384 lines
16 KiB
Python
384 lines
16 KiB
Python
# SPDX-FileCopyrightText: 2025 QinHan
|
||
# SPDX-License-Identifier: MPL-2.0
|
||
# docutranslate/sdk.py
|
||
|
||
import asyncio
|
||
import base64
|
||
import os
|
||
from pathlib import Path
|
||
from typing import Optional, Literal, Dict, Any, List, Union
|
||
|
||
from pydantic import TypeAdapter
|
||
|
||
from docutranslate.agents.agent import ThinkingMode
|
||
from docutranslate.agents.provider import ProviderType
|
||
|
||
from docutranslate.core.schemas import TranslatePayload, GlossaryAgentConfigPayload, WorkflowType, InsertMode
|
||
from docutranslate.core.factory import create_workflow_from_payload
|
||
from docutranslate.translator import default_params
|
||
from docutranslate.global_values.conditional_import import DOCLING_EXIST
|
||
|
||
# --- 映射配置 ---
|
||
# 格式说明: {workflow_type: {save_type: (method_name, default_suffix)}}
|
||
# 注意: 第一个格式为默认格式
|
||
_WORKFLOW_MAPPINGS = {
|
||
"markdown_based": {
|
||
"markdown": ("save_as_markdown", "md"), # 默认格式: 内嵌 base64 图片
|
||
"markdown_zip": ("save_as_markdown_zip", "zip"), # 非内嵌图片，分离存储
|
||
"html": ("save_as_html", "html"),
|
||
},
|
||
"docx": {"docx": ("save_as_docx", "docx")},
|
||
"xlsx": {"xlsx": ("save_as_xlsx", "xlsx")},
|
||
"pptx": {"pptx": ("save_as_pptx", "pptx")},
|
||
"epub": {"epub": ("save_as_epub", "epub")},
|
||
"txt": {"txt": ("save_as_txt", "txt")},
|
||
"json": {"json": ("save_as_json", "json")},
|
||
"srt": {"srt": ("save_as_srt", "srt")},
|
||
"ass": {"ass": ("save_as_ass", "ass")},
|
||
"html": {"html": ("save_as_html", "html")},
|
||
}
|
||
|
||
# 每种工作流支持的输出格式列表
|
||
_WORKFLOW_SUPPORTED_FORMATS = {
|
||
wf: list(formats.keys()) for wf, formats in _WORKFLOW_MAPPINGS.items()
|
||
}
|
||
|
||
|
||
class TranslationResult:
|
||
"""
|
||
封装翻译结果，负责后续的保存或导出操作。
|
||
"""
|
||
|
||
def __init__(self, workflow: Any, workflow_type: str, original_filename: str):
|
||
self._workflow = workflow
|
||
self._workflow_type = workflow_type
|
||
self._original_filename = original_filename
|
||
self._mapping = _WORKFLOW_MAPPINGS.get(workflow_type)
|
||
self._supported_formats = _WORKFLOW_SUPPORTED_FORMATS.get(workflow_type, [])
|
||
|
||
@property
|
||
def supported_formats(self) -> List[str]:
|
||
"""获取当前工作流支持的输出格式"""
|
||
return self._supported_formats
|
||
|
||
def save(
|
||
self,
|
||
output_dir: str = "./output",
|
||
name: Optional[str] = None,
|
||
fmt: Optional[str] = None
|
||
) -> str:
|
||
"""
|
||
保存结果到文件系统。
|
||
|
||
:param output_dir: 输出目录。
|
||
:param name: 文件名 (如 'result.html')。若为 None，使用原文件名 + 对应后缀。
|
||
:param fmt: 输出格式 (如 'html', 'markdown', 'markdown_zip')。
|
||
若为 None，使用工作流默认格式。
|
||
:return: 保存文件的完整路径 (仅供参考)。
|
||
"""
|
||
if not self._mapping:
|
||
raise ValueError(f"工作流 {self._workflow_type} 不支持自动保存")
|
||
|
||
# 确定使用的格式
|
||
if fmt:
|
||
if fmt not in self._mapping:
|
||
raise ValueError(
|
||
f"格式 '{fmt}' 不支持。可用格式: {self._supported_formats}"
|
||
)
|
||
method_name, default_suffix = self._mapping[fmt]
|
||
else:
|
||
# 使用默认第一个格式
|
||
fmt = self._supported_formats[0]
|
||
method_name, default_suffix = self._mapping[fmt]
|
||
|
||
# 生成文件名
|
||
if not name:
|
||
base_name = os.path.splitext(self._original_filename)[0]
|
||
name = f"{base_name}.{default_suffix}"
|
||
|
||
Path(output_dir).mkdir(parents=True, exist_ok=True)
|
||
|
||
method = getattr(self._workflow, method_name, None)
|
||
if method:
|
