Building a PDF Generator on AWS Lambda with Python3 and wkhtmltopdf
June 24, 2019    |    APIs    |    AWS    |    Serverless

Update 07/18/2019:
Source Code:

UPDATE 06/28/2019:
For anyone attempting to follow this please read a follow-up post about a font bug on the generated PDFs:


Creating scalable APIs in 2019 is easier than ever before with serverless auto-scaling compute power being widely accessible. In this article, I’ll show you how I created a PDF generator API that can handle 1000 concurrent requests.

The purpose of this article is to show you how I accomplished creating the API so that you will see how easy it is to create your own serverless APIs. When this project is complete, you will have an API endpoint in which you can POST a JSON object to, and receive a link to a PDF on Amazon S3. The JSON object looks like this:

  "filename": "sample.pdf",
  "html": "<html><head></head><body><h1>It works! This is the default PDF.</h1></body></html>"


What you’ll need:

  • Python3
  • Aws CLI installed and configured
  • Serverless installed


The first thing you need to do is initialize a new Serverless project. Serverless is a tool that greatly simplifies using AWS, GCloud, and Azure services so you can create APIs without worrying about managing a server. If you need more information about Serverless visite their website: In your terminal run the following command to initialize your project.

Note: sls is short for serverless and can be used interchangeably.

sls create --template aws-python3

This command will bootstrap a Python3 Lambda setup for you to work from.


Next, we need to get the binary for turning HTML into PDFs. For this I used WKHTMLTOPDF.

Important: You have to use version 0.12.4 because later versions of wkhtmltopdf require dynamic dependencies of the host system that cannot be installed on AWS Lambda.

Download version 0.12.4 here:

Once you have extracted this tar file, copy the binary wkhtmltopdf to the binary folder in your project.

cp wkhtmltox/bin/wkhtmltopdf ./binary/wkhtmltopdf

More information about wkhtmltopdf can be found on their website:

Python3 Dependencies

Now that we have the WKHTMLtoPDF binary, we need the Python library, pdfkit to use it. Since we are using Serverless and AWS Lambda we cannot just run pip install pdfkit. We need a Serverless plugin to install our dependencies on Lambda.

In our project folder install the python plugin requirements module for Serverless.

sls plugin install -n serverless-python-requirements

Now, in your serverless.yml file, you need to add a custom section in the yaml:

    dockerizePip: true

Once the serverless plugin requirements is installed, you can add a requirements.txt file to your project and it will be automatically installed on lambda when you deploy.

Your requirements.txt for this project only needs to have pdfkit.



For any issues with this module checkout the repository issue:
Serverless Python Requirements

Ready, Set, Code

We are ready to code now. We only need to work with two files for this project: serverless.yml ad Typically you want your serverless projects to be lightweight and independent of other parts of your codebase.


The serverless.yml file is our main configuration file. A serverless.yml file is broken down into a few important sections. If you are new to YAML, checkout the documentation for YAML.

For this serverless.yml configuration, you’ll need to create a file called config.yml. This will store the S3 bucket name. The serverless.yml will reference the config.yml to setup the correct bucket for your project.

Contents of config.yml

BucketName: 'your-s3-bucket-name'

Here is a high-level overview of a serverless.yml file:

service: pdf-services # name or reference to our project
provider: # It is possible to use Azure, GCloud, or AWS
functions: # Array of functions to deploy as Lambdas
resources: # S3 buckets, DynamoDB tables, and other possible resources to create
plugins: # Plugins for Serverless
custom: # Custom variables used by you or plugins during setup and deployment

Our serverless configuration will do a few things for us when we deploy:

  1. Create an S3 bucket called pdf-service-bucket to store our PDFs
  2. Create a function that will create the PDFs
  3. Give our function access to the S3 bucket
  4. Setup an API endpoint for our Lambda function at:

Here is the full serverless.yml configuration. I’ve added a couple important comments in the code.

service: pdf-service
  name: aws
  runtime: python3.7
  # Set environment variable for the S3 bucket
    S3_BUCKET_NAME: ${file(./config.yml):BucketName}
  # Gives our functions full read and write access to the S3 Bucket
    -  Effect: "Allow"
         - "s3:*"
          - arn:aws:s3:::${file(./config.yml):BucketName}
          - arn:aws:s3:::${file(./config.yml):BucketName}/*
    handler: handler.generate_pdf
      - http:
          path: new-pdf
          method: post
          cors: true
 # Creates an S3 bucket in our AWS account
     Type: AWS::S3::Bucket
       BucketName: ${file(./config.yml):BucketName}
    dockerizePip: true
  - serverless-python-requirements is the only Python file we need. It contains one function to generate the PDF and save it to Lambda. You can create more functions in this file to split the code into reusable parts, but for this example, one function was enough. Lambda functions take two arguments by default: Event and Context.

