Quick Start

Create a file named hello.jde:

let x = 10
let y = 32
let answer = x + y
print(answer)

Run it:

jade run hello.jde

Output:

42

jade run also takes the name of a script from jade.toml, and with no argument at all it runs the current project's entry file. See the CLI Reference.

Seeing what a program left behind

Pass --verbose to print every global variable after the program finishes:

jade run hello.jde --verbose

Output:

42
Grammar = {"new": <builtin new>}
answer = 42
x = 10
y = 32

The program's own output comes first, then the globals in alphabetical order. Built-in globals such as Grammar are listed alongside your own.

A slightly bigger program

Functions, f-strings, arrays, and for loops, in greet.jde:

fn greet(name) {
    return f"Hello, {name}!"
}

let people = ["Ada", "Grace", "Alan"]

for p in people {
    print(greet(p))
}

jade run greet.jde

Hello, Ada!
Hello, Grace!
Hello, Alan!

To type-check without running:

jade check greet.jde

greet.jde: ok

Compiling to a binary

The same file compiles to a native executable that needs no jade installed to run:

jade build greet.jde -o greet
./greet

built: greet
Hello, Ada!
Hello, Grace!
Hello, Alan!

jade run and jade build are two engines for one language, and a program means the same thing under either.

Starting a project

Once a program outgrows a single file, jade new lays out a project:

jade new myapp

created: myapp/
  jade.toml
  main.jde
  .gitignore

Get started:
  cd myapp
  jade run

cd myapp
jade run

Hello from myapp!

jade run with no argument runs the entry file named in jade.toml, so from here on it is the only command you need.

Calling a model

Jade's reason for existing is the next step. A prompt is a value, ? sends it to your provider, and |> int makes the result an int or raises:

prompt p = "How many moons does Mars have? Reply with just the number."

let n = ?p |> int
print(n + 1)

This one needs an inference provider registered first — run jade register. See LLM Integration for the whole picture.