Render and save a PDF

The single most common thing you’ll do with Pagr: take a published template, feed it a JSON object, and end up with a PDF on disk. This guide walks the whole path, including the two decisions that trip people up — whether to get the bytes inline, and whether to store the result.

Before you start

You need three things:

• A published template version. A template with nothing published renders as 404 NoPublishedVersion — see Versions & publishing.
• The template’s id — copy it from the template list or the editor URL.
• An API key. Use a pagr_test_ key while you build: output is watermarked and consumes no credit. See API Keys.

How it works

1. Construct a client with your API key.

   The base URL defaults to the hosted Pagr API, so the key is the only required argument. The key’s prefix decides test vs. production — there is no separate flag.

2. Call render with the template id and your data.

   Your data is a plain object whose keys match the template’s bindings. Not sure what those are? Fetch the version’s sampleData — it matches the bindings by construction. See Templates & versions.

3. Ask for the bytes inline with includeDocument.

   By default the response carries metadata only and the PDF stays server-side. Pass includeDocument: true and the PDF rides along as Base64, which the SDK decodes for you. Without it, save() has nothing to write.

4. Check result.ok before touching result.document.

   A document that fails validation is a normal outcome, not an exception: result.ok is false, result.document is null, and result.issues explains why. Only protocol failures raise.

5. Save it.

   Pass a directory and the SDK names the file from documentName, appending .pdf. Pass a full path and it writes exactly there.

Full example

cURL

curl -X POST "https://pagr-prd-api-public.azurewebsites.net/v1/render/8bec66ff-6f3d-4c1e-9a2b-1f0e5d7c4a90" \
  -H "Authorization: Bearer pagr_test_xxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "documents": [
      { "Title": "Acme Q3 Invoice", "Amount": 42 }
    ],
    "includeDocument": true
  }' | jq -r '.documents[0].documentBase64' | base64 -d > invoice.pdf

Raw HTTP gives you Base64 in a JSON field — you decode it yourself. To skip that step entirely, send Accept: application/pdf instead and write the body straight to a file: see Render a raw PDF.

Python

import asyncio
from pagr import PagrApiClient, PagrError

TEMPLATE_ID = "8bec66ff-6f3d-4c1e-9a2b-1f0e5d7c4a90"

async def main():
    async with PagrApiClient("pagr_test_xxxxxxxx") as client:
        try:
            result = await client.render(
                TEMPLATE_ID,
                {"Title": "Acme Q3 Invoice", "Amount": 42},
                include_document=True,          # ← without this, save() has nothing to write
            )
        except PagrError as exc:                # protocol failures only
            print("Request failed:", exc)
            return

        if not result.ok:                       # business outcome
            print(result.status, result.message)
            for issue in result.issues:
                print(" ", issue)
            return

        doc = result.document
        print(f"{doc.document_name} — {doc.page_count} page(s), {doc.file_size_bytes} bytes")
        path = doc.save("out/")                 # existing directory → out/<name>.pdf
        print("Wrote", path)

asyncio.run(main())

TypeScript

import { PagrApiClient, PagrError } from 'pagr';

const TEMPLATE_ID = '8bec66ff-6f3d-4c1e-9a2b-1f0e5d7c4a90';

const client = new PagrApiClient('pagr_test_xxxxxxxx');

try {
  const result = await client.render(
    TEMPLATE_ID,
    { Title: 'Acme Q3 Invoice', Amount: 42 },
    { includeDocument: true },            // ← without this, save() has nothing to write
  );

  if (!result.ok) {                       // business outcome
    console.log(result.status, result.message);
    for (const issue of result.issues) console.log(' ', issue.description);
  } else {
    const doc = result.document!;
    console.log(`${doc.documentName} — ${doc.pageCount} page(s), ${doc.fileSizeBytes} bytes`);
    console.log('Wrote', await doc.save('./out'));
  }
} catch (err) {                           // protocol failures only
  if (err instanceof PagrError) console.log('Request failed:', err.message);
  else throw err;
}

Node.js only — save() uses node:fs/promises, so it won’t run in a browser bundle.