Context contains environment variables and system information.

Event contains request data that is sent to the lambda function.

In this project, we will send our function generate_pdf a filename and HTML, and it will return the URI for a PDF it creates.

import json
import pdfkit
import boto3
import os
client = boto3.client('s3')

# Get the bucket name environment variables to use in our code
S3_BUCKET_NAME = os.environ.get('S3_BUCKET_NAME')

def generate_pdf(event, context):

    # Defaults
    key = 'deafult-filename.pdf'
    html = "<html><head></head><body><h1>It works! This is the default PDF.</h1></body></html>"

    # Decode json and set values for our pdf    
    if 'body' in event:
        data = json.loads(event['body'])
        key = data['filename']
        html = data['html'] 

    # Set file path to save pdf on lambda first (temporary storage)
    filepath = '/tmp/{key}'.format(key=key)

    # Create PDF
    config = pdfkit.configuration(wkhtmltopdf="binary/wkhtmltopdf")
    pdfkit.from_string(html, filepath, configuration=config, options={})

    # Upload to S3 Bucket
    r = client.put_object(
        Body=open(filepath, 'rb'),

    # Format the PDF URI
    object_url = "https://{0}{1}".format(S3_BUCKET_NAME, key)

    # Response with result
    response = {
        "headers": {
            "Access-Control-Allow-Origin": "*",
            "Access-Control-Allow-Credentials": True,
        "statusCode": 200,
        "body": object_url

    return response


Now that you have all the code ready it is time to deploy. In your project folder run the following command from your terminal:

sls deploy
# or
serverless deploy

After you run deploy, Serverless will create everything for you in AWS. You will get HTTP POST endpoint that you will use to generate PDFs. The endpoint will look something like this:

You can use curl to test your function. The following curl command posts a JSON object to the lambda endpoint. The JSON object contains a filename and some html to turn into a PDF.

curl -d '{"filename":"my-sample-filename.pdf", "html":"<html><head></head><body><h1>Custom HTML -> Posted From CURL as {JSON}</h1></body></html>"}' -H "Content-Type: application/json" -X POST REPLACE-WITH-YOUR-ENDPOINT

Note: Replace “REPLACE-WITH-YOUR-ENDPOINT” with the endpoint you receive from Serverless.

After running this command you should receive the URI to your generated PDF.


Creating a scalable PDF generator is easy with AWS and Serverless. In addition to having a scalable API, you don’t have to worry about server maintenance. I encourage you to create more projects with AWS Lambda to get comfortable with the platform and configuration.

Thanks for reading!

Next Steps


  • Serverless Environment Variables –
  • Serverless CORS Guide –
  • Fonts AWS Lambda –
  • Boto3 API –
  • S3 Argument Reference –

Was this article helpful?

26 responses to “Building a PDF Generator on AWS Lambda with Python3 and wkhtmltopdf”

  1. Grayden says:

    Super helpful. Found this because I was looking for a workaround to configuring pdfkit on Heroku.
    Thanks for putting the guide together!

  2. v says:

    Thank you for putting it here however, I am getting error while trying to deploy. Please help
    Error Message:
    Error: docker not found! Please install it.

    • richard says:

      Hi v, thanks for the comment. I must have made the assumption that Docker was installed:

      You may be able to remove this from the serverless.yaml:

      dockerizePip: true

      Or change it to:

      dockerizePip: false

  3. Fabio says:

    Awesome article thank you 🙂

    I have a small problem and maybe you can help.

    sls deploys the function but when it executes I get a

    [ERROR] Runtime.ImportModuleError: Unable to import module ‘handler’: No module named ‘pdfkit’

    I have created a binary folder in my project and moved wkhtmltopdf in there but maybe I am missing a step here.

