, field_validator from docutranslate import __version__ from docutranslate.agents.agent import ThinkingMode from docutranslate.cacher import md_based_convert_cacher +from docutranslate.exporter.md.types import ConvertEngineType # --- 核心代码 Imports --- from docutranslate.global_values.conditional_import import DOCLING_EXIST from docutranslate.workflow.base import Workflow @@ -70,23 +71,6 @@ def _create_default_task_state() -> Dict[str, Any]: } -# --- [KEPT FOR TEMP ENDPOINT] 旧的 Workflow 工厂函数 (仅为临时接口保留) --- -def _get_workflow_for_file(filename: str, logger: logging.Logger) -> Workflow: - """根据文件名后缀选择并返回合适的 Workflow 实例。这是扩展点。""" - suffix = Path(filename).suffix.lower() - if suffix == '.txt': - logger.info("检测到 .txt 文件，使用 TXTWorkflow。") - # 为临时接口创建虚拟config - return TXTWorkflow(config=TXTWorkflowConfig(translator_config=None, html_exporter_config=None, logger=logger)) - else: - # 默认为基于 Markdown 的流程（处理 .pdf, .docx, .md 等） - logger.info(f"检测到 {suffix} 文件，使用 MarkdownBasedWorkflow。") - # 为临时接口创建虚拟config - return MarkdownBasedWorkflow(config=MarkdownBasedWorkflowConfig( - convert_engine=None, converter_config=None, translator_config=None, html_exporter_config=None, logger=logger - )) - - # --- 日志处理器 (保持不变) --- class QueueAndHistoryHandler(logging.Handler): def __init__(self, queue_ref: asyncio.Queue, history_list_ref: List[str], max_history_items: int, task_id: str): @@ -192,6 +176,8 @@ class BaseWorkflowParams(BaseModel): chunk_size: int = Field(default=default_params["chunk_size"], description="文本分割的块大小（字符）。") concurrent: int = Field(default=default_params["concurrent"], description="并发请求数。") temperature: float = Field(default=default_params["temperature"], description="LLM温度参数。") + thinking: ThinkingMode = Field(default_params["thinking"], description="是否启用深度思考", + examples=["default", "enable", "disable"]), custom_prompt: Optional[str] = Field(None, description="用户自定义的翻译Prompt。", alias="custom_prompt") @@ -200,10 +186,10 @@ class MarkdownWorkflowParams(BaseWorkflowParams): workflow_type: Literal['markdown_based'] = Field(..., description="指定使用基于Markdown的翻译工作流。") # --- Markdown-specific Converter Params --- - convert_engine: Optional[Literal["mineru", "docling"]] = Field( - None, - description="文档解析引擎。`mineru`在线服务, `docling`本地引擎。如果输入文件是.md，此项可为`null`或不传。", - examples=["mineru", "docling"] + convert_engine: ConvertEngineType = Field( + "identity", + description="文档解析引擎。`identity`处理markdown文件,`mineru`在线服务, `docling`本地引擎。如果输入文件是.md，此项可为`null`或不传。", + examples=["identity", "mineru", "docling"] ) mineru_token: Optional[str] = Field(None, description="当 `convert_engine` 为 'mineru' 时必填的API令牌。") formula_ocr: bool = Field(True, description="是否对公式进行OCR识别。对 `mineru` 和 `docling` 均有效。") @@ -251,6 +237,7 @@ class TranslateServiceRequest(BaseModel): "chunk_size": 3000, "concurrent": 10, "temperature": 0.1, + "thinking": "enable", "custom_prompt": "将所有技术术语翻译为业界公认的中文对应词汇。" } } @@ -294,7 +281,7 @@ async def _perform_translation( translator_config = MDTranslatorConfig( **payload.model_dump(include={ 'base_url', 'api_key', 'model_id', 'to_lang', 'custom_prompt', - 'temperature', 'timeout', 'chunk_size', 'concurrent' + 'temperature', 'thinking', 'timeout', 'chunk_size', 'concurrent' }, exclude_none=True) ) @@ -328,7 +315,7 @@ async def _perform_translation( translator_config = TXTTranslatorConfig( **payload.model_dump(include={ 'base_url', 'api_key', 'model_id', 'to_lang', 'custom_prompt', - 'temperature', 'timeout', 'chunk_size', 'concurrent' + 'temperature', 'thinking', 'timeout', 'chunk_size', 'concurrent' }, exclude_none=True) ) @@ -1076,62 +1063,41 @@ async def temp_translate( model_id: str = Body(..., description="使用的模型ID。", examples=["gpt-4-turbo"]), mineru_token: Optional[str] = Body(None, description="Mineru引擎的Token。"), file_name: str = Body(..., - description="文件名，用以判断文件类型。当后缀为txt时该接口返回普通文本，为其他后缀时返回翻译后的markdown文本", + description="文件名，用以判断文件类型。", examples=["test.txt", "test.md", "test.pdf"]), file_content: str = Body(..., description="文件内容，可以是纯文本或Base64编码的字符串。"), to_lang: str = Body("中文", description="目标语言。", examples=["中文", "英文", "English"]), concurrent: int = Body(default_params["concurrent"], description="ai翻译请求并发数"), temperature: float = Body(default_params["temperature"], description="ai翻译请求温度"), - thinking: ThinkingMode = Body(default_params["thinking"], description="是否启用深度思考", examples=["default", "enable", "disable"]), + thinking: ThinkingMode = Body(default_params["thinking"], description="是否启用深度思考", + examples=["default", "enable", "disable"]), chunk_size: int = Body(default_params["chunk_size"], description="文本分块大小（bytes）"), custom_prompt: Optional[str] = Body(None, description="翻译自定义提示词", - examples=["人名保持原文不翻译"]), + examples=["人名保持原文不翻译"]), ): """一个用于快速测试的同步翻译接口。""" + file_name = Path(file_name) try: decoded_content = base64.b64decode(file_content) except (ValueError, binascii.Error): decoded_content = file_content.encode('utf-8') try: - # [MODIFIED] 使用旧的辅助函数，仅为这个临时接口服务 - workflow = _get_workflow_for_file(file_name, global_logger) - workflow.read_bytes(decoded_content, Path(file_name).stem, Path(file_name).suffix) - - # Manually set up configs for the workflow instance for this temp endpoint - if isinstance(workflow, MarkdownBasedWorkflow): - translator_config = MDTranslatorConfig( - base_url=base_url, api_key=api_key, model_id=model_id, to_lang=to_lang, - custom_prompt=custom_prompt, temperature=temperature,thinking=thinking, - chunk_size=chunk_size, concurrent=concurrent, logger=global_logger, timeout=2000 - ) - convert_config = ConverterMineruConfig(mineru_token=mineru_token, - formula_ocr=True) if mineru_token else None - - # Update workflow's internal config for translate() to work - workflow.config.translator_config = translator_config - workflow.config.converter_config = convert_config - workflow.config.convert_engine = 'mineru' if mineru_token and convert_config else 'identity' - - await workflow.translate_async() - return {"success": True, "content": workflow.export_to_markdown()} - - elif isinstance(workflow, TXTWorkflow): - translator_config = TXTTranslatorConfig( - base_url=base_url, api_key=api_key, model_id=model_id, to_lang=to_lang, - custom_prompt=custom_prompt, temperature=temperature,thinking=thinking, - chunk_size=chunk_size, concurrent=concurrent, logger=global_logger, timeout=2000 - ) - - # Update workflow's internal config - workflow.config.translator_config = translator_config - - await workflow.translate_async() - return {"success": True, "content": workflow.export_to_txt()} - - else: - raise NotImplementedError(f"Temp endpoint does not support workflow type {type(workflow).__name__}") - + workflow_config = MarkdownBasedWorkflowConfig(convert_engine="mineru", + converter_config=ConverterMineruConfig(mineru_token=mineru_token), + translator_config=MDTranslatorConfig(base_url=base_url, + api_key=api_key, + model_id=model_id, + to_lang=to_lang, + custom_prompt=custom_prompt, + temperature=temperature, + thinking=thinking, + chunk_size=chunk_size, + concurrent=concurrent), + html_exporter_config=MD2HTMLExporterConfig()) + workflow = MarkdownBasedWorkflow(workflow_config) + workflow.read_bytes(content=decoded_content, stem=file_name.stem, suffix=file_name.stem) + return {"success": True, "content": workflow.export_to_markdown()} except Exception as e: global_logger.error(f"临时翻译接口出现错误：{e.__repr__()}", exc_info=True) return {"success": False, "reason": e.__repr__()} @@ -1163,4 +1129,4 @@ def run_app(port: int | None = None): if __name__ == "__main__": - run_app() \ No newline at end of file + run_app() diff --git a/docutranslate/static/index.html b/docutranslate/static/index.html index 03fdd69..1efe1ca 100644 --- a/docutranslate/static/index.html +++ b/docutranslate/static/index.html @@ -1 +1 @@ - DocuTranslate - 交互式文档翻译

