⚡ LUMI-JOY
The Lightning-Fast, Zero-Lag AI Coding Assistant
An AI coding companion that feels instant, lets you undo any mistake with one click, and slashes your AI bills by up to 90%.
Dedicated to the open-source community and the creators at Nous Research & Hermes. Made with love so anyone can build with fast, reliable AI.
— William Andrew Cruz (bozoegg / CardSorting) · Read Author's Note
🌟 Why Developers Love It
| What You Get |
Why It Helps You Every Day |
How Fast & Simple It Is |
| ⚡ Instant Responses |
Runs directly on your computer with no waiting, no lag, and no spinning wheels. |
Instant (< 0.15 ms) |
| 🔍 Fast Code Search |
Finds any file, function, or word instantly without freezing your screen. |
5–10x faster than normal search |
| ⏪ 1-Click Undo |
Made a bad edit? One command snaps all your files back to how they were. |
Instant time-travel undo |
| 💸 90% Cheaper Bills |
Remembers your project so you only pay for your new questions. |
Cuts API bills up to 90% |
| 🛡️ No Stuck Ports |
If a background server is stuck on port 3000, LUMI unlocks it automatically. |
Zero "port in use" headaches |
| 🚀 Smooth & Reliable |
Dedicated memory management ensures your screen never stutters or crashes. |
Over 6,800 actions every second |
🚀 Quick Start in 60 Seconds
# 1. Download & Build
git clone https://github.com/CardSorting/LUMI-JOY.git && cd LUMI-JOY
npm install && npm run build
# 2. Add your AI API key (friendly step-by-step wizard)
npx tsx src/index.ts --setup
# 3. Start coding!
npx tsx src/index.ts --profile coder
╔══════════════════════════════════════════════════════════════════════════════════════════╗
║ ⚡ LUMI-JOY v1.0.0 │ 👤 [💻 Coder] │ 🧠 [gpt-5.6-luna] │ ⏱️ 0.12ms │ 💰 $0.0018 │ ⭐ Fav ║
╠══════════════════════════════════════════════════════════════════════════════════════════╣
║ 👤 You: Refactor auth.ts to add expiration checks ║
║ ⚡ LUMI (Coder): Checked auth.ts and applied clean edits with unit tests (0.04ms) ║
╠══════════════════════════════════════════════════════════════════════════════════════════╣
║ 💡 Shortcuts: [Ctrl+M] Switch Model [Ctrl+P] Setup [/profile] Change Role [Ctrl+C] Quit ║
╚══════════════════════════════════════════════════════════════════════════════════════════╝
🎯 What You Get (Everyday Superpowers)
| The Common Problem |
How LUMI Solves It |
Your Real-World Benefit |
| Accidental bad edits or AI mistakes |
1-Click Instant Undo |
Snap all files, memory, and chat history back instantly with zero lost work. |
| Fear of AI breaking working code |
Safe Preview Sandbox |
Look over proposed edits visually before anything is saved to your computer. |
| Searches freezing on big projects |
High-Speed In-Memory Search |
Scans millions of lines of code in a flash with zero terminal freezing. |
| Typos in function or variable names |
Typo-Friendly Smart Matching |
Finds what you're looking for even if you or the AI misspell the name. |
| Ghost servers locking port 3000 |
Automatic Port Unlocking |
Automatically frees up stuck background ports so your apps just run. |
| Expensive AI bills on long chats |
Smart Memory Caching |
Locks heavy project rules in memory so you pay up to 90% less per turn. |
| Need a smarter model for hard bugs |
Swap AI Models Mid-Chat |
Switch between Claude, ChatGPT, or local models without losing conversation context. |
| Want a custom AI teammate |
Drag & Drop Customization |
Drop any notes or script files into a folder to create a personalized AI assistant. |
👤 Pick an AI Specialist
| Specialist |
Best For |
What It Does For You |
💻 coder (Lead Engineer) |
Coding & Refactoring |
Writes clean code, fixes bugs, cleans up old code, and makes sure tests pass. |
🔬 researcher (Deep Researcher) |
Research & Learning |
Searches documentation, summarizes web findings, and gives you verified facts. |
🛡️ sre (Reliability Engineer) |
Fixing Crashes & Errors |
Diagnoses error logs, checks system health, and fixes system problems automatically. |
✍️ writer (Technical Author) |
Guides & Documentation |
Writes crystal-clear documentation, user guides, and release notes. |
🎓 student (Patient Mentor) |
Learning & Onboarding |
Breaks down difficult code into friendly, step-by-step explanations. |
⚡ minimal (Speed Script) |
Fast Automation |
Gets straight to work on quick file edits using the lowest possible token cost. |
📋 What's New in Plain English
| Feature |
What It Means in Everyday Words |
Why You'll Love It |
| ⚡ High-Speed Search & Smart Tools (ADR-136) |
• Instant Code Search: Search your whole project with zero terminal lockups.• Typo Forgiveness: Finds the right code even with spelling mistakes.• Automatic Port Unlock: Never get stuck on port already in use errors.• Safe Sandbox Review: Inspect visual diffs before saving changes to your computer. |
Faster coding flow, zero terminal freezes, and complete peace of mind when editing code. |
| 🧠 Smart Cost-Saving Memory (ADR-135) |
• Up to 90% Cheaper: Never pay to re-send instructions the AI already read.• Instant Answers: Starts answering immediately without re-reading from scratch.• Live Dollar Savings: Shows your exact daily and monthly savings in real dollars. |
Massively cuts your monthly AI subscription and API costs. |
📜 See the complete release history in CHANGELOG.md.
📚 Advanced Developer Reference
Latest Verified Workspace Baseline
The authoritative run was generated on 2026-08-17T04:06:43.562Z using Node.js v23.5.0 on macOS ARM64. It passed:
| Verification lane |
Latest result |
| Pass 192 composition manifest |
566/566 components |
| Runtime capability smoke |
9/9 checks |
| Heterogeneous benchmark suite |
5/5 cases |
| Complete Flappy Bird React + TypeScript + Vite case |
8/8 assertions; 12/12 files |
| Architecture and performance guardrails |
6/6 checks |
(Measured fast-path execution throughput: 8506.11 frames/sec).
🏛️ Open Source & License
No comments yet
Be the first to share your take.