    Thank you very much!!


  4. Jeremy says:


    I have been exploring this on an existing serverless project that i have so the setup was pretty easy. I am getting this error:

    [Errno 13] Permission denied: ‘binary/wkhtmltopdf’

    I checked my deployment zip and binary/wkhtmltopdf is present. You approached this platform specific binary problem different than i have in the past, which i have created a layer to solve this issue.

    So im curious what you might think about this problem?


  5. Erick Tapia says:

    Permission denied: ‘binary/wkhtmltopdf 🙁

  6. Antonio Bautista says:

    Hello, thanks for the post. Can you help me?

    When i try to test in lambda

    “errorMessage”: “[Errno 13] Permission denied: ‘binary/wkhtmltopdf'”,
    “errorType”: “PermissionError”,

    • richard says:

      Can you tell me the steps to replicate this issue? Thank you.

      • Lachlan says:

        Heya Richard, thanks for the tutorial and the code I’m having the same problem as everyone else when deploying your package,

        Sun Jul 12 03:02:59 UTC 2020 : Lambda execution failed with status 200 due to customer function error: [Errno 13] Permission denied: ‘binary/wkhtmltopdf’.

        Developing on a windows machine, I tried previously to cobble together a version of your project and what solved it for me was to modify the permissions for the wkhtmltopdf file to allow for rwx for everyone. But had a separate issue of: “error while loading shared libraries:”

      • richard says:

        Can you try making the wkhtmltopdf binary executable?

        `sudo chmod +x ./binary/wkhtmltopdf`

  7. Ganesh says:

    Hi, How can we resolve the issue error while loading error while loading shared libraries:”

  8. Audrey says:

    Hi Richard, fantastic tutorial, thank you!

    I find when trying to delete the cloudformation stack generated from aws console (log on as same iam user as the one running sls deploy), deletion will fail due to the S3 sls deployment bucket, how to make this sls deployment bucket deletable along with the stack? Thank you.

  9. Anderson Bispo says:

    For everyone who is getting the error “error while loading shared libraries:”, I solved it adding one layer in the serveless.yml.

    handler: handler.generate_pdf
    – arn:aws:lambda:sa-east-1:347599033421:layer:wkhtmltopdf-0_12_6:1


    This layer I got (here)[]. Look for your region there.

    Besides, I changed the binary path in the from binary/wkhtmltopdf to /opt/bin/wkhtmltopdf.

    And it worked! But I still had problem with the fonts. So, I added this layer as well:
    – arn:aws:lambda:sa-east-1:347599033421:layer:amazon_linux_fonts:1
    That I got (here)[]

    And I added this value to the environment in the serveless.yml:
    FONTCONFIG_PATH: /opt/etc/fonts

    Now, it worked like a charm.

    Thanks Richard, awesome tutorial!

  10. Abdullah Khan says:

    I dont usually comment on blogs, but, honestly, great & comprehensive article.

  11. Jean S. says:

    Hi Richard,

    Thanks for this excellent step by step tutorial. I just have a question though, after I sls deploy, I tried using the curl command above and I got {“message”: “Internal server error”} message. Is this because I am running this locally? Thank you.

  12. Raj says:

    I am getting
    OSError: No wkhtmltopdf executable found:

    • richard says:

      Did you do the part where you download the wkhtmltopdf binary?

      • Vik says:

        Richard, I have the same issue as Raj does. I did the part where I had to download the wkhtmltopdf binary. I might have put the unpacked folder in the wrong place. Would you be so kind to verify where the binary folder of my project might be? Please, let me know if there is any additional information that I can provide to help you help me.

      • richard says:

        The location of the binary should be “ROOT_OF_PROJECT/binary/wkhtmltopdf” where wkhtmltopdf is the actual binary. You can try cloning the project to test. I wonder if this code still works.

  13. Vik says:

    Richard, I have the same issue as Raj does. I did the part where I had to download the wkhtmltopdf binary. I might have put the unpacked folder in the wrong place. Would you be so kind to verify where the binary folder of my project might be? Please, let me know if there is any additional information that I can provide to help you help me.

Leave a Reply

Your email address will not be published. Required fields are marked *