lumira

Real-time statusline for Claude Code and Qwen Code

Lumira is a zero-dependency statusline plugin that turns the bottom of your AI-coding terminal into a live cockpit — surfacing model, git state, token usage, cost, rate limits, active tools, and todo progress in one glance.

Why I built it

Working long sessions in Claude Code, I kept losing track of which model was active, how much context I had left, and what the run was costing me. The default statusline didn't answer those questions, and the third-party options either pulled in heavy dependencies or didn't speak Qwen Code's payload format. I wanted a single binary I could npx in any project and trust to render fast.

What it does

• Two render modes — 3-line custom layout for wide terminals + 1-line minimal mode that auto-switches under 70 columns.
• Powerline mode with 7 separator presets (arrow, flame, slant, round, diamond, compatible, plain) and a 3-tier color system (named ANSI / 256-color / truecolor) that auto-detects terminal capability.
• 7 hand-curated themes — Dracula, Nord, Tokyo Night, Catppuccin, Monokai, Gruvbox, Solarized.
• Context bar with color thresholds and an actionable /compact? hint when fill gets dangerous.
• Git status with branch, staged/modified/untracked counts (5s TTL cache); branch turns red on dirty repos.
• Token + cost telemetry — input/output counts, tok/s, $ spent, burn rate, 5h/7d rate-limit usage with reset countdown.
• Transcript parsing for active tools, agents, and todo progress, plus optional GSD task integration.
• OSC 8 hyperlinks — clickable directory and version in iTerm2, WezTerm, Kitty, VS Code, Alacritty.
• Interactive installer — npx lumira install walks through preset, theme, and icons with a live preview.

Stack

TypeScript (strict), Node.js. Zero runtime dependencies — everything from ANSI rendering to JSON parsing is hand-rolled to keep startup latency in the single-digit milliseconds.

Status

Actively maintained, published to npm, dual-platform (Claude Code + Qwen Code).