Minimal AI Terminal

A small, GPU-accelerated AI terminal built from scratch with Rust, Tauri, and React. This is the companion source for the video How to Build Your Own AI Terminal.

It is intentionally minimal: the smallest solid terminal that still does the real thing, a PTY-backed shell, GPU rendering, and an AI agent that can read the terminal and run commands. For the full product built on these ideas, see Terax.

If the video does not play inline, watch terminal-preview.mp4.

Features

• Real shell in a pseudo-terminal (PTY) via portable-pty, cross-platform.
• Raw bytes streamed from Rust to the frontend over a Tauri Channel, no JSON round trip.
• GPU rendering with xterm.js and its WebGL addon.
• An AI panel (OpenAI) that reads recent terminal output and runs commands through the same PTY, using tool calling.

Stack

Rust + Tauri v2 · React 19 + TypeScript + Vite · xterm.js (WebGL) · portable-pty · OpenAI SDK · Tailwind v4 + streamdown

How it works

• Backend (src-tauri) spawns a shell inside a PTY and exposes three commands: pty_spawn, pty_write, pty_resize. A dedicated thread reads the shell's output and streams the raw bytes to the frontend over a Tauri Channel.
• Frontend (src) renders that byte stream with xterm.js on the GPU, sends keystrokes back through pty_write, and reflows on resize.
• AI agent (src/lib) gives the model two tools, run_command and read_terminal, and loops: the model asks for a tool, we run it against the PTY, feed the result back, and repeat until it answers.

The full walkthrough is in the video.

Prerequisites

• Rust
• Node.js and pnpm

Run

git clone https://github.com/crynta/yt-video-terminal.git
cd yt-video-terminal
pnpm install
pnpm tauri dev

The terminal works out of the box. To enable the AI panel, add an OpenAI key:

cp .env.example .env.local
# then set VITE_OPENAI_API_KEY in .env.local

Note: this calls OpenAI from the frontend, so the key lives in the app. That is fine for a demo, but in production you would proxy the request through the Rust backend. The model is set in src/lib/agent.ts.

Build

pnpm tauri build

Project structure

src-tauri/src/
  pty.rs            PTY session: spawn, read thread, write, resize
  lib.rs            Tauri commands + shared session state
src/
  components/
    Terminal.tsx    xterm.js + WebGL, wired to the PTY
    AiPanel.tsx     the AI chat panel
  lib/
    agent.ts        OpenAI tool-calling loop
    terminalBridge.ts  read terminal output + run commands
  App.tsx           terminal + AI panel layout

License

MIT