399 lines
17 KiB
Python
399 lines
17 KiB
Python
"""
|
||
MaxKB平台适配器 - 实现MaxKB特定的功能
|
||
|
||
这个适配器实现了MaxKB平台的特定功能，包括：
|
||
- MaxKB的trace上下文管理（如果有）
|
||
- MaxKB的锁机制（如果有）
|
||
- MaxKB的文件存储客户端
|
||
- MaxKB的模型客户端
|
||
"""
|
||
|
||
import contextlib
|
||
import os
|
||
from typing import Any, Dict, Optional
|
||
from .logger import get_module_logger
|
||
logger = get_module_logger('adapter')
|
||
|
||
# 导入MaxKB特有的模块
|
||
from .file_storage_client import FileStorageClient
|
||
from .maxkb_model_client import maxkb_model_client
|
||
|
||
# 导入基类从上级目录
|
||
from ..base_parser import PlatformAdapter, BaseMinerUExtractor
|
||
|
||
|
||
class MaxKBAdapter(PlatformAdapter):
|
||
"""MaxKB平台的适配器实现"""
|
||
|
||
def __init__(self):
|
||
"""初始化MaxKB适配器"""
|
||
self.file_storage = FileStorageClient()
|
||
self.model_client = maxkb_model_client
|
||
|
||
# 导入配置以获取存储路径
|
||
from .config_maxkb import MaxKBMinerUConfig
|
||
self.config = MaxKBMinerUConfig()
|
||
self.storage_path = self.config.file_storage_path
|
||
|
||
@contextlib.asynccontextmanager
|
||
async def trace_context(self, trace_id: str):
|
||
"""MaxKB的trace上下文 - 如果没有特殊实现，使用简单的上下文"""
|
||
# MaxKB可能没有trace_context，这里提供一个简单的实现
|
||
logger.info(f"MaxKB: Starting trace {trace_id}")
|
||
try:
|
||
yield
|
||
finally:
|
||
logger.info(f"MaxKB: Ending trace {trace_id}")
|
||
|
||
async def lock_enter(self, temp_dir: str):
|
||
"""MaxKB的锁机制进入 - 如果没有特殊实现，创建目录即可"""
|
||
# MaxKB可能没有特殊的锁机制，确保目录存在即可
|
||
os.makedirs(temp_dir, exist_ok=True)
|
||
logger.debug(f"MaxKB: Entered lock for {temp_dir}")
|
||
|
||
async def lock_release(self, temp_dir: str):
|
||
"""MaxKB的锁机制释放 - 如果没有特殊实现，简单记录即可"""
|
||
logger.debug(f"MaxKB: Released lock for {temp_dir}")
|
||
|
||
async def upload_file(self, file_path: str, options: Any = None) -> str:
|
||
"""使用MaxKB的文件存储上传文件 - 直接复制文件到存储目录"""
|
||
import shutil
|
||
import uuid
|
||
|
||
logger.info(f"MaxKB: upload_file called with path={file_path}, options={options}")
|
||
|
||
# 如果在测试模式下，直接返回原图地址
|
||
#if os.getenv('MINERU_TEST_FILE'):
|
||
# logger.info(f"MaxKB: Test mode - returning original path: {file_path}")
|
||
# return file_path
|
||
|
||
try:
|
||
# 确保文件存在
|
||
if not os.path.exists(file_path):
|
||
logger.warning(f"MaxKB: File not found: {file_path}")
|
||
return file_path
|
||
|
||
# 获取knowledge_id（如果在options中提供）
|
||
knowledge_id = None
|
||
if options and isinstance(options, (tuple, list)) and len(options) > 0:
|
||
knowledge_id = options[0]
|
||
|
||
# 创建存储目录结构
|
||
# 使用 knowledge_id 或 'mineru' 作为子目录
|
||
sub_dir = knowledge_id if knowledge_id else 'mineru'
|
||
storage_dir = os.path.join(self.storage_path, sub_dir, 'images')
|
||
|
||
# 确保存储目录存在
|
||
os.makedirs(storage_dir, exist_ok=True)
|
||
|
||
# 生成唯一的文件名，保留原始扩展名
|
||
file_ext = os.path.splitext(file_path)[1]
|
||
file_name = f"{uuid.uuid4().hex}{file_ext}"
