Certificate PDFs in Go

Why certificates become PDFs

A certificate is meant to be shown to a third party who has no relationship with the issuing system, and often printed. A link that could rot is not a credential. That is the difference between a document and a view of a database, and it is why training providers, course platforms and event organisers keep asking for this.

What a certificate has to carry

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

• the recipient full name exactly as they gave it
• the achievement, named precisely enough to be checked
• the issue date, and an expiry where the credential lapses
• a unique certificate ID that a verifier can look up
• the issuing body and an authorised signature or seal

The Go implementation

net/http ships with the runtime, so this adds no dependency. The render happens in an http.HandlerFunc that copies the body through, and you io.Copy from the response body to the ResponseWriter.

func CertificatePDF(w http.ResponseWriter, r *http.Request) {
	html, err := renderCertificate(r.PathValue("id"))
	if err != nil {
		http.Error(w, err.Error(), http.StatusInternalServerError)
		return
	}

	payload, _ := json.Marshal(map[string]any{
		"html": html,
		"options": map[string]any{
			"format":          "A4",
			"printBackground": true,
		},
	})

	req, _ := http.NewRequestWithContext(r.Context(), "POST",
		"https://api.pdfpipe.xyz/v1/pdf", bytes.NewReader(payload))
	req.Header.Set("Authorization", "Bearer "+os.Getenv("PDFPIPE_KEY"))
	req.Header.Set("Content-Type", "application/json")

	resp, err := http.DefaultClient.Do(req)
	if err != nil {
		http.Error(w, err.Error(), http.StatusBadGateway)
		return
	}
	defer resp.Body.Close() // leaks a connection per request without this

	w.Header().Set("Content-Type", "application/pdf")
	io.Copy(w, resp.Body)
}

The CSS that makes a certificate page correctly

The layout problem specific to this document is that the design is fixed and usually landscape, so it must fit exactly one page regardless of how long the recipient's name is. These rules handle it.

/* Exactly one landscape page, whatever the name's length. */
@page {
  size: A4 landscape;
  margin: 0;
}

.certificate {
  width: 297mm;
  height: 210mm;
  break-after: avoid;
  display: grid;
  place-items: center;
}

/* Long names shrink rather than wrap onto a second line. */
.recipient {
  font-size: clamp(28px, 5vw, 54px);
  text-wrap: balance;
}

What goes wrong in Go

Leaving resp.Body unclosed leaks a connection per request, and the leak only shows up as slow exhaustion under sustained traffic rather than as an error you can catch in testing.

What people try first

Most Go projects reach for chromedp, which drives a real Chrome you then have to install, supervise and keep from accumulating zombie processes. 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 certificate lives afterwards

Rendering is the short part. The recipient keeps it indefinitely and may present it years later to an employer who will want to verify it. That is the argument for the ID field: the PDF is the artefact, but the ID is what makes it checkable after your system has changed.

Getting the document right

• Check the recipient's name, which must render correctly for every alphabet your learners actually use.
• Generation is triggered by a course or assessment being completed, 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

Related

Other Go documents, and the same certificate in other stacks.