Skip to content

README.md

Latest commit

 

History

History
103 lines (73 loc) · 3.74 KB

README.md

File metadata and controls

103 lines (73 loc) · 3.74 KB

late

Companion CLI for late.sh — a cozy terminal clubhouse for developers.

Connects to the SSH session and streams lofi audio locally with a live visualizer synced to your terminal.

Install

macOS / Linux / Termux:

curl -fsSL https://cli.late.sh/install.sh | bash

On Termux, the installer detects the environment and downloads the Android build.

Windows PowerShell (x64):

irm https://cli.late.sh/install.ps1 | iex

The PowerShell installer places late.exe in %LOCALAPPDATA%\Programs\late and prints a PATH hint if that directory is not available in the current shell.

Build from source

git clone https://github.com/mpiorowski/late-sh
cd late-sh
cargo build --release --bin late
# binary at target/release/late (late.exe on Windows)

What it does

  1. Opens an SSH session to late.sh
  2. Streams audio (lofi/ambient/jazz/classical) to your local speakers
  3. Runs a real-time FFT audio analyzer
  4. Sends visualizer data back to the TUI over WebSocket
  5. Syncs mute/volume controls between terminal and audio

Usage

late

That's it. On first run it will generate a dedicated SSH key at ~/.ssh/id_late_sh_ed25519. If you want to use a different key, pass --key /path/to/key. For YubiKey/FIDO security-key identities, use --ssh-mode openssh so OpenSSH handles PIN and touch prompts. In openssh mode, omitting --key lets OpenSSH use your normal ~/.ssh/config, agent, and default identity discovery.

Options

--ssh-target <host>        SSH target (default: late.sh)
--ssh-port <port>          SSH port override
--ssh-user <user>          SSH username override
--key <path>               SSH identity file override
--ssh-mode <mode>          SSH transport: native (default), openssh, or old
--ssh-bin <command>        SSH client command for openssh/old modes (default: ssh)
--audio-base-url <url>     Audio stream URL
--api-base-url <url>       API URL for WebSocket pairing
-v, --verbose              Debug logging to stderr

Requirements

  • Linux, macOS, or Windows x64 (WSL works too)
  • Working audio output device
  • Rust toolchain (if building from source)

--ssh-mode openssh uses a system OpenSSH client with an internal ControlMaster connection. It is the recommended mode for YubiKey/FIDO security-key identities and other OpenSSH-managed auth flows because OpenSSH owns the PIN/passphrase, touch prompt, terminal echo, agent, and ~/.ssh/config handling.

--ssh-mode old keeps the legacy OpenSSH-through-PTY behavior and still depends on a system ssh binary. --ssh-mode native uses an embedded russh client, records host keys in ~/.ssh/known_hosts with accept-new semantics, fetches the pairing token over a dedicated SSH exec handshake, and does not require OpenSSH on $PATH. Native mode intentionally does not fall back to the legacy LATE_SESSION_TOKEN= banner protocol, so it will fail fast against an older server.

If your audio device does not support the stream's native 44.1 kHz output rate, the CLI now falls back to a supported device rate such as 48 kHz and resamples locally. Native 44.1 kHz playback is still preferred when available.

On WSL, audio startup failures now include a targeted hint covering DISPLAY, WAYLAND_DISPLAY, and PULSE_SERVER so users get an actionable fix path instead of only raw ALSA errors.

Privacy

The CLI connects to late.sh using your SSH key. Only your key fingerprint is stored — not the full public key. No IP logging, no tracking.

If you'd rather not use your real key:

ssh-keygen -t ed25519 -f ~/.ssh/late_throwaway
late --key ~/.ssh/late_throwaway

License

This repo is source-available under FSL-1.1-MIT. See LICENSING.md for the plain-English usage policy.