Fastly Compute Python SDK

Experimental Python SDK for Fastly Compute services.

Highlights

• Support for WSGI Frameworks. Flask and Bottle examples are included for reference.
• Access from Python to Fastly's API
• Full type hints and IDE support

Quick Start

Install Dependencies

To work with fastly-compute in Python, you must install two system dependencies:

1. The Fastly CLI.
2. The uv Python package manager.

Additional dependencies, including the Python SDK for compute and build tooling, will be installed and managed by uv in an isolated environment.

Set Up Your Project

For this basic project, we'll use the Flask microframework. We will create our project and add the SDK and build tooling by adding fastly-compute and flask:

$ uv init my-compute-service
...
$ cd my-compute-service
$ uv add fastly-compute flask
...

Write Your Service

uv init automatically creates a main.py file in your project directory. Replace its contents with the following Flask application code:

import platform
from flask import Flask
from fastly_compute.wsgi import WsgiHttpIncoming

app = Flask(__name__)


@app.route("/")
def index():
    version = platform.python_version()
    return f"Hello from Python {version} on Fastly Compute!"

HttpIncoming = WsgiHttpIncoming(app)

Configure Compute Service Entry Point

The fastly-compute-py build tool, provided as part of the fastly-compute package, needs to be told about the module containing the compute service we just created. We can do this by modifying the pyproject.toml as follows:

# Add to end of pyproject.toml
[tool.fastly-compute]
entry = "main"

Then, let's do a quick test to make sure we are able to build a WebAssembly (Wasm) component:

$ uv run fastly-compute-py build
Building Python application for Fastly Compute...
  Entry point: main
  Output: bin/main.wasm
  Resolving Python dependencies...
  Componentizing Python application...
  Composing final WebAssembly module...
  Injecting Fastly metadata...
✓ Build complete: bin/main.wasm

Test and Deploy Your Service Using the Fastly CLI

Now that we have the skeleton of our service, let's test it using the Fastly CLI.

1. Run fastly compute init

2. Say yes when warned "The current directory isn't empty." Answer "Other" when it asks for Language.

3. Add this to the bottom of fastly.toml:

   [scripts]
   build = "uv run fastly-compute-py build"

With that in place, we can now run fastly compute serve to test our service locally. When you're ready, you can use the Fastly CLI to deploy the service to the production fleet and perform other actions.

See the examples/ directory for more examples.

Run Some Examples on Your Own Machine

We ship a few simple examples you can run locally to get a taste of what's possible.

Status

Currently supports...

• Building pure Python into WebAssembly components
• Creating Python bindings from Fastly's WIT files
• Hosting web frameworks by adapting Fastly's API to WSGI
• Hosting non-WSGI applications by writing directly against Fastly's API

Caveats

• So our memory-snapshotting build process can retain them, all the packages needed at runtime must get imported when your entrypoint (e.g. flask-app.py) is imported. This can happen transitively. But beware of deferred imports like non-top-level ones; if they aren't triggered by importing your entrypoint, they will fail at runtime. (If you have third-party code that uses non-top-level imports, you can ensure they work by importing them at the top level in your own code.)
• Third-party C extension modules are not yet supported.
• Our in-Python API may change backward-incompatibly during this beta period.

Contributing

See CONTRIBUTING.md.