I've been building Legato, a realtime audio graph framework for Rust. The idea is to sit somewhere between PureData/MaxMSP (graph-based, visual routing) and something like JUCE.

The core of it is a minimal runtime and DSL for graph definitions. There are no loops, branching or evaluations. It's purely for wiring nodes together. If you want real logic, you write a custom node in Rust, register it, and drop it into the graph.

This way, you can easily extend the framework with Rust, as opposed to having to learn SuperCollider or CSound.

In addition to adding nodes in Rust, you can compose macros/templates/patches(still unsure which name I like best) of nodes, and then call them like regular nodes:

patch voice(attack = 200.0) { in freq gate audio { sine: mod, sine: carrier, adsr { attack: $attack, ... } }

    freq >> mod.freq
    mod >> carrier.freq
    gate >> adsr.gate
    carrier >> adsr[1]

    { adsr }

}

Patches inline into the same flat allocation, so spawning voice * 5 for polyphony is not too hard on memory layout.

There are better examples in the linked repository. They also "compose", so you can use patch a in patch b, etc.

The primary target right now is hardware — the idea is that you could build a VST, software synth, or groovebox and deploy it on embedded Linux (Raspberry Pi, etc.) using the NixOS images I'm working on (although not necessary).

Currently I am using CPAL to handle cross platform audio APIs.

I'm currently working on a few grooveboxes I will open source, likely in the tracker space.

The license is AGPLv3 with an additional permissions file that waives source disclosure for most creative projects (VSTs, synths, grooveboxes) without DAW/AI functionality.

Still early (no 0.1.0 yet), but the DSL parser, graph executor, midi tools, and a handful of nodes are working, and for myself I am already finding it productive.

Looking for feedback, contributors, and people looking to dogfood the project.

I will likely start posting some video demos on various platforms in the upcoming weeks, as it's at a point where it's actually quite fun to perform with.

Repo: https://github.com/legato-dsp/legato Docs: https://legato.gg