keybr-tui 0.2.1

Adaptive terminal (TUI) typing trainer in Rust that ports the keybr.com algorithm, runs offline, and imports your keybr.com data
keybr-tui-0.2.1 is not a library. 
 _              _
| | _____ _   _| |__  _ __
| |/ / _ \ | | | '_ \| '__|
|   <  __/ |_| | |_) | |
|_|\_\___|\__, |_.__/|_|
          |___/

  practice the keys that slow you down

keybr-tui

CI Crates.io docs.rs License: MIT MSRV

Adaptive touch-typing practice for the terminal. keybr-tui times every key you press, finds the ones slowing you down, and builds drills aimed straight at them. It is a faithful port of the keybr.com learning algorithm, runs entirely offline, and can import your keybr.com history.

Install

curl -sSf https://y0sif.github.io/keybr-tui/install.sh | sh

The script installs the latest prebuilt binary for your platform. Prefer Cargo? Run cargo install keybr-tui.

Prebuilt binary. Every release ships binaries for Linux (x86_64), macOS (Intel and Apple Silicon), and Windows (x86_64). Unix archives are .tar.gz, Windows is .zip.

tar -xzf keybr-tui-x86_64-unknown-linux-gnu.tar.gz
./keybr-tui

From source.

git clone https://github.com/y0sif/keybr-tui.git
cd keybr-tui
cargo install --path .

Pin a version for the install script with KEYBR_TUI_VERSION=v0.2.0.

How it adapts

Every session feeds a per-key model of how fast and accurately you type. keybr-tui reads that model to decide what you practice next.

  1. Start small. You begin with a handful of letters, not the whole keyboard.
  2. Earn the rest. A new letter unlocks only once your active set reaches a confidence threshold, measured against your target speed.
  3. Drill the weak spot. The slowest key in your active set becomes the focus key and shows up more often in the text you are given.
  4. Read like words. A phonetic Markov chain builds pronounceable pseudo-words from your active letters, so practice never feels like random noise.
  5. Learn from each key. Every keystroke's reaction time updates a smoothed per-key average, the same calculation keybr.com runs.

Because the engine is a faithful port of the keybr.com algorithm, the practice you get in the terminal matches what the website would give you.

Features

Feature What you get
Adaptive drills Phonetic Markov text built from your active letters, biased toward your weakest key
Per-key tracking Reaction times smoothed into a confidence score for every key
Progressive unlock Letters open up as you prove proficiency, in keybr.com's frequency order
keybr.com import Bring your full learning state over from a keybr.com data export
Two error modes Forgive mistakes and keep moving, or stop until you correct them
Lesson summary WPM, accuracy, and your weakest keys after every round
Local and offline No account, no network, stats saved as plain local files
Terminal-native Minimal ratatui interface in ANSI colors, no chrome in the way

Usage

keybr-tui [OPTIONS]
Flag Description
--target-wpm <N> Target typing speed in words per minute (default: 35)
--error-mode <MODE> move-on (default) or stop-on-error
--reset Delete saved stats and start fresh
--data-dir Print the data directory path and exit
--import <FILE> Import a keybr.com data export and exit (see below)
--force With --import, replace existing stats (a .bak backup is kept)
--help Show help
--version Show version

In-session keys: Esc returns to the menu or quits, Enter selects a menu item or dismisses the lesson summary, the arrow keys navigate menus and settings, and Left/Right adjust setting values.

Bring your keybr.com history

Already practicing on keybr.com? Carry your progress over and continue in the terminal.

  1. On keybr.com, open your profile and click Download data. You get a typing-data.json file with your full practice history.

  2. Import it:

    keybr-tui --import typing-data.json

The importer replays every session through the same per-key smoothing keybr.com uses, so your unlocked letters, per-key speeds, and focus letter come out exactly as the website computed them.

A few details worth knowing:

  • Your target speed is not part of the export. If you use a non-default target on keybr.com, pass --target-wpm <N> alongside --import, since the unlocked set is derived against it. With no flag, the target saved in your config is used.
  • Sessions from non-English layouts are skipped, since this TUI is English-only for now.
  • Existing local stats are never overwritten unless you pass --force, which still writes a stats.json.bak backup first.

Configuration

keybr-tui runs with zero config. To change the defaults, edit ~/.config/keybr-tui/config.toml (the XDG path on Linux):

target_wpm = 35
error_mode = "forgive-mistakes"  # or "stop-on-error"
fragment_length = 100
Key Default Description
target_wpm 35 Target speed in words per minute; drives the confidence threshold
error_mode "forgive-mistakes" "forgive-mistakes" advances past typos, "stop-on-error" blocks until you correct them
fragment_length 100 Characters of practice text generated per lesson
natural_words true Prefer real English words, falling back to the phonetic model when none fit your active letters
daily_goal_minutes 30 Daily practice goal in minutes; 0 hides the daily-goal indicator
alphabet_size 0.0 Fraction of the non-starter alphabet to force-unlock regardless of confidence (keybr's alphabetSize), clamped to the range 0.0 to 1.0

Stats are stored separately. Run keybr-tui --data-dir to find them.

Documentation

Credits

Algorithm inspired by keybr.com by aradzie. Built with ratatui and crossterm.

License

MIT