pdfjetdev 0.1.1

Python client for the PDFjet API (pdfjet.dev): turn PDFs into Markdown, CSV, structured JSON (with provenance), Word, and searchable PDFs — with LangChain and LlamaIndex loaders for RAG.

pdfjetdev — Python client for PDFjet

Turn any PDF into Markdown, CSV, structured JSON (with provenance), editable Word, or a searchable PDF with a few lines of Python. Wraps the PDFjet API. Files are never stored.

• Zero dependencies — the core client uses only the Python standard library.
• Structured JSON with provenance — element types, text, page numbers, and per-element bounding boxes.
• RAG-ready — LangChain and LlamaIndex loaders included.

Install

pip install pdfjetdev
# with framework loaders:
pip install "pdfjetdev[langchain]"
pip install "pdfjetdev[llamaindex]"

Quick start

from pdfjetdev import Client

client = Client("pj_live_...")          # or set the PDFJET_API_KEY env var

# Structured parse with bounding-box provenance
doc = client.extract_json("invoice.pdf")
print(doc.page_count, "pages, scanned:", doc.scanned)
for block in doc.pages[0].blocks:
    print(block.type, block.bbox, block.text or block.items or "")

# Chunk-ready Markdown for a RAG pipeline
print(doc.to_markdown())

# Other outputs
md   = client.extract_markdown("invoice.pdf")   # str
csv  = client.extract_csv("invoice.pdf")        # str
docx = client.extract_docx("invoice.pdf")       # bytes
pdf  = client.extract_searchable("scan.pdf")    # bytes (searchable PDF)

file can be a path, bytes, or an open binary file. Requests to /extract/* and the utility endpoints automatically retry on 429/5xx with exponential backoff.

Pass provenance=False to extract_json for a leaner text+structure payload (no bounding boxes).

LangChain

from pdfjetdev.integrations.langchain import PDFJetLoader

loader = PDFJetLoader("report.pdf", api_key="pj_live_...")   # one Document per page
docs = loader.load()   # page_content is layout-preserving Markdown; page number in metadata

LlamaIndex

from pdfjetdev.integrations.llamaindex import PDFJetReader

reader = PDFJetReader(api_key="pj_live_...")
docs = reader.load_data("report.pdf")   # one LlamaIndex Document per page

Errors

All errors subclass pdfjetdev.PDFJetError: AuthenticationError (401/403), PaymentRequiredError (402, quota), InvalidInputError (400/413), RateLimitError (429), ServiceUnavailableError (5xx).

License

MIT · pdfjet.dev