Files
riptune/README.md
T
2026-07-12 01:59:55 +02:00

102 lines
4.2 KiB
Markdown

# riptune
*Rip and tune.* A terminal-native music client for [Subsonic-compatible](http://www.subsonic.org/pages/api.jsp) servers (Navidrome, Airsonic, Gonic), built for the [niri](https://github.com/YaLTeR/niri) scrollable-tiling compositor. Because if id Software's engine can run on a smart fridge, your music player can run on a compositor IPC socket.
> **Status:** early scaffold — architecture and interfaces are in place, implementation is in progress. See [Roadmap](#roadmap).
## Why
Most terminal music clients treat the desktop as a black box: no MPRIS, no awareness of tiling layout, no way to hook into compositor-level keybinds. `riptune` is built the other way around — it assumes it's running inside a modern Wayland tiling setup and integrates accordingly.
## Features
- **MPRIS2 D-Bus interface** — waybar, GNOME/KDE media widgets, and hardware media keys control riptune out of the box, no riptune-specific config needed on their end.
- **Decoupled async pipeline** — network I/O (tokio) and audio decode/playback (dedicated OS thread) never share a runtime, so a slow server response can't stutter playback or freeze the UI.
- **niri IPC integration** — listens on niri's event-stream socket (not polling) to react to workspace changes; exposes a local control socket so niri keybinds can trigger in-app actions like favoriting a track.
- **Local SQLite index** — the remote library is mirrored locally so the TUI opens instantly, even on servers with slow `getArtists`/`getAlbum` responses over large libraries.
## Architecture
```mermaid
flowchart TB
subgraph Main["Main Thread"]
TUI["TUI (ratatui)<br/>render loop, input handling"]
end
subgraph AudioThread["Dedicated OS Thread"]
Decoder["symphonia decoder"]
Sink["cpal output stream"]
Decoder --> Sink
end
subgraph Tokio["Tokio Runtime"]
Subsonic["Subsonic client<br/>(reqwest)"]
Cache["SQLite cache<br/>(sqlx)"]
Mpris["MPRIS2 server<br/>(zbus)"]
Niri["niri IPC listener<br/>(event-stream socket)"]
end
TUI <-- "AppEvent (mpsc)" --> Tokio
TUI <-- "reads local state" --> Cache
Tokio -- "AudioCommand (mpsc)" --> AudioThread
AudioThread -- "AudioEvent: position, buffering" --> TUI
Subsonic -- "sync_library()" --> Cache
Subsonic -- "authenticated stream URL" --> AudioThread
Mpris -- "translates D-Bus calls into" --> Tokio
Niri -- "WorkspaceChanged" --> TUI
External1["waybar / notification daemon"] -.->|D-Bus| Mpris
External2["niri compositor"] -.->|IPC socket| Niri
```
**Why three execution contexts instead of one?** Audio playback is latency-sensitive in a way that async task scheduling doesn't guarantee — a busy tokio runtime (e.g. mid-sync with the server) shouldn't be able to introduce a buffer underrun. Giving audio its own OS thread with a bounded channel as the only interface makes that impossible by construction, not by careful scheduling.
## Repository layout
```
riptune/
├── src/ # binary crate: wires everything together
│ ├── main.rs
│ ├── app.rs # async orchestration (tokio runtime)
│ ├── audio.rs # audio thread entrypoint
│ ├── config.rs
│ └── events.rs
├── crates/
│ ├── riptune-core/ # Subsonic REST client, auth, models
│ ├── riptune-cache/ # SQLite local index
│ ├── riptune-mpris/ # MPRIS2 D-Bus server
│ ├── riptune-niri/ # niri IPC event-stream listener
│ └── riptune-tui/ # ratatui UI
├── Cargo.toml # workspace root
└── config.example.toml
```
## Configuration
`~/.config/riptune/config.toml`:
```toml
[server]
url = "https://music.example.com"
username = "you"
# Either a legacy Subsonic password (salted-token auth is computed at runtime,
# never sent in the clear) or a Navidrome API key, depending on server support.
password = "your-password-or-api-key"
[niri]
enabled = true
workspace_notifications = true
[cache]
path = "~/.cache/riptune/library.sqlite"
```
## Building
```sh
cargo build --release
```
Requires a D-Bus session bus (for MPRIS) and, optionally, a running niri instance (for IPC features — riptune degrades gracefully without them).