Java

import org.example.PagrApiClient;
import org.example.RenderOptions;
import org.example.exception.PagrException;
import org.example.models.RenderResult;
import java.nio.file.Path;
import java.util.Map;
import java.util.UUID;

UUID templateId = UUID.fromString("8bec66ff-6f3d-4c1e-9a2b-1f0e5d7c4a90");

try (PagrApiClient client = new PagrApiClient("pagr_test_xxxxxxxx")) {

    RenderResult result = client.render(
        templateId,
        Map.of("Title", "Acme Q3 Invoice", "Amount", 42),
        RenderOptions.builder().includeDocument(true).build());   // ← required to save

    if (!result.isOk()) {                                          // business outcome
        System.out.println(result.getStatus() + " " + result.getMessage());
        result.getIssues().forEach(i -> System.out.println("  " + i));
    } else {
        var doc = result.getDocument();
        System.out.printf("%s — %d page(s), %d bytes%n",
            doc.getDocumentName(), doc.getPageCount(), doc.getFileSizeBytes());
        System.out.println("Wrote " + doc.save(Path.of("out")));
    }
} catch (PagrException exc) {                                      // protocol failures only
    System.out.println("Request failed: " + exc.getMessage());
}

A separate PagrAsyncApiClient mirrors every call, returning CompletableFuture<T>.

C#

using Pagr.Sdk;
using Pagr.Sdk.Exceptions;

var templateId = Guid.Parse("8bec66ff-6f3d-4c1e-9a2b-1f0e5d7c4a90");

// The client owns a pooled HttpClient — create one and reuse it.
using var client = new PagrApiClient("pagr_test_xxxxxxxx");

try
{
    var result = await client.RenderAsync(
        templateId,
        new { Title = "Acme Q3 Invoice", Amount = 42 },
        includeDocument: true);              // ← without this, SaveAsync has nothing to write

    if (!result.Ok)                          // business outcome
    {
        Console.WriteLine($"{result.Status} {result.Message}");
        foreach (var issue in result.Issues)
            Console.WriteLine($"  {issue}");
    }
    else
    {
        var doc = result.Document!;
        Console.WriteLine($"{doc.DocumentName} — {doc.PageCount} page(s), {doc.FileSizeBytes} bytes");
        Console.WriteLine($"Wrote {await doc.SaveAsync("out")}");
    }
}
catch (PagrApiException exc)                 // protocol failures only
{
    Console.WriteLine($"Request failed: {exc.Message}");
}

Ruby

require "pagr"

TEMPLATE_ID = "8bec66ff-6f3d-4c1e-9a2b-1f0e5d7c4a90"

client = Pagr::Client.new("pagr_test_xxxxxxxx")

begin
  result = client.render(
    TEMPLATE_ID,
    { "Title" => "Acme Q3 Invoice", "Amount" => 42 },
    include_document: true,             # ← without this, save has nothing to write
  )

  if result.ok?
    doc = result.document
    puts "#{doc.document_name} — #{doc.page_count} page(s), #{doc.file_size_bytes} bytes"
    puts "Wrote #{doc.save('out')}"     # existing directory → out/<name>.pdf
  else
    puts "#{result.status} #{result.message}"
    result.issues.each { |issue| puts "  #{issue}" }
  end
rescue Pagr::Error => e                 # protocol failures only
  warn "Request failed: #{e.message}"
end

C++

#include <iostream>
#include "pagr/PagrApiClient.hpp"
#include "pagr/exceptions.hpp"

const std::string kTemplateId = "8bec66ff-6f3d-4c1e-9a2b-1f0e5d7c4a90";

pagr::PagrApiClient client("pagr_test_xxxxxxxx");

try {
    const auto result = client.render(
        kTemplateId,
        R"({"Title": "Acme Q3 Invoice", "Amount": 42})",
        {.include_document = true});         // ← without this, save() has nothing to write

    if (!result.ok()) {                      // business outcome
        std::cout << result.status << "\n";
        for (const auto& issue : result.issues) {
            std::cout << "  " << issue.description << "\n";
        }
    } else {
        const auto& doc = *result.document;
        std::cout << doc.document_name << " — " << doc.page_count << " page(s)\n";
        doc.save("out/" + doc.document_name + ".pdf");
    }
} catch (const pagr::PagrApiException& exc) {  // protocol failures only
    std::cout << "Request failed: " << exc.what() << "\n";
}

Every call also has a _async form (render_async) returning std::future<T>.

Variation: don’t store the document

Set persist to false when the PDF is transient — a download you stream straight to a user, a preview you throw away. Nothing is stored: the render doesn’t appear in Renders, and id / viewUrl come back null.

cURL

curl -X POST ".../v1/render/{templateId}?persist=false" \
  -H "Authorization: Bearer pagr_test_xxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{ "documents": [ { "Title": "Preview" } ] }'

Python

result = await client.render(TEMPLATE_ID, data, persist=False)
# include_document is unnecessary — the bytes are forced inline
pdf_bytes = result.document.to_bytes()
assert result.document.id is None and result.document.view_url is None

TypeScript

const result = await client.render(TEMPLATE_ID, data, { persist: false });
const pdfBytes = result.document!.toBytes();
// result.document.id and .viewUrl are null — nothing was stored

Java

RenderResult result = client.render(templateId, data,
    RenderOptions.builder().persist(false).build());
byte[] pdf = result.getDocument().toBytes();

C#

var result = await client.RenderAsync(templateId, data, persist: false);
byte[] pdf = result.Document!.ToBytes();

Ruby

result = client.render(TEMPLATE_ID, data, persist: false)
pdf = result.document.to_bytes

C++

const auto result = client.render(kTemplateId, json_data, {.persist = false});
const auto pdf = result.document->to_bytes();

persist=false forces the bytes inline

It has to: with nothing stored, the inline copy is the only copy. So to_bytes() / save() work whether or not you passed includeDocument. This is the one case where the flag is redundant.

Variation: render a specific version

Rendering defaults to the latest published version, so template updates will immediately affect your output. When strict reproducibility matters (e.g., generating exact copies of past invoices), specify the exact version number in your request.

cURL

curl -X POST ".../v1/render/{templateId}/versions/3" …

Python

result = await client.render(TEMPLATE_ID, data, version=3)

TypeScript

const result = await client.render(TEMPLATE_ID, data, { version: 3 });

Java

RenderResult result = client.render(templateId, data,
    RenderOptions.builder().version(3).build());

C#

var result = await client.RenderAsync(templateId, data, version: 3);

Ruby

result = client.render(TEMPLATE_ID, data, version: 3)

C++

const auto result = client.render(kTemplateId, json_data, {.version = 3});

Other considerations

• documentName is data, not a path. It’s generated from the version’s document-name template, so it can embed values bound from your payload — including slashes and dots. TODO: how is it sanitized??
• When .pdf gets appended differs slightly by SDK. TODO: SHOULD BE CHANGED; Python, TypeScript, Ruby and C++ check the .pdf suffix, so Invoice 2024.10 becomes Invoice 2024.10.pdf. C# and Java check for any extension / dot, so the same name is left as Invoice 2024.10. Pass an explicit full path when the exact filename matters.
• Render output is always PDF. documentName carries no extension because there’s no other format to distinguish.
• A slow template needs a bigger timeout, not a retry. The client default is 30 seconds; a document may legitimately take up to the server’s 60-second budget. Pass a per-call timeout rather than raising the client-wide default — see Configure the client. Writes are never retried, so a timed-out render must not be blindly re-sent: it may have rendered and charged already.
• Warnings block production but not test. A payload that renders on your test key can be rejected on a production key, because production also blocks on Warning-severity issues. See Validate before rendering.
• Reuse the client. In C# it owns a pooled HttpClient; in Python it owns an httpx connection pool. Create one per process, not one per render.

Related articles

• Render a document — the field-level reference for this endpoint.
• Render a raw PDF — skip Base64 entirely.
• Validate before rendering — catch data problems for free.
• Render a batch — more than one document per request.
• Handle errors and retries — what raises, what doesn’t.