Source code for langchain_community.document_loaders.firecrawl

from typing import Iterator, Literal, Optional

from langchain_core.document_loaders import BaseLoader
from langchain_core.documents import Document
from langchain_core.utils import get_from_env


class FireCrawlLoader(BaseLoader):
    """使用FireCrawl将网页加载为文档。

    必须安装Python包`firecrawl`并拥有FireCrawl API密钥。请参见
        https://www.firecrawl.dev/ 了解更多。"""

    def __init__(
        self,
        url: str,
        *,
        api_key: Optional[str] = None,
        mode: Literal["crawl", "scrape"] = "crawl",
        params: Optional[dict] = None,
    ):
        """使用API密钥和URL进行初始化。

参数：
    url: 要爬取的URL。
    api_key: Firecrawl API密钥。如果未指定，将从环境变量FIREWALL_API_KEY中读取。获取API密钥
    mode: 加载程序运行的模式。默认为“crawl”。
         选项包括“scrape”（单个URL）和
         “crawl”（所有可访问的子页面）。
    params: 传递给Firecrawl API的参数。
        示例包括crawlerOptions。
        有关更多详细信息，请访问：https://github.com/mendableai/firecrawl-py
"""

        try:
            from firecrawl import FirecrawlApp
        except ImportError:
            raise ImportError(
                "`firecrawl` package not found, please run `pip install firecrawl-py`"
            )
        if mode not in ("crawl", "scrape"):
            raise ValueError(
                f"Unrecognized mode '{mode}'. Expected one of 'crawl', 'scrape'."
            )
        api_key = api_key or get_from_env("api_key", "FIREWALL_API_KEY")
        self.firecrawl = FirecrawlApp(api_key=api_key)
        self.url = url
        self.mode = mode
        self.params = params

    def lazy_load(self) -> Iterator[Document]:
        if self.mode == "scrape":
            firecrawl_docs = [self.firecrawl.scrape_url(self.url, params=self.params)]
        elif self.mode == "crawl":
            firecrawl_docs = self.firecrawl.crawl_url(self.url, params=self.params)
        else:
            raise ValueError(
                f"Unrecognized mode '{self.mode}'. Expected one of 'crawl', 'scrape'."
            )
        for doc in firecrawl_docs:
            yield Document(
                page_content=doc.get("markdown", ""),
                metadata=doc.get("metadata", {}),
            )