DocuTranslate

如果上传的文件本身是.md格式，此项可不选。

控制翻译时是否启用多步思考。部分模型可能不支持。

GitHub主页(欢迎star❤):
https://github.com/xunbu/docutranslate

交流QQ群: 1047781902

任务列表

当前没有任务，点击“新建任务”开始吧！

预览
原文
译文
\ No newline at end of file + DocuTranslate - 交互式文档翻译

DocuTranslate

如果上传的文件本身是.md格式，此项可不选。

GitHub主页(欢迎star❤):
https://github.com/xunbu/docutranslate

交流QQ群: 1047781902

任务列表

当前没有任务，点击“新建任务”开始吧！

预览
原文
译文
\ No newline at end of file diff --git a/docutranslate/translator/ai_translator/txt_translator.py b/docutranslate/translator/ai_translator/txt_translator.py index 4ded8c8..243e13b 100644 --- a/docutranslate/translator/ai_translator/txt_translator.py +++ b/docutranslate/translator/ai_translator/txt_translator.py @@ -18,15 +18,16 @@ class TXTTranslator(Translator): super().__init__(config=config) self.chunk_size = config.chunk_size agent_config = TXTTranslateAgentConfig(custom_prompt=config.custom_prompt, - to_lang=config.to_lang, - baseurl=config.base_url, - key=config.api_key, - model_id=config.model_id, - system_prompt=None, - temperature=config.temperature, - max_concurrent=config.concurrent, - timeout=config.timeout, - logger=self.logger) + to_lang=config.to_lang, + baseurl=config.base_url, + key=config.api_key, + model_id=config.model_id, + system_prompt=None, + temperature=config.temperature, + thinking=config.thinking, + max_concurrent=config.concurrent, + timeout=config.timeout, + logger=self.logger) self.translate_agent = TXTTranslateAgent(agent_config) def translate(self, document: Document) -> Self: