【AI RAG本地知识库管理-confluence空间加载】

在搭建本地知识库过程中，需要把confluence空间的内容加载到 本地知识库中，下面介绍如何保存confluence内容

from atlassian import Confluence
from langchain.schema import Document
import requests


class ConfluenceLoader:
    """Confluence文档加载器，支持空间、指定页面或目录加载"""
    
    def __init__(self, url, username, password, space_key=None, page_ids=None, directory_page_ids=None):
        self.url = url
        self.username = username
        self.password = password
        self.space_key = space_key
        self.page_ids = page_ids if page_ids else []
        self.directory_page_ids = directory_page_ids if directory_page_ids else []

              # 创建自定义Session，强制添加Host头
        self.session = requests.Session()
        self.session.headers.update({"Host": "conf2.XXXXX.com"})
        self.session.auth = (self.username, self.password)

    def load(self):
        """加载Confluence文档"""
        try:
            confluence = Confluence(
                url=self.url,
                username=self.username,
                password=self.password,
                session=self.session,  # 注入自定义Session
                cloud=False  # 如果是云版本需要设置为True
            )
            documents = []
            
            if self.space_key:
                self._load_space_content(confluence, documents)
            elif self.directory_page_ids:
                for directory_page_id in self.directory_page_ids:
                    print(f"*******Confluence加载目录******: {directory_page_id}")
                    self._load_directory_content(confluence, documents, directory_page_id)
                
            return documents
        except Exception as e:
            print(f"Confluence加载失败: {str(e)}")
            return []

    def _load_space_content(self, confluence, documents):
        """加载指定空间的所有页面"""
        start = 0
        limit = 50
        while True:
            pages = confluence.get_all_pages_from_space(
                self.space_key,
                start=start,
                limit=limit,
                expand='body.storage'
            )
            if not pages:
                break
            for page in pages:
                self._process_page(confluence, page['id'], documents)
            start += limit

        def _load_directory_content(self, confluence, documents, parent_page_id):
        """先加载子页面，再加载父页面（避免权限继承问题）"""
        try:
            # 先处理子页面
            child_page_ids = self._get_child_page_ids(confluence, parent_page_id)
            for page_id in child_page_ids:
                self._load_directory_content(confluence, documents, page_id)
            # 最后处理当前目录
            self._process_page(confluence, parent_page_id, documents)
        except Exception as e:
            print(f"加载目录失败: {str(e)}")

    def _get_child_page_ids(self, confluence, parent_page_id):
        try:
            children = confluence.get_page_child_by_type(
                parent_page_id,
                type="page",
                start=0,
                limit=50
            )
            return [child["id"] for child in children]
        except Exception as e:
            print(f"获取子页面失败: {str(e)}")
            return []

    def _process_page(self, confluence, page_id, documents):
        """处理单个页面内容"""
        try:
            page = confluence.get_page_by_id(page_id, expand='body.storage')
            content = page['body']['storage']['value']
            metadata = {
                "source": f"{self.url}/pages/viewpage.action?pageId={page_id}",
                "title": page['title'],
                "page_id": page_id
            }
            print(f"加载页面，id: {page_id}")
            documents.append(Document(page_content=content, metadata=metadata))
        except Exception as e:
            print(f"加载页面 {page_id} 失败: {str(e)}")


# 使用示例
if __name__ == "__main__":
    # 加载指定目录
    loader = ConfluenceLoader(
    url="http://ip:port",
    username="xxx",
    password="xxx",
    directory_page_id=["xxxx"]  # 目录页面的ID
    )   
    documents = loader.load()

注意的问题

• 如果confleunce 的访问，是不需要使用ip加端口的方式的，url直接填conf的域名地址即可
• confluence不允许直接ip访问，才需要session注入的方式，否则直接ip加端口访问即可‘

可以用下面的命令验证:

curl -v -H "Host: conf2.xxx.com" -u "user:pwd" http://ip:port/pages/viewpage.action?pageId=xxx

从域名查看ip和端口 可以用 :

curl -v 域名地址