||
method(name=name, output_dir=output_dir)
|
||
return str(Path(output_dir) / name)
|
||
raise AttributeError(f"Workflow 缺少方法 {method_name}")
|
||
|
||
def export(self, fmt: Optional[str] = None) -> str:
|
||
"""
|
||
导出为 Base64 编码的字符串 (用于 API 传输或无需落盘的场景)。
|
||
|
||
:param fmt: 输出格式。若为 None，使用工作流默认格式。
|
||
:return: Base64 编码的结果。
|
||
"""
|
||
if not self._mapping:
|
||
raise ValueError(f"工作流 {self._workflow_type} 不支持导出")
|
||
|
||
# 确定使用的格式
|
||
if fmt:
|
||
if fmt not in self._mapping:
|
||
raise ValueError(
|
||
f"格式 '{fmt}' 不支持。可用格式: {self._supported_formats}"
|
||
)
|
||
export_key = fmt
|
||
else:
|
||
# 使用默认第一个格式
|
||
fmt = self._supported_formats[0]
|
||
export_key = fmt
|
||
|
||
# 构建方法名
|
||
if export_key == "markdown_zip":
|
||
method_name = "export_to_markdown_zip"
|
||
else:
|
||
method_name = f"export_to_{export_key}"
|
||
|
||
method = getattr(self._workflow, method_name, None)
|
||
if method:
|
||
content = method()
|
||
if isinstance(content, str):
|
||
content = content.encode('utf-8')
|
||
return base64.b64encode(content).decode('utf-8')
|
||
raise AttributeError(f"Workflow 缺少方法 {method_name}")
|
||
|
||
@property
|
||
def workflow(self):
|
||
"""获取底层 Workflow 实例以访问高级功能 (如附件)"""
|
||
return self._workflow
|
||
|
||
|
||
class Client:
|
||
"""
|
||
DocuTranslate SDK。
|
||
"""
|
||
|
||
def __init__(
|
||
self,
|
||
api_key: Optional[str] = None,
|
||
base_url: Optional[str] = None,
|
||
model_id: Optional[str] = None,
|
||
to_lang: str = "中文",
|
||
concurrent: int = default_params["concurrent"],
|
||
timeout: int = default_params["timeout"],
|
||
retry: int = default_params["retry"],
|
||
thinking: ThinkingMode = default_params["thinking"],
|
||
system_proxy_enable: bool = default_params["system_proxy_enable"],
|
||
convert_engine: Literal["identity", "mineru", "docling", "mineru_deploy"] = "identity",
|
||
mineru_token: str = "",
|
||
**kwargs
|
||
):
|
||
"""
|
||
初始化 SDK 实例。
|
||
此处设置的参数将作为全局默认值，可在调用 translate 时被覆盖。
|
||
"""
|
||
self.defaults = {
|
||
"api_key": api_key, "base_url": base_url, "model_id": model_id,
|
||
"to_lang": to_lang, "concurrent": concurrent, "timeout": timeout,
|
||
"retry": retry, "thinking": thinking,
|
||
"system_proxy_enable": system_proxy_enable,
|
||
"convert_engine": convert_engine, "mineru_token": mineru_token,
|
||
**kwargs
|
||
}
|
||
self.defaults = {k: v for k, v in self.defaults.items() if v is not None}
|
||
|
||
def translate(
|
||
self,
|
||
file_path: str,
|
||
*,
|
||
# --- 为了获得 IDE 提示，必须在这里显式列出所有参数 ---
|
||
api_key: Optional[str] = None,
|
||
base_url: Optional[str] = None,
|
||
model_id: Optional[str] = None,
|
||
to_lang: Optional[str] = None,
|
||
workflow_type: WorkflowType = "auto",
|
||
skip_translate: bool = False,
|
||
concurrent: Optional[int] = None,
|
||
chunk_size: Optional[int] = None,
|
||
temperature: Optional[float] = None,
|
||
timeout: Optional[int] = None,
|
||
retry: Optional[int] = None,
|
||
thinking: Optional[ThinkingMode] = None,
|
||
custom_prompt: Optional[str] = None,
|
||
source_lang: Optional[str] = None,
|
||
system_proxy_enable: Optional[bool] = None,
|
||
force_json: Optional[bool] = None,
|
||
rpm: Optional[int] = None,
|
||
tpm: Optional[int] = None,
|
||
provider: Optional[Union[ProviderType, str]] = None,
|
||
insert_mode: Literal["replace", "append", "prepend"] = "replace",
|
||
separator: str = "\n",
|
||
segment_mode: Literal["line", "paragraph", "none"] = "line",
|
||
translate_regions: Optional[List[str]] = None,
|
||
convert_engine: Literal["identity", "mineru", "docling", "mineru_deploy"] = "identity",
|
||
mineru_token: str = "",
|
||
model_version: Literal["pipeline", "vlm"] = "vlm",
|
||
formula_ocr: bool = True,
|
||
code_ocr: bool = True,
|
||
mineru_deploy_base_url: str = "http://127.0.0.1:8000",
|
||
mineru_deploy_backend: Literal["pipeline", "vlm-auto-engine", "vlm-http-client", "hybrid-auto-engine", "hybrid-http-client"] = "hybrid-auto-engine",
|
||
mineru_deploy_parse_method: Literal["auto", "txt", "ocr"] = "auto",
|
||
mineru_deploy_formula_enable: bool = True,
|
||
mineru_deploy_table_enable: bool = True,
|
||
mineru_deploy_start_page_id: int = 0,
|
||
mineru_deploy_end_page_id: int = 99999,
|
||
mineru_deploy_lang_list: Optional[List[str]] = None,
|
||
mineru_deploy_server_url: str = "",
|
||
json_paths: Optional[List[str]] = None,
|
||
glossary_generate_enable: Optional[bool] = None,
|
||
glossary_dict: Optional[Dict[str, str]] = None,
|
||
glossary_agent_config: Optional[Union[GlossaryAgentConfigPayload, Dict[str, Any]]] = None,
|
||
**kwargs
|
||
) -> TranslationResult:
|
||
"""
|
||
同步执行翻译。参数说明请参考 translate_async。
|
||
"""
|
||
# 获取当前函数的所有参数（这包含了你传入的 api_key, model_id 等）
|
||
# 排除掉 self，剩下的就是传给 async 函数的参数
|
||
args = locals()
|
||
call_params = {k: v for k, v in args.items() if k != 'self'}
|
||
|
||
# 剔除 kwargs 避免双重传递 (因为 call_params 已经包含了 kwargs 的内容)
|
||
if 'kwargs' in call_params:
|
||
extra = call_params.pop('kwargs')
|
||
call_params.update(extra)
|
||
|
||
return asyncio.run(self.translate_async(**call_params))
|
||
|
||
async def translate_async(
|
||
self,
|
||
file_path: str,
|
||
*,
|
||
# --- 核心覆盖参数 ---
|
||
api_key: Optional[str] = None,
|
||
base_url: Optional[str] = None,
|
||
model_id: Optional[str] = None,
|
||
to_lang: Optional[str] = None,
|
||
|
||
# --- 流程控制 ---
|
||
workflow_type: WorkflowType = "auto",
|
||
skip_translate: bool = False,
|
||
|
||
# --- LLM 参数 ---
|
||
concurrent: Optional[int] = None,
|
||
chunk_size: Optional[int] = None,
|
||
temperature: Optional[float] = None,
|
||
timeout: Optional[int] = None,
|
||
retry: Optional[int] = None,
|
||
thinking: Optional[ThinkingMode] = None,
|
||
custom_prompt: Optional[str] = None,
|
||
source_lang: Optional[str] = None,
|
||
system_proxy_enable: Optional[bool] = None,
|
||
force_json: Optional[bool] = None,
|
||
rpm: Optional[int] = None,
|
||
tpm: Optional[int] = None,
|
||
provider: Optional[Union[ProviderType, str]] = None,
|
||
|
||
# --- 格式参数 (Docx/Excel/Txt) ---
|
||
insert_mode: Literal["replace", "append", "prepend"] = "replace",
|
||
separator: str = "\n",
|
||
segment_mode: Literal["line", "paragraph", "none"] = "line",
|
||
translate_regions: Optional[List[str]] = None,
|
||
|
||
# --- 解析引擎 (PDF/OCR) ---
|
||
convert_engine: Literal["identity", "mineru", "docling", "mineru_deploy"] = "identity",
|
||
mineru_token: str = "",
|
||
model_version: Literal["pipeline", "vlm"] = "vlm",
|
||
formula_ocr: bool = True,
|
||
code_ocr: bool = True,
|
||
|
||
# --- Mineru 本地部署参数 ---
|
||
mineru_deploy_base_url: str = "http://127.0.0.1:8000",
|
||
mineru_deploy_backend: Literal["pipeline", "vlm-auto-engine", "vlm-http-client", "hybrid-auto-engine", "hybrid-http-client"] = "hybrid-auto-engine",
|
||
mineru_deploy_parse_method: Literal["auto", "txt", "ocr"] = "auto",
|
||
mineru_deploy_formula_enable: bool = True,
|
||
mineru_deploy_table_enable: bool = True,
|
||
mineru_deploy_start_page_id: int = 0,
|
||
mineru_deploy_end_page_id: int = 99999,
|
||
mineru_deploy_lang_list: Optional[List[str]] = None,
|
||
mineru_deploy_server_url: str = "",
|
||
|
||
# --- JSON / 术语表 ---
|
||
json_paths: Optional[List[str]] = None,
|
||
glossary_generate_enable: Optional[bool] = None,
|
||
glossary_dict: Optional[Dict[str, str]] = None,
|
||
glossary_agent_config: Optional[Union[GlossaryAgentConfigPayload, Dict[str, Any]]] = None,
|
||
|
||
**kwargs
|
||
) -> TranslationResult:
|
||
"""
|
||
异步执行翻译任务。
|
||
|
||
:param file_path: 输入文件路径 (必需)。
|
||
:param workflow_type: 工作流类型 (auto, docx, markdown_based, xlsx, json, txt)。
|
||
:param skip_translate: 若为 True，仅进行解析/OCR，不调用 LLM 翻译。
|
||
:param concurrent: LLM 请求并发数。
|
||
:param json_paths: [Json专用] JsonPath 列表 (如 '$.data.*')。
|
||
:param translate_regions: [Excel专用] 翻译区域 (如 'Sheet1!A1:B10')。
|
||
:param insert_mode: [Docx/Xlsx/Txt] 译文插入模式 (replace, append, prepend)。
|
||
:param convert_engine: [PDF/OCR] 解析引擎 (mineru, docling)。
|
||
:param mineru_token: [Mineru Cloud] API Token。
|
||
:param mineru_deploy_base_url: [Mineru Local] 本地服务地址。
|
||
"""
|
||
|
||
# 1. 获取所有参数
|
||
current_args = locals()
|
||
call_params = {
|
||
k: v for k, v in current_args.items()
|
||
if k not in ['self', 'file_path', 'kwargs'] and v is not None
|
||
}
|
||
call_params.update({k: v for k, v in kwargs.items() if v is not None})
|
||
|
||
# 2. 参数层级合并
|
||
final_params = {**default_params, **self.defaults, **call_params}
|
||
|
||
# 3. 文件校验
|
||
path_obj = Path(file_path)
|
||
if not path_obj.exists():
|
||
raise FileNotFoundError(f"文件不存在: {file_path}")
|
||
|
||
# 4. 自动检测 Workflow 类型
|
||
if final_params.get("workflow_type", "auto") == "auto":
|
||
final_params["workflow_type"] = self._detect_workflow(path_obj)
|
||
|
||
# 5. 智能补全
|
||
self._patch_defaults(final_params, path_obj)
|
||
|
||
# 6. Pydantic 校验
|
||
try:
|
||
payload = TypeAdapter(TranslatePayload).validate_python(final_params)
|
||
except Exception as e:
|
||
raise ValueError(f"参数配置校验失败: {e}")
|
||
|
||
# 7. 创建 Workflow
|
||
workflow = create_workflow_from_payload(payload)
|
||
|
||
# 8. 执行逻辑
|
||
workflow.read_path(str(path_obj))
|
||
|
||
# 仅当 skip_translate 为 False 时才执行翻译
|
||
if not payload.skip_translate:
|
||
await workflow.translate_async()
|
||
|
||
return TranslationResult(workflow, final_params["workflow_type"], path_obj.name)
|
||
|
||
def _detect_workflow(self, path: Path) -> str:
|
||
ext = path.suffix.lower().lstrip(".")
|
||
if ext in ["md", "pdf", "png", "jpg"]: return "markdown_based"
|
||
if ext in ["xlsx", "csv", "xls"]: return "xlsx"
|
||
if ext in ["docx", "doc"]: return "docx"
|
||
if ext in ["html", "htm"]: return "html"
|
||
if ext in ["pptx", "ppt"]: return "pptx"
|
||
if ext in ["txt", "json", "srt", "epub", "ass"]: return ext
|
||
return "txt"
|
||
|
||
def _patch_defaults(self, params: Dict[str, Any], path: Path):
|
||
wf = params.get("workflow_type")
|
||
if wf == "json" and not params.get("json_paths"):
|
||
params["json_paths"] = ["$..*"]
|
||
if wf == "markdown_based" and "convert_engine" not in params:
|
||
ext = path.suffix.lower()
|
||
if ext == ".pdf":
|
||
params["convert_engine"] = "mineru" if not DOCLING_EXIST else "docling"
|
||
else:
|
||
params["convert_engine"] = "identity"
|
||
|