How It Works

Zephyr is in practice three things: a language, a compiler (the CLI), and a "runtime" (the Zephyr cloud).

The Language

Zephyr is a domain-specific language with both declarative and functional qualities. It uses syntax familiar to developers, such as variables and function calls. Instead of objects or classes, its first-class entities are cloud architecture components such as services, operations (serverless functions), resources, and events.

Here is a sample:

Zephyr code sample

// Import tools for AWS lambda
use aws::lambda

// Configure AWS
cfg aws!({
  region: 'us-east-1'
})

// Define a serverless function
op addWidget(widget) = lambda!(
  #pkg::src('./operations/add-widget'), widget
)

// Use it in an API
api 'widgets-service.example.com' = REST {
  POST /widgets widget = addWidget(widget)
}

The Compiler

The Zephyr compiler translates your high-level Zephyr code into specific, low-level cloud resources. It deploys these some of these resources to your account with your cloud provider, and some to the Zephyr cloud.

The Zephyr compiler ships as a CLI, which supports related actions, such as logging in to and managing your Zephyr account.

The Zephyr Cloud

Zephyr applications also require a "runtime" in the form of the Zephyr Cloud. This is to support high-level features such as architectural observability and chains. As a rule of thumb:

• Architecture components - such as servers, serverless compute, and databases or storage - will be deployed to your cloud account. This simplifies usage-based billing, and allows you to readily integrate your favorite monitoring tools.
• Architecture connectors - such as asynchronous pipelines and gateways - will be deployed to Zephyr's cloud.

Zephyr will also provide a local runtime, for experimentation and testing.