AWS Python uv Workspaces

Deploy Python Lambda functions from a uv workspace that uses package = true and a src/ layout.

SST traverses up from the handler path to find the nearest pyproject.toml.

├── sst.config.ts
├── pyproject.toml
├── handler.py
├── src
│   └── myapp
│       ├── __init__.py
│       └── utils.py
└── packages
    ├── api
    │   ├── pyproject.toml
    │   └── src/api
    │       ├── __init__.py
    │       └── handler.py
    ├── shared
    │   ├── pyproject.toml
    │   └── src/shared
    │       ├── __init__.py
    │       └── models.py
    └── worker
        ├── pyproject.toml
        └── src/worker
            ├── __init__.py
            └── handler.py

With package = true, the package is importable by name instead of src.*.

[tool.hatch.build.targets.wheel]
packages = ["src/myapp"]

Then import it normally in your handler.

from myapp import utils

Each workspace member can become its own function.

new sst.aws.Function("PackageHandler", {
  handler: "packages/api/src/api/handler.lambda_handler",
  runtime: "python3.11",
  url: true,
});

const rootHandler = new sst.aws.Function("RootHandler", {
  handler: "handler.lambda_handler",
  runtime: "python3.11",
  url: true,
});

const packageHandler = new sst.aws.Function("PackageHandler", {
  handler: "packages/api/src/api/handler.lambda_handler",
  runtime: "python3.11",
  url: true,
});

const workspaceHandler = new sst.aws.Function("WorkspaceHandler", {
  handler: "packages/worker/src/worker/handler.lambda_handler",
  runtime: "python3.11",
  url: true,
});

import json

from shared import models


def lambda_handler(event, context):
    return {
        "statusCode": 200,
        "body": json.dumps(models.response("api package")),
    }

import json

import arrow
from shared import models


def lambda_handler(event, context):
    body = models.response("worker package")
    body["timestamp"] = arrow.utcnow().isoformat()

    return {
        "statusCode": 200,
        "body": json.dumps(body),
    }

View the full example.