The FynePDF API for Python: compress, convert, OCR, redact, merge, split and generate PDFs, and read them with AI. Typed from the API's own spec.
pip install fynepdffrom fynepdf import FynePDF
fyne = FynePDF() # reads FYNEPDF_API_KEY
file = fyne.files.upload("report.pdf")
job = fyne.pdf.compress(file_id=file["id"], target_size_bytes=5_000_000)
print(job["receipt"]["steps"][0]) # what it did and cost
fyne.files.download_to(job["outputs"][0], "report-small.pdf")Get a key at fynepdf.com/developers. Test keys are free (results are watermarked); live keys get 250 free credits a month.
All 34 actions are methods on fyne.pdf, with their parameters as keyword arguments your editor completes:
fyne.pdf.merge(file_ids=[a["id"], b["id"]])
fyne.pdf.ocr(file_id=scan["id"])
fyne.pdf.pdf_to_excel(file_id=statement["id"], mode="statement")
fyne.pdf.redact(file_id=contract["id"], text=["Jane Doe", "555-0142"])
fyne.pdf.html_to_pdf(url="https://example.com")
fyne.pdf.create_pdf(
template="invoice",
data={"invoice_number": "INV-104", "currency": "USD",
"line_items": [{"description": "Consulting", "quantity": 3, "unit_price": 120}], "tax_rate": 10},
)Each waits up to 30 seconds for the result. Pass wait=0 to get the job at once, then fyne.jobs.wait(job), or give a webhook_url.
steps = [
{"action": "merge"},
{"action": "compress", "params": {"target_size_bytes": 5_000_000}},
{"action": "page_numbers", "params": {"position": "bottom-right"}},
]
quote = fyne.jobs.quote(inputs=[a["id"], b["id"]], steps=steps) # the price, nothing run
job = fyne.jobs.wait(fyne.jobs.create(inputs=[a["id"], b["id"]], steps=steps))files.upload(path_or_bytes_or_file, filename=...): in the request body up to 25 MB, through an upload URL above that.files.from_url(url): the API fetches a public link.files.download(file_or_id)returns bytes;files.download_to(file_or_id, path)writes them. Your key is never sent to storage.
answer = fyne.ai.ask(file_id=file["id"], question="What's the notice period?")
invoice = fyne.ai.extract(file_id=file["id"], type="invoice")Errors are FynePDFError with status, a code to branch on (insufficient_credits, validation_failed, rate_limited…), detail and request_id. Rate limits, timeouts and server errors are retried (twice by default, honouring Retry-After), and every POST carries an Idempotency-Key, so a retry never runs or charges twice.
from fynepdf import verify_webhook
event = verify_webhook(request.body, request.headers, os.environ["FYNEPDF_WEBHOOK_SECRET"])
if event["type"] == "job.succeeded":
print(event["data"]["outputs"][0]["download_url"])FynePDF is also an MCP server for Claude, ChatGPT, Cursor and others: fynepdf.com/developers/mcp.
MIT