Python

HTML to PDF in Python

One POST, a PDF back. No headless browser to install, no Chromium to keep patched, no fonts to install on the box — the render happens on a warm browser that is already running.

With the official client

There is a published Python client, so the shortest version is two lines. It has no dependencies of its own, retries 429s and 5xxs for you, and turns the API’s error envelope into a typed error you can branch on. The raw HTTP below still works and is still supported — plenty of people would rather send a POST than take a dependency.

pip install pdfcraft

from pdfcraft import PDFCraft

client = PDFCraft(os.environ["PDFCRAFT_API_KEY"])
pdf = client.render(html="<h1>Invoice 1042</h1>")
open("invoice.pdf", "wb").write(pdf)

data = client.extract_pdf(open("statement.pdf", "rb").read())
print(data["tables"][0]["header"])

On PyPI.

Render HTML to a PDF with plain HTTP

import os, requests

response = requests.post(
    "https://api.pdfcraft.dev/v1/render",
    headers={"Authorization": f"Bearer {os.environ['PDFCRAFT_API_KEY']}"},
    json={
        "html": "<h1>Invoice 1042</h1>",
        "options": {"printBackground": True, "margin": {"top": "20mm", "bottom": "20mm"}},
    },
    timeout=60,
)
response.raise_for_status()

with open("invoice.pdf", "wb") as handle:
    handle.write(response.content)

Read a PDF back as JSON

The same key works for extraction. Send a PDF, name the fields you want by the label printed on the page, and get them back with the raw text, a coerced value, a confidence and a bounding box.

import base64, os, requests

with open("statement.pdf", "rb") as handle:
    encoded = base64.b64encode(handle.read()).decode()

response = requests.post(
    "https://api.pdfcraft.dev/v1/extract",
    headers={"Authorization": f"Bearer {os.environ['PDFCRAFT_API_KEY']}"},
    json={
        "file": encoded,
        "schema": {"account_number": "string", "closing_balance": "currency"},
        "options": {"rows_as_objects": True},
    },
    timeout=60,
)
body = response.json()

print(body["fields"]["closing_balance"]["value"])
for row in body["tables"][0]["rows_as_objects"]:
    print(row)

What bites in Python

`response.content`, not `response.text`. `.text` decodes the bytes as a string and the PDF is destroyed — this is the single most common way a Python integration produces a file that will not open.
Pass `timeout=`. requests has no default timeout at all, so a hung connection hangs your worker indefinitely rather than raising.
For async use httpx with the same shape; the API has no streaming requirement, so the sync client is fine for most jobs.

Errors are one shape

Every failure is {"error":{"code","message","docs_url"}} with a stable code, so you can switch on error.code rather than parsing prose. The two worth handling explicitly are rate_limited — honour Retry-After — and render_failed, which means your HTML broke rather than ours did.

The same thing in another language

Or try it with no code at all in the playground. A free key is 100 renders a month.