NapGPT

🔗 Live Demo

The AI that just... doesn't feel like it right now.

A cozy, intentionally-lazy AI chatbot built with Next.js 15, TypeScript, and Tailwind CSS. NapGPT wraps a normal LLM with a "Nap Wrapper" that modulates responses based on an Effort Level (0-100), sometimes refusing to help, stalling, napping mid-reply, or giving one-liners.

Features

• 🎯 Effort-based responses: Adjust the effort slider (0-100) to control how helpful NapGPT is
• ☕ Boost button: Temporarily increase effort for one reply (with cooldown)
• ⚡ Energy meter: Visual indicator that drains during responses and refills over time
• 💾 Persistent State: Beans, effort, and energy levels are saved across sessions
• 😴 Nap timer: Auto-sleep overlay after 30 seconds of inactivity
• 🛌 Blanket mode: Automatic cozy overlay when effort is low or idle for too long
• 🎨 Cozy UI: Warm palette, rounded corners, soft shadows, gentle animations
• 🎭 Easter eggs: /nap command, /dream command for surreal responses
• 🔄 LLM adapter: Works with Anthropic Claude (primary) or OpenAI API, falls back to mock mode (no API key required)

Quick Start

Prerequisites

• Node.js 18+ and pnpm (or npm/yarn)
• (Optional) Upstash Redis for production rate limiting

Installation

1. Clone the repository:

git clone https://github.com/rjspence3/napGPT.git
cd napGPT

2. Install dependencies:

pnpm install

3. (Optional) Set up Anthropic API key:

# Create a .env.local file in the root directory (never commit this file)
echo "ANTHROPIC_API_KEY=your-api-key-here" > .env.local

If you don't set an API key, the app will run in mock mode with deterministic responses.

Run Locally

pnpm dev

Open http://localhost:3000 in your browser.

Development with Auto-Restart Monitor

For development with automatic server restart on failures:

pnpm dev:monitored

This runs a monitor script that checks server health every 5 seconds and automatically restarts the dev server if it becomes unhealthy (after 3 consecutive failures).

Build for Production

pnpm build
pnpm start

Testing

Run the Playwright e2e tests:

# Run all e2e tests
pnpm test:e2e

# Run UI element tests only
pnpm test:ui

# Run CDP-based tests (Puppeteer)
pnpm test:ui-cdp

See tests/README.md for detailed testing documentation.

Usage

Effort Levels

• 0-15: Instant refusal ("meh… too tired")
• 16-35: One-sentence answer, suggests googling
• 36-70: Short, low-effort answer with mild sarcasm
• 71-85: Normal answer but with lazy tone
• 86-100: Full answer (rare), then "going back to sleep" line

Commands

• /nap - Returns a nap message and triggers a brief nap state
• /dream - Switches to surreal, dreamy tone for one reply (mock mode only)

Wake-Up Triggers

• Click the ☕ Boost button to temporarily increase effort for the next reply (10 second cooldown)
• Type keywords like "motivate me", "urgent", "deadline", "important", "please help" to add +15 effort for one reply

Nap Timer

Toggle the 🌙 Nap Timer button to enable/disable automatic napping. When enabled, NapGPT will automatically show a napping overlay after 30 seconds of inactivity.

Blanket Mode

Blanket mode automatically activates a cozy visual overlay when:

• Effort level is below 20, OR
• Nap timer is enabled and idle for 30+ seconds

The blanket overlay creates a warm, dimmed atmosphere with a subtle gradient and texture effect. It automatically toggles based on effort and idle state.

Default Values

• Initial Effort: 50 (moderate laziness)
• Initial Energy: 100% (fully charged)
• Energy Drain: 10 points per message
• Energy Refill: 0.5 points per 100ms
• Boost Cooldown: 10 seconds
• Idle Threshold: 30 seconds (for nap timer)

Project Structure

napgpt/
  src/
    app/
      api/
        chat/route.ts       # Chat API endpoint (POST/PUT)
        mode/route.ts       # Mode detection endpoint (GET)
      layout.tsx            # Root layout
      page.tsx              # Main page
      globals.css           # Global styles
    components/
      ChatWindow.tsx        # Main chat interface
      MessageBubble.tsx     # Message display component
      EffortBar.tsx         # Effort controls
      EnergyMeter.tsx       # Energy indicator
      IdleOverlay.tsx       # Nap overlay
      BlanketOverlay.tsx     # Blanket mode overlay
    lib/
      llm/
        adapter.ts          # LLM interface + factory
        openai.ts           # OpenAI implementation
        mock.ts             # Mock LLM (no network)
      nap/
        engine.ts           # Effort bands, strategies
        tone.ts             # Tone templates
        state.ts            # Zustand store
        utils.ts            # Helper functions
        blanket.ts          # Blanket mode auto-activation
      rate-limit.ts         # Rate limiting logic (Redis/Memory)
  scripts/
    monitor.js              # Dev server health monitor
  tests/
    e2e/
      smoke.spec.ts         # Basic smoke tests
      ui-elements.spec.ts   # Comprehensive UI tests
      ui-elements-cdp.ts    