langchain-superpdf

superpdf.ai · Docs · GitHub

The LangChain document loader for superPDF AI: local PDFs as Markdown, page by page, ready for RAG. Headings stay Markdown headings, tables stay tables, charts drawn in the PDF come out as tables of their values, and scanned pages are read with OCR in 15 languages. Everything runs on your computer; nothing is uploaded.

Install

pip install -U langchain-superpdf

Python 3.10 or later on macOS, Linux or Windows. Node.js comes with the package; the first load installs the superPDF reader (about 220 MB) into your cache folder, once.

Use

from langchain_superpdf import SuperPDFLoader

docs = SuperPDFLoader("report.pdf").load()  # one Document per page
docs[0].page_content  # "# Quarterly report\n\nRevenue grew in every region…"
docs[0].metadata
# {"source": "report.pdf", "total_pages": 4, "creationdate": "", "creator": "…",
#  "producer": "…", "page": 0, "page_label": "1"}

page counts from 0 and page_label from 1, like LangChain's other PDF loaders.

Several files at once: SuperPDFLoader(["a.pdf", "b.pdf"]). A folder:

from pathlib import Path

docs = SuperPDFLoader(sorted(Path("papers").glob("**/*.pdf"))).load()

PDFs that arrive as bytes, such as uploads, go through the parser:

from langchain_core.documents.base import Blob
from langchain_superpdf import SuperPDFParser

blob = Blob.from_data(pdf_bytes, path="upload.pdf")
docs = list(SuperPDFParser().lazy_parse(blob))

Split by section

from langchain_text_splitters import MarkdownHeaderTextSplitter

splitter = MarkdownHeaderTextSplitter([("#", "title"), ("##", "section")])
chunks = []
for doc in docs:
    for chunk in splitter.split_text(doc.page_content):
        chunk.metadata.update(doc.metadata)  # keep source and page for citations
        chunks.append(chunk)

Agents (LangGraph)

For agents, use superPDF AI's MCP server: its 30 PDF tools (read, tables to Excel, Word export, merge, fill forms, compare and more) load as LangChain tools through langchain-mcp-adapters and run in any LangGraph agent. The tools can only use the folders you pass with --allow-dir.

from langchain.agents import create_agent
from langchain_mcp_adapters.client import MultiServerMCPClient

client = MultiServerMCPClient(
    {
        "superpdf": {
            "command": "npx",
            "args": [
                "-y",
                "@superpdf/cli",
                "mcp",
                "--allow-dir",
                "/Users/you/Documents/superPDF",
            ],
            "transport": "stdio",
        }
    }
)
tools = await client.get_tools()
agent = create_agent("anthropic:claude-sonnet-5-5", tools)

More, including search with page citations: superpdf.ai/docs/python/langchain-llamaindex.

License

Apache-2.0. Part of superPDF AI.