|
||
dest_path = os.path.join(storage_dir, file_name)
|
||
|
||
# 复制文件到存储目录
|
||
shutil.copy2(file_path, dest_path)
|
||
|
||
# 返回相对路径或URL格式
|
||
# 生成相对于storage根目录的路径
|
||
relative_path = os.path.relpath(dest_path, self.storage_path)
|
||
# 确保路径使用正斜杠（兼容所有系统）
|
||
relative_path = relative_path.replace(os.path.sep, '/')
|
||
|
||
# 根据环境配置生成完整的URL
|
||
# 检查是否配置了基础URL
|
||
base_url = os.getenv('MAXKB_BASE_URL', '')
|
||
if base_url:
|
||
# 如果有基础URL，生成完整的URL
|
||
result_url = f"{base_url.rstrip('/')}/storage/{relative_path}"
|
||
else:
|
||
# 生成相对URL，直接使用/storage/路径
|
||
result_url = f"/storage/{relative_path}"
|
||
|
||
logger.info(f"MaxKB: Copied file {file_path} -> {dest_path}")
|
||
logger.info(f"MaxKB: Returning URL: {result_url}")
|
||
|
||
return result_url
|
||
|
||
except Exception as e:
|
||
logger.error(f"MaxKB: Failed to copy file {file_path}: {str(e)}")
|
||
# 如果复制失败，返回本地路径
|
||
return file_path
|
||
|
||
def get_logger(self):
|
||
"""获取MaxKB的日志器"""
|
||
return logger
|
||
|
||
def get_settings(self) -> Dict[str, Any]:
|
||
"""获取MaxKB的配置"""
|
||
# 返回MaxKB的配置，可以从环境变量或配置文件读取
|
||
return {
|
||
'api_key': os.getenv('MAXKB_API_KEY'),
|
||
'api_url': os.getenv('MAXKB_API_URL', 'https://api.maxkb.com'),
|
||
'model': os.getenv('MAXKB_DEFAULT_MODEL', 'gpt-4o'),
|
||
# 添加其他MaxKB特有的配置
|
||
}
|
||
|
||
def get_learn_type(self, params: Dict[str, Any]) -> int:
|
||
"""获取learn_type参数 - MaxKB使用model_id映射到learn_type"""
|
||
# MaxKB使用llm_model_id和vision_model_id
|
||
# 这里可以根据model_id映射到learn_type
|
||
llm_model_id = params.get('llm_model_id')
|
||
vision_model_id = params.get('vision_model_id')
|
||
|
||
# 根据模型ID映射到learn_type（这里是示例映射）
|
||
if llm_model_id:
|
||
# 可以根据具体的模型ID返回不同的learn_type
|
||
return 9 # 默认返回9
|
||
return 9
|
||
|
||
def set_trace_id(self, trace_id: str):
|
||
"""设置trace ID用于日志跟踪"""
|
||
# MaxKB可能有自己的trace机制
|
||
# 这里简单记录日志
|
||
logger.debug(f"MaxKB: Setting trace ID to: {trace_id}")
|
||
|
||
|
||
class MinerUExtractor(BaseMinerUExtractor):
|
||
"""
|
||
MaxKB平台的MinerU解析器
|
||
|
||
继承自基类，使用MaxKB适配器
|
||
"""
|
||
|
||
def __init__(self, llm_model_id: Optional[str] = None,
|
||
vision_model_id: Optional[str] = None):
|
||
"""
|
||
初始化MaxKB的MinerU解析器
|
||
|
||
Args:
|
||
llm_model_id: 大语言模型ID（MaxKB特有参数）
|
||
vision_model_id: 视觉模型ID（MaxKB特有参数）
|
||
"""
|
||
# 创建MaxKB适配器
|
||
adapter = MaxKBAdapter()
|
||
|
||
# 导入并创建MaxKB特定的配置，传递模型ID
|
||
from .config_maxkb import MaxKBMinerUConfig
|
||
config = MaxKBMinerUConfig(llm_model_id=llm_model_id, vision_model_id=vision_model_id)
|
||
|
||
# 调用基类初始化，传递适配器、配置和MaxKB特有参数
|
||
super().__init__(
|
||
adapter,
|
||
config=config,
|
||
llm_model_id=llm_model_id,
|
||
vision_model_id=vision_model_id
|
||
)
|
||
|
||
# 保存MaxKB特有的参数
|
||
self.llm_model_id = llm_model_id
|
||
self.vision_model_id = vision_model_id
|
||
|
||
# 如果需要，可以在这里初始化MaxKB特有的组件
|
||
self.model_client = adapter.model_client
|
||
self.file_storage = adapter.file_storage
|
||
|
||
async def process_file_with_models(self, filepath: str, src_name: str = None,
|
||
upload_options: Any = None) -> Any:
|
||
"""
|
||
MaxKB特有的处理方法 - 使用指定的模型处理文件
|
||
|
||
这个方法展示了如何在子类中添加平台特有的功能
|
||
"""
|
||
logger.info(f"MaxKB: Processing with LLM: {self.llm_model_id}, Vision: {self.vision_model_id}")
|
||
|
||
# 调用基类的处理方法
|
||
result = await self.process_file(filepath, src_name, upload_options)
|
||
|
||
# 可以在这里添加MaxKB特有的后处理
|
||
if self.llm_model_id:
|
||
logger.info(f"MaxKB: Used LLM model {self.llm_model_id} for processing")
|
||
if self.vision_model_id:
|
||
logger.info(f"MaxKB: Used Vision model {self.vision_model_id} for image processing")
|
||
|
||
return result
|
||
|
||
|
||
class MinerUAdapter:
|
||
"""
|
||
MinerU文档处理适配器 - 用于MaxKB中处理文档
|
||
"""
|
||
|
||
def __init__(self):
|
||
"""初始化MinerU适配器"""
|
||
self.extractor = None
|
||
self._init_extractor()
|
||
|
||
def _init_extractor(self):
|
||
"""初始化MinerU解析器"""
|
||
try:
|
||
# 获取配置的模型ID
|
||
llm_model_id = os.environ.get('MINERU_LLM_MODEL_ID')
|
||
vision_model_id = os.environ.get('MINERU_VISION_MODEL_ID')
|
||
|
||
# 创建解析器
|
||
self.extractor = MinerUExtractor(
|
||
llm_model_id=llm_model_id,
|
||
vision_model_id=vision_model_id
|
||
)
|
||
logger.info("MinerU适配器初始化成功")
|
||
except Exception as e:
|
||
logger.error(f"MinerU适配器初始化失败: {str(e)}")
|
||
raise
|
||
|
||
def process_document(self, file_content: bytes, file_name: str,
|
||
save_image_func=None, **kwargs) -> Dict[str, Any]:
|
||
"""
|
||
处理文档并返回结构化内容
|
||
|
||
Args:
|
||
file_content: 文件内容字节流
|
||
file_name: 文件名
|
||
save_image_func: 保存图片的函数
|
||
**kwargs: 额外参数，包括llm_model_id和vision_model_id
|
||
|
||
Returns:
|
||
包含sections的字典，每个section包含content、title和images
|
||
"""
|
||
import tempfile
|
||
import asyncio
|
||
import threading
|
||
|
||
try:
|
||
# 创建临时文件
|
||
with tempfile.NamedTemporaryFile(suffix=os.path.splitext(file_name)[1],
|
||
delete=False) as tmp_file:
|
||
tmp_file.write(file_content)
|
||
tmp_file_path = tmp_file.name
|
||
|
||
try:
|
||
# 在新线程中运行异步代码，避免事件循环冲突
|
||
result = None
|
||
exception = None
|
||
|
||
def run_async():
|
||
nonlocal result, exception
|
||
try:
|
||
loop = asyncio.new_event_loop()
|
||
asyncio.set_event_loop(loop)
|
||
try:
|
||
# 提取模型ID参数
|
||
llm_model_id = kwargs.get('llm_model_id')
|
||
vision_model_id = kwargs.get('vision_model_id')
|
||
if llm_model_id and vision_model_id:
|
||
logger.info(f"使用指定模型处理文档: LLM={llm_model_id}, Vision={vision_model_id}")
|
||
# TODO: 将模型ID传递给extractor
|
||
# 目前暂时使用默认配置，后续可以在这里设置模型
|
||
|
||
result = loop.run_until_complete(
|
||
self.extractor.process_file(tmp_file_path, file_name)
|
||
)
|
||
finally:
|
||
loop.close()
|
||
except Exception as e:
|
