Hello World

Let's write your first Haira program.

A Simple Program

Create a file called hello.haira:

import "io"

fn main() {
    io.println("Hello, World!")
}

Build & Run

# Compile to binary
haira build hello.haira

# Run the binary
./hello

Or use haira run to compile and run in one step:

haira run hello.haira

Output:

Hello, World!

Variables and String Interpolation

import "io"

fn main() {
    name = "Haira"
    version = 0.1
    io.println("Welcome to ${name} v${version}!")
}

Haira uses ${expr} for string interpolation — any expression can go inside the braces.

Your First Agent

Here's a minimal AI agent with an HTTP endpoint:

import "io"
import "http"

provider openai {
    api_key: env("OPENAI_API_KEY")
    model: "gpt-4o"
}

agent Assistant {
    provider: openai
    system: "You are a helpful assistant. Be concise."
}

@post("/chat")
workflow Chat(message: string) -> { reply: string } {
    reply, err = Assistant.ask(message)
    if err != nil {
        return { reply: "Something went wrong." }
    }
    return { reply: reply }
}

fn main() {
    server = http.Server([Chat])
    io.println("Running on :8080")
    server.listen(8080)
}

# Set your API key
export OPENAI_API_KEY="sk-..."

# Build and run
haira run agent.haira

Your agent is now serving at http://localhost:8080/chat. To view the UI, run:

haira webui -c localhost:8080 -p 3000

Then open http://localhost:3000 in your browser.

What Just Happened?

In ~20 lines you defined:

1. A provider — configures which LLM backend to use
2. An agent — an LLM entity with a system prompt
3. A workflow — an HTTP endpoint that calls the agent
4. A server — serves everything as a native binary

No frameworks. No boilerplate. Just four keywords.

Next Steps

• Key Concepts — understand the four primitives
• Providers — configure LLM backends
• Agents — build intelligent agents
• Examples — see more real-world patterns