KaizenReply

Messages don't get rewritten. They evolve. Improve every message, one Kaizen at a time.

改善 — Kaizen is a Japanese word meaning "change for the better" or "continuous improvement". It's a philosophy rooted in making small, consistent refinements rather than sweeping overhauls. Applied to communication — every message you send can be a little clearer, a little more fitting, a little better than the last draft. That's the idea behind KaizenReply.

---

💭 Why I Built This

Honestly? I was tired of copy-pasting my messages into ChatGPT just to fix grammar or make them sound less casual before sending. Every time — open a new tab, paste the message, type "fix this", wait, copy the result back. It works but it's annoying when you're doing it ten times a day.

So I thought — why not just build my own thing? A minimal web tool where I paste my message, pick a tone, and get the improved version instantly. No extra tabs, no prompting, no context switching.

Once I had that working, I figured it'd actually be useful for others too. Not everyone is comfortable with AI tools, and a lot of people just want a simple "make this message sound better" button without any of the AI wrapper complexity.

The Kaizen philosophy felt like the right name for it — the idea that every message can be made slightly better, one small refinement at a time. Not a full rewrite, just a continuous improvement. That's exactly what this does.

I also thought about adding screenshot support — paste a screenshot of a message and have it extract + improve the text. That would be genuinely useful. Not implementing it right now, but it's on the list for when I get around to it.

---

✨ Features

• 🧠 AI-Powered Evolution — Powered by Groq + Llama 3.3 70B. Fast, free, high quality.
• 🎭 10 Tone Presets — Casual, Professional, Polite, Formal, Friendly, Gen Z, Persuasive, Assertive, Diplomatic, Concise. Plus custom tone input.
• 📱 Platform-Aware Output — WhatsApp, LinkedIn, Email, Telegram, Instagram, Facebook, SMS, Discord, X (Twitter), or platform-neutral.
• 📏 Hard Character Limits — Enforces SMS (160) and X/Twitter (280) constraints at the prompt level.
• 💬 Conversation Context — Optional chat history and recipient description for smarter tone matching.
• 🔄 Reply Mode — Suggests 3 distinct, ready-to-send replies for any message you receive.
• 📊 Kaizen Score — Side-by-side Before → After comparison with a breakdown across Clarity, Tone, Professionalism, and Readability.
• 🤖 AI Tone Suggestion — Let the model analyse your draft and recommend the best tone.
• 📋 One-Click Copy & Share — Copy evolved text or share via Web Share API.
• 🌗 Dark / Light Mode — Persisted in localStorage, smooth transitions.
• 🔒 Anti-Spam Guard — 1-second per-IP cooldown + 5-minute response cache.

---

🧱 Tech Stack

Layer	Technology
Frontend	Plain HTML5, Vanilla CSS, ES6 JavaScript — zero build steps
Backend	FastAPI (Python, ASGI) + Uvicorn
AI Model	Groq llama-3.3-70b-versatile
HTTP Client	httpx (async)

---

🚀 Quick Start

1. Clone & Enter

git clone https://github.com/MKishoreDev/KaizenReply.git
cd KaizenReply

2. Set Up Virtual Environment

python -m venv .venv
source .venv/bin/activate        # Windows: .venv\Scripts\activate

3. Install Dependencies

pip install -r requirements.txt

4. Configure Environment

Create a .env file in the root:

# Get your free key at https://console.groq.com/keys
GROQ_API_KEY=your-groq-api-key-here
GROQ_MODEL=llama-3.3-70b-versatile

Groq Free Tier: ~14,400 requests/day · 30 RPM · No credit card required.

5. Run the Server

python -m uvicorn app.main:app --reload --port 8000

Open http://localhost:8000 in your browser.

---

🌐 Deployment

For Railway, Render, or Fly.io — set env vars in the platform dashboard, no .env file needed. The app reads os.getenv() directly, with .env as a local-only fallback.

GROQ_API_KEY=your-groq-api-key-here
GROQ_MODEL=llama-3.3-70b-versatile

---

🔌 API Reference

POST /api/improve

Evolves a draft message based on tone, platform, and context.

Request:

{
  "message": "bro send that file asap",
  "tone": "Polite",
  "platform": "Email",
  "conversationContext": "",
  "recipient": "Manager"
}

Response:

{
  "improved": "Hi, could you please send the file when you get a chance? Thanks!",
  "score": {
    "before": 52,
    "after": 91,
    "breakdown": { "clarity": 18, "tone": 24, "professionalism": 14, "readability": 21 }
  }
}

---

POST /api/reply

Generates 3 ready-to-send replies for a received message.

Request: Same fields as /api/improve.

Response:

{ "suggestions": ["Reply 1", "Reply 2", "Reply 3"] }

---

POST /api/analyze

Recommends the best tone for a draft message.

Request: { "message": string }

Response: { "tone": string, "platform": "", "reason": string }

---

GET /health

{ "status": "ok", "model": "llama-3.3-70b-versatile" }

---

📁 Project Structure

KaizenReply/
├── app/
│   ├── main.py          # FastAPI routes, Groq client, anti-spam, cache
│   └── models.py        # Pydantic request & response schemas
├── static/
│   ├── index.html       # Full SaaS UI (no framework)
│   ├── styles.css       # Vanilla CSS design system — dark/light, responsive
│   ├── app.js           # Frontend logic — modes, chips, copy, share, cooldown
│   ├── logo.png
│   └── banner.png
├── requirements.txt
├── .env.example                # Local dev only — not committed
├── Dockerfile
├── LICENSE
└── README.md

---

🐳 Docker

docker build -t kaizenreply .
docker run -p 8000:8000 -e GROQ_API_KEY=your-key kaizenreply

---

🤝 Contributing

1. Fork this repository
2. Create a branch — git checkout -b feature/YourFeature
3. Commit — git commit -m 'Add YourFeature'
4. Push — git push origin feature/YourFeature
5. Open a Pull Request

Ideas to Hack On

• 📸 Screenshot support — paste an image, extract text, improve it
• 🔌 Swap in Claude, Gemini, or a local Ollama model
• 🌍 Multi-language support
• 🎨 More theme presets

---

📄 License

MIT License — see LICENSE for details.