AI assistance in your Bash terminal and local voice dictation for your desktop.

Integration	What it does	Requirements
Text	Routes natural-language input to Pi; shell commands run normally.	Bun, Bash 4.4+ in Emacs editing mode, a configured Pi provider
Voice	Hold Copilot to record; release to transcribe locally and paste. Never submits the text.	Bun, Ubuntu/Debian Wayland, x86_64 or ARM64, a supported Copilot keyboard

Install

Install Bun, then run the command for your integration:

Text

curl -fsSL https://github.com/xLongLink/uncode/releases/latest/download/text.sh | bash

Voice

curl -fsSL https://github.com/xLongLink/uncode/releases/latest/download/voice.sh | bash

Installers verify the archive's SHA-256 checksum and keep release files under ${XDG_DATA_HOME:-~/.local/share}/uncode/releases. They print the uninstall command. Voice uses sudo for system setup; do not move its installed directory. See each integration's README for setup and permissions.

These commands execute a downloaded script; inspect it first if needed. To pin a version, replace latest/download with download/v0.0.0. Archives and checksums are also available for manual installation on the releases page.

From source

git clone https://github.com/xLongLink/uncode.git
cd uncode
bun run up:text     # or bun run up:voice

Use bun run up to install both. Uninstall with bun run down:text, bun run down:voice, or bun run down.

Development

bun install
bun install --cwd text
bun install --cwd voice
bun run check
bun run fmt

text/ and voice/ are independent Bun packages with their own dependencies, lockfiles, checks, tests, licenses, and release tooling. Either folder can be copied out and developed without the root or the other package. Inside either folder, run bun install, bun run check, bun test, or bun run build. Root commands are convenience wrappers; this is not a Bun workspace.

Checks cover formatting, linting, types, builds, Bash syntax, and both integration test suites. Run suites separately with bun run test:text or bun run test:voice. Ordinary tests do not contact AI providers. For native inference, run VOICE_NATIVE_TEST=1 bun test inside voice/.

Release

Commit and push changes, then push an annotated version tag:

git tag -a v0.0.1
git push origin v0.0.1