||
exception = e
|
||
|
||
thread = threading.Thread(target=run_async)
|
||
thread.start()
|
||
thread.join(timeout=300) # 5分钟超时
|
||
|
||
if exception:
|
||
raise exception
|
||
|
||
if result is None:
|
||
logger.warning("MinerU处理超时或无结果")
|
||
return {'sections': []}
|
||
|
||
logger.info(f"MinerU返回结果类型: {type(result)}")
|
||
|
||
# 转换结果格式
|
||
sections = []
|
||
|
||
# 检查result的类型
|
||
if isinstance(result, list):
|
||
# result是一个页面文档列表（Langchain Document对象）
|
||
for page_doc in result:
|
||
# 处理Langchain Document对象
|
||
if hasattr(page_doc, 'page_content') and hasattr(page_doc, 'metadata'):
|
||
page_content = page_doc.page_content
|
||
metadata = page_doc.metadata
|
||
|
||
if page_content:
|
||
# 提取页码
|
||
page_num = metadata.get('page', metadata.get('page_num', ''))
|
||
|
||
# 提取图片列表
|
||
images = metadata.get('images', [])
|
||
|
||
sections.append({
|
||
'content': page_content,
|
||
'title': f"Page {page_num}" if page_num else '',
|
||
'images': images
|
||
})
|
||
elif isinstance(page_doc, dict):
|
||
# 字典格式的文档
|
||
page_content = page_doc.get('text', page_doc.get('page_content', ''))
|
||
if page_content:
|
||
sections.append({
|
||
'content': page_content,
|
||
'title': f"Page {page_doc.get('page_num', '')}",
|
||
'images': page_doc.get('images', [])
|
||
})
|
||
elif isinstance(result, dict):
|
||
# result是一个字典
|
||
if 'content' in result:
|
||
content = result['content']
|
||
if isinstance(content, str):
|
||
sections.append({
|
||
'content': content,
|
||
'title': '',
|
||
'images': []
|
||
})
|
||
elif isinstance(content, list):
|
||
for item in content:
|
||
if isinstance(item, dict):
|
||
sections.append({
|
||
'content': item.get('text', ''),
|
||
'title': item.get('title', ''),
|
||
'images': item.get('images', [])
|
||
})
|
||
else:
|
||
sections.append({
|
||
'content': str(item),
|
||
'title': '',
|
||
'images': []
|
||
})
|
||
|
||
# 处理图片（如果有保存函数）
|
||
if save_image_func and sections:
|
||
for section in sections:
|
||
if section.get('images'):
|
||
saved_images = []
|
||
for img_path in section['images']:
|
||
try:
|
||
# 检查图片文件是否存在
|
||
if os.path.exists(img_path):
|
||
with open(img_path, 'rb') as f:
|
||
img_content = f.read()
|
||
saved_path = save_image_func(img_content)
|
||
saved_images.append(saved_path)
|
||
else:
|
||
saved_images.append(img_path)
|
||
except Exception as e:
|
||
logger.warning(f"保存图片失败 {img_path}: {str(e)}")
|
||
saved_images.append(img_path)
|
||
section['images'] = saved_images
|
||
|
||
logger.info(f"MinerU处理完成，提取了{len(sections)}个sections")
|
||
return {'sections': sections}
|
||
|
||
finally:
|
||
# 清理临时文件
|
||
if os.path.exists(tmp_file_path):
|
||
os.unlink(tmp_file_path)
|
||
|
||
except Exception as e:
|
||
logger.error(f"MinerU处理文档失败: {str(e)}")
|
||
# 返回空结果而不是抛出异常，让调用方可以回退到其他处理器
|
||
return {'sections': []}
|