Examples

Showcase: larger demos

The showcase/ track combines several capabilities at once (remote LLMs, speech, distributed compute, CI/CD) into larger demos, without stopping to explain every concept along the way.

Example	What it shows	Needs beyond installing Mélodium
Smart LLM Router	A JavaScript routing function sends each prompt to the cheapest LLM tier that can handle it	an LLM provider API key
AI Voice Assistant	Local speech-to-text, a remote LLM, and text-to-speech chained across two entrypoints (`chat` and `voice`)	an LLM provider API key; `voice` also needs an ElevenLabs key and a microphone
CI Pipeline	A three-stage CI pipeline (build, test, package) running entirely on provisioned cloud containers	a Mélodium Services API token, or a local `podman`/`docker compose` setup
CI Failure Triage	CI step output classified by a small JS function, with only genuine failures sent to an LLM for diagnosis	a Mélodium Services API token and an LLM provider API key
Distributed LLM Cluster	An HTTP front-end that provisions a remote worker on demand to run the LLM, keeping `ml` off the front-end entirely	a Mélodium Services API token and an LLM provider API key

Learning path: the tutorial track

Create a file named hello-world.mel, and write this code:


#!/usr/bin/env melodium
#! name = hello_world
#! version = 0.1.0
#! require = std:0.10.*
    
use std/engine/util::startup
use std/engine/log::logInfoMessage
    
treatment main() {
    startup()
    logInfoMessage(message = "Hello World!")
    
    startup.trigger -> logInfoMessage.trigger
}

Then run it using melodium run hello-world.mel.


$ melodium run hello-world.mel
[2012-12-21T01:32:45.678Z] info: info: Hello World!

That single file is as far as inline snippets go. Real Mélodium projects live in a directory with a Compo.toml, and the best way to learn the language from there is the tutorial/ track in the Mélodium repository: ten short, self-contained steps, each introducing one or two new concepts, in increasing order of difficulty. Clone the repository to run them:


git clone https://github.com/melodium-tech/melodium.git
cd melodium/examples/tutorial

Each step is its own directory. Run a step from inside its own directory, because several of them read or write files relative to the current working directory:


cd 01_hello_melodium
melodium run Compo.toml [arguments...]

Step	Introduces	Needs beyond installing Mélodium
01: Hello, Mélodium	Treatments & connections, `Block<T>` vs `Stream<T>`, functions, `startup()`	nothing
02: Flow & Generics	Generic treatments, `std/flow` combinators, arithmetic & comparison	nothing
03: Text & Files	File I/O, regular expressions, text composition	nothing
04: JSON Toolkit	Parsing & validating JSON, `Option` unwrapping	nothing
05: HTTP Client	Calling a remote HTTP API, telling network failure from application error	internet access
06: HTTP Server API	The `HttpServer` model, routing, the `@HttpRequest` context	a free TCP port and `curl`
07: SQL CRUD API	A `SqlPool` model shared across requests, combined with an HTTP server	a reachable PostgreSQL database
08: JavaScript Transform	The `JavaScriptEngine` model, transforming structured data with JS	nothing
09: Process Pipeline	Running an external command as a treatment, streaming through its stdin/stdout	the `sort` command in `PATH`
10: Distributed Computation	The `distrib` package: running a treatment on another engine	a second Mélodium engine (`melodium dist`)

Every example, tutorial or showcase, is a self-contained Mélodium project (Compo.toml + one or more .mel files) with its own README explaining what it does and why it is built the way it is. Run any of them the same way:


melodium run <example-dir>/Compo.toml [entrypoint] [-- --arg value …]

See the examples README for the full capability map (which package demonstrates what) and a note on how each example was verified.