Class PlasmateLoader

scrapegraphai/docloaders/plasmate.py:46–203

Fetches pages using Plasmate — a lightweight Rust browser engine that outputs Structured Object Model (SOM) instead of raw HTML. Advantages over ChromiumLoader for static / server-rendered pages: - No Chrome/Playwright required — single binary, installs via pip - ~64MB RAM per sessi

class PlasmateLoader(BaseLoader):
    """Fetches pages using Plasmate — a lightweight Rust browser engine that outputs
    Structured Object Model (SOM) instead of raw HTML.

    Advantages over ChromiumLoader for static / server-rendered pages:
    - No Chrome/Playwright required — single binary, installs via pip
    - ~64MB RAM per session vs ~300MB for Chromium
    - 10-100x fewer tokens per page (SOM strips nav, ads, boilerplate)
    - Drops into existing ScrapeGraphAI workflows with minimal config changes

    For SPAs or pages that require JavaScript rendering, set ``fallback_to_chrome=True``
    to automatically retry with ChromiumLoader on empty or error responses.

    Attributes:
        urls: List of URLs to fetch.
        output_format: Plasmate output format — ``"text"`` (default, most compatible),
            ``"som"`` (full JSON), or ``"markdown"``.
        timeout: Per-request timeout in seconds. Defaults to 30.
        selector: Optional ARIA role or CSS id selector to scope extraction
            (e.g. ``"main"`` or ``"#content"``).
        extra_headers: Optional dict of HTTP headers to pass to each request.
        fallback_to_chrome: If True, retry with ChromiumLoader when Plasmate
            returns empty content (useful for JS-heavy SPAs). Defaults to False.
        chrome_kwargs: Extra kwargs forwarded to ChromiumLoader when fallback is used.

    Example::

        from scrapegraphai.docloaders import PlasmateLoader

        loader = PlasmateLoader(
            urls=["https://docs.python.org/3/library/json.html"],
            output_format="text",
            timeout=30,
        )
        docs = loader.load()
        print(docs[0].page_content[:500])
    """

    def __init__(
        self,
        urls: List[str],
        *,
        output_format: str = "text",
        timeout: int = 30,
        selector: Optional[str] = None,
        extra_headers: Optional[dict] = None,
        fallback_to_chrome: bool = False,
        **chrome_kwargs,
    ):
        if output_format not in ("som", "text", "markdown", "links"):
            raise ValueError(
                f"output_format must be one of 'som', 'text', 'markdown', 'links'; got {output_format!r}"
            )
        self.urls = urls
        self.output_format = output_format
        self.timeout = timeout
        self.selector = selector
        self.extra_headers = extra_headers or {}