Serverless Framework Python Lambda Monorepo

Example of setting up Python lambdas using Serverless Framework using serverless-python-requirements with common shared module without using lambda layers.

Structure

/
├─ services/        - services, each is a Serverless stack
│   ├─ first/       - stack with first lambda
│   ├─ pycommon/    - scripts reused in first and second modules
│   ├─ second/      - stack with second lambda
│   ... more Serverless stacks

Common module

When working with monorepo often we use some same code, configuration and libraries across different modules. Imports in Python can be tricky not as straightforward as for similar repository architecture when using NodeJS.

Some related threads found online:

• #256
• #606
• How to share code in serverless with Python

Most answers advice to use Lambda Layers but this project show example how to do this without it.

It can be achieved using poetry (or pip) and with Vendor library directory feature of serverless-python-requirements. The trick is to:

• create common module, e.g: common
• for testing locally reference common module in poetry.toml
• to include files in the package uploaded to AWS add the common directory to vendor section of serverless.yml

Platform specific dependencies

Some libraries like numpy or pandas are OS specific and in order to work on AWS Lambda a project need to be build on Linux environment. So if you work on Mac like myself to make it work you need to add dockerizePip configuration to your serverless.yml

Requirements

On local machine install:

• Poetry
• Docker
• Python3+

Development

Install:

yarn install

Test:

yarn run test

Deploy:

yarn run deploy