PDFPipe

FastAPI / receipts

Receipt PDFs in FastAPI

Post your receipt HTML to a render API from an async endpoint returning StreamingResponse, then return StreamingResponse over the httpx byte iterator. No browser binary in your FastAPI deployment.

Why receipts become PDFs

A receipt is proof a payment happened, and customers attach it to expense claims. It has to survive being forwarded, printed and filed by someone who never had an account with you. That is the difference between a document and a view of a database, and it is why ecommerce and point-of-sale teams keep asking for this.

What a receipt has to carry

Before any of the code below matters, the template has to produce a document that is actually a receipt. These are the fields that make it one, and the ones a reader or an auditor will look for first.

  • the payment reference from the processor, which is what a dispute is traced by
  • the amount actually captured, which is not always the amount ordered
  • the payment method, usually the card brand and last four digits
  • the timestamp with a timezone, since expense systems reconcile by date
  • the merchant legal name, which may differ from the brand on the site

Turning your Jinja2Templates template into a document

You already have the markup. Templates.get_template("invoice.html").render(context) gives you it as a string, which is the only input the render needs. TemplateResponse is for returning HTML to a browser. When the HTML is going to a renderer you want the string, so call the underlying Jinja template directly rather than unwrapping a Response you never send.

The FastAPI implementation

httpx is the client most projects here already have. The render happens in an async endpoint returning StreamingResponse, and you return StreamingResponse over the httpx byte iterator.

python
# routers/receipts.py
import httpx
from fastapi import APIRouter, Depends
from fastapi.responses import StreamingResponse
from fastapi.templating import Jinja2Templates

router = APIRouter()
templates = Jinja2Templates(directory="templates")


@router.get("/receipts/{receipt_id}.pdf")
async def receipt_pdf(receipt_id: int, db=Depends(get_db)):
    receipt = await db.get_receipt(receipt_id)

    # The template directly, not TemplateResponse: we want the string, and
    # TemplateResponse is for HTML you are actually sending to a browser.
    html = templates.get_template("receipt.html").render(receipt=receipt)

    client = httpx.AsyncClient(timeout=60.0)
    request = client.build_request(
        "POST",
        "https://api.pdfpipe.xyz/v1/pdf",
        headers={"Authorization": f"Bearer {settings.pdfpipe_key}"},
        json={"html": html, "options": {"format": "A4", "printBackground": True}},
    )
    upstream = await client.send(request, stream=True)
    upstream.raise_for_status()

    return StreamingResponse(
        upstream.aiter_bytes(),
        media_type="application/pdf",
        background=BackgroundTask(upstream.aclose),
    )

The CSS that makes a receipt page correctly

The layout problem specific to this document is that receipts are short and usually want a narrow page, so the default A4 canvas leaves an expanse of white unless the page size is set deliberately. These rules handle it.

css
/* A receipt on A4 is mostly white space. Set the page to the
   content, and let it grow only as far as the items require. */
@page {
  size: 80mm 200mm;
  margin: 6mm;
}

body { font-size: 11px; }

What goes wrong in FastAPI

Calling a synchronous HTTP client inside an async def endpoint blocks the event loop for the whole render, which stalls every other request on that worker. Either use httpx.AsyncClient, or declare the endpoint def rather than async def so it runs in the threadpool.

What people try first

Most FastAPI projects reach for WeasyPrint called from a background task, which moves the memory cost off the request but not out of the process. That works until it is running on more than one machine, at which point the browser becomes the thing you operate rather than the thing you use.

Where the receipt lives afterwards

Rendering is the short part. The customer keeps it, not you, and they submit it to an expense system months later. It has to be readable by software that never saw your site, which is why the amount and reference belong in selectable text rather than inside an image.

Getting the document right

  • Check the amount paid and the payment reference, which is what an expense system matches against.
  • Generation is triggered by a successful payment webhook, so size the timeout for that path rather than for a health check.
  • Volume arrives in bursts, so queue the render rather than doing it inside the request that triggered it.
  • Backgrounds are painted by default here, so a design that uses colour needs nothing set. Only an explicit print_background of false turns them off.

Frequently asked

Do I need Chromium installed to generate receipts from FastAPI?

No. The render happens over HTTP, so your FastAPI deployment stays the size it is now. That is the main reason to use an API rather than WeasyPrint called from a background task, which moves the memory cost off the request but not out of the process.

How do I stop a long receipt using all the memory?

Return StreamingResponse over the httpx byte iterator. Calling a synchronous HTTP client inside an async def endpoint blocks the event loop for the whole render, which stalls every other request on that worker. Either use httpx.AsyncClient, or declare the endpoint def rather than async def so it runs in the threadpool.

What has to be on a receipt?

At minimum: the payment reference from the processor, which is what a dispute is traced by; the amount actually captured, which is not always the amount ordered; the payment method, usually the card brand and last four digits. The one to get right before anything else is the amount paid and the payment reference, which is what an expense system matches against.

Do I need to store the generated receipts?

Depends on the document, and this one has a clear answer: the customer keeps it, not you, and they submit it to an expense system months later. It has to be readable by software that never saw your site, which is why the amount and reference belong in selectable text rather than inside an image.

Can I keep my existing receipt template?

Yes, if it produces HTML. Whatever renders your receipt view today can render the same markup for the PDF, which is why the CSS above is the only new thing you write.

Related

Other FastAPI documents, and the same receipt in other stacks.

100 free documents a month, no card.