serverless-pydeps v0.1.0
Serverless Python Dependencies
Serverless Framework plugin to automatically add Python dependencies as layers to your functions.
Based on Lovage - a Python-only serverless library that's more RPC-like and less HTTP service oriented.
Install
sls plugin install -n serverless-pydeps
This will automatically add the plugin to your project's package.json
and the plugins section of its
serverless.yml
. That's all you need. There is no need to install or configure anything else, not even Python.
The plugin will automatically create a layer from the dependencies listed in Pipfile
or requirements.txt
and attach
it to all functions with Python runtime when you run sls deploy
.
How Does It Work?
Dependencies are collected and packaged in a special Lambda function. This means you won't have to wait on dependencies to download to your local computer, get zipped up, and uploaded back to S3. This makes deployment, and especially updates, really fast. Dependencies will only be built and uploaded when they change. Your code packages will contain your code and nothing else.
Dependencies will be installed by the same system that runs them, so you never have to worry about running the right Python version on the right OS.
Known Limitations
- Only AWS is supported since we use Lambda Layers
- serverless-prune-plugin will not clean-up old layers
Other Options
serverless-pydeps | serverless-python-requirements | |
---|---|---|
Setup | Install plugin | Install plugin, Python, and potentially Docker for consistent requirements |
Package speed | Fast as dependencies are not packaged locally | Slow as every code change repackages all dependencies |
Upload Speed | Never uploads anything locally | Uploads all dependencies on every code change |
Supported providers | AWS | All providers supported by Serverless (AWS, GCP, Azure, etc.) |
Private repositories | Not supported as dependencies are downloaded in a Lambda function | Supported |
Native code dependencies (*.so files) | Only supported if dependency has proper binary wheel | Supported with custom Dockerfile , dockerExtraFiles , etc. |
Local disk usage | Zero | All dependencies are cached and also stored in .serverless |