Skip to main content

Kind of like Celery, but simpler and with more Lambda

Project description


Actions Status PyPI PyPI pyversions PyPI status GitHub stars

Python-only serverless library that's more RPC-like and less HTTP service oriented.

Status: Usable but not battle tested. PRs are welcome!


Lovage is a Python library with no external dependencies. Just install and use.

pip install lovage

You will have to set up your AWS credentials to deploy with either environment variables, shared credentials files or any other method that works with boto3.


Lovage is a Python 3 library that makes it very easy to offload normal Python functions to the cloud using AWS Lambda functions.

Call Functions Easily

Lovage lets you call functions without knowing anything about AWS API. You define the function as part of your codebase, use @app.task decorator, deploy it, and then just call the function with .invoke() or invoke_async(). Function arguments, return values, and exceptions can still be used as usual. You don't need to worry about serialization or AWS API. Everything just works as it normally does with normal Python functions.

Note: exceptions are only supported when using PickleSerializer. With the default JSONSerializer all exceptions are converted to LovageRemoteException.

import lovage.backends

app = lovage.Lovage(lovage.backends.AwsLambdaBackend("lovage-test"))

def hello(x):
    return x + 1

if __name__ == "__main__":
    app.deploy(root=".", requirements=["requests"])
    print("hello.invoke(1) returned", hello.invoke(1))

Compartmentalize Functions

It's easy to define separate IAM policies for each function to enhance your security with compartmentalization. You can give granular access to each function to just the resources it needs.

import boto3
import lovage.backends
import os.path

app = lovage.Lovage(lovage.backends.AwsLambdaBackend("lovage-test"))

# let this function send emails using SES as
    "Version": "2012-10-17",
    "Statement": [
            "Effect": "Allow",
            "Action": [
            "Resource": "*",
            "Condition": {
                "StringEquals": {
                    "ses: FromAddress": ""

def send_email(x):
    boto3.client("ses").send_email(Source="", ...)

if __name__ == "__main__":
    app.deploy(root=os.path.dirname(__file__), requirements=["boto3==1.12.25"])

Requirements Layer Generated in Lambda

Unlike other solutions, Lovage collects and packages required libraries in Lambda itself. Each deployment has a custom resource that gets the requirements list as a parameter, downloads all of them in Lambda, uploads it directly to S3, and finally creates a Lambda layer containing all the dependencies. This gives you:

  • Much faster cloud-local dependencies downloads and uploads
  • No local development dependencies but Python (no need for Docker, no need to run on Linux, etc.)
  • Faster code updates as you don't have to zip up the requirements and upload them along with your code
  • Cleaner working directory with no dependencies being duplicated from your site-packages and no hidden folders
import boto3
import lovage.backends

app = lovage.Lovage(lovage.backends.AwsLambdaBackend("lovage-test"))

if __name__ == "__main__":
    app.deploy(requirements=["boto3==1.12.25", "requests", "Django>=2.0.0"])
    # or...

Other Features

  • CloudFormation stack leaves nothing behind and can be deleted without any special treatment
  • Easy to test locally without deploying anything
  • No need for Node.js
  • Versatile configuration in code


This script will deploy one function to AWS using Lambda, S3 and CloudFormation. It will then execute the function twice. At first it will wait for the function to finish and print its answer. Then it will execute it asynchronously and return control to your script immediately.

import lovage
import lovage.backends

app = lovage.Lovage(lovage.backends.AwsLambdaBackend("lovage-test"))

def hello(x):
    print("hello world!")
    return x + 1

if __name__ == "__main__":
    print("hello.invoke(1) returned", hello.invoke(1))

To delete the functions, simply delete the lovage-test CloudFormation stack. You can choose the name when creating the AwsLambdaBackend object.

Testing Locally

Sometimes you don't want to wait for a full deployment and just want to iterate locally. Lovage makes this simple with LocalBackend which is the default backend. app.deploy() will do nothing and any function call will be executed locally. When using invoke_async() a new thread will be created and the function will execute there.

import platform

import lovage

app = lovage.Lovage()

def hello():
    print("Hello locally from", platform.node())

if __name__ == "__main__":
    app.deploy()  # doesn't do anything

Ignoring Files

Lovage will package all files from the current working directory for the Lambda function and upload them for you. If you want to avoid including some files because they are not required, you can create a file named .lovageignore which works just like ce.gitignore. Any pattern listed there will be excluded from the package.

Separate Environments

A common use-case in cloud development is having a separate environment for development, QA and production. Sometimes even a separate environment for each developer. Lovage uses a self-contained CloudFormation stack for each environment. There are no local or remote side-effects to worry about. As soon as you delete the stack, everything is gone.

The environment name is set by the first parameter given to AwsLambdaBackend().

app_dev = lovage.Lovage(lovage.backends.AwsLambdaBackend("lovage-dev"))
app_prod = lovage.Lovage(lovage.backends.AwsLambdaBackend("lovage-prod"))

Caveat: if you use AwsLambdaBackend.add_resource() to add additional CloudFormation resources to your stack, you may have to delete those manually. For example, if you add a bucket, you have to make sure it's empty before deleting the stack.

Available Configuration

Configuration can be passed to the @app.task() decorator. For example:

def hello_world():
  return 42

Some configuration is platform-specific and will therefore have a prefix like aws_.

Configuration Purpose Default Value
timeout Set Lambda timeout in seconds. Every Lambda function has a maximum execution time. 3
aws_policies List of IAM policy documents to attach to the Lambda function. []
aws_vpc_subnet_ids List of VPC subnets to attach to the Lambda function. Must be used together with aws_vpc_security_group_ids. []
aws_vpc_security_group_ids List of VPC security groups to attach to the Lambda function. Must be used along with aws_vpc_subnet_ids. []

Best Practices

  • Always specify root so you are sure which files are packaged. You can use something like from pathlib import Path; app.deploy(root=Path(__file__).parent.parent) to easily get your root folder.
  • Always use if __name__ == "__main__": in files with Lovage tasks. Global code will be executed both locally and in Lambda. This may cause some unwanted side-effects.
  • You should probably have a separate script to call app.deploy(). No-op deploys are pretty quick, but still take time to zip up the code, check if the latest is already available on S3, and finally update the CloudFormation stack.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for lovage, version 0.0.7
Filename, size File type Python version Upload date Hashes
Filename, size lovage-0.0.7-py3-none-any.whl (20.9 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size lovage-0.0.7.tar.gz (20.7 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring DigiCert DigiCert EV certificate Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page