New examples: from hello world to a distributed LLM cluster
Mélodium’s example programs got reorganized in 0.10.3, and it’s a good excuse to point at what’s actually in there: one path to learn the language, and a set of bigger demos to see it used for real.
Learn the language in ten steps
Start at hello world, end at distributed computation, and pick up everything the language actually does along the way:
- Hello, Mélodium
- Flow & Generics
- Text & Files
- JSON Toolkit
- HTTP Client
- HTTP Server API
- SQL CRUD API
- JavaScript Transform
- Process Pipeline
- Distributed Computation
Each step introduces exactly one idea, so nothing is competing for attention while you’re learning it.
Five bigger demos
Each one combines several standard library packages into a single program:
- Smart LLM Router: a JavaScript decision function sizes up how complex an incoming prompt actually is and routes it to one of three LLM tiers, instead of paying for the biggest model every time.
- AI Voice Assistant: remote LLM chat, local speech-to-text, and remote text-to-speech, chained into one pipeline.
- CI Pipeline: a three-stage CI pipeline that runs entirely on provisioned cloud containers.
- Distributed LLM Cluster: provisions a remote engine on demand and runs inference on it, no manually started worker required.
- CI Failure Triage: classifies each CI step’s output deterministically first, and only calls an LLM for the ones actually flagged as failed. The steps that pass cost nothing beyond the classification.
Every example ships with its own README explaining what it does and how to run it, so picking one up is never more than cd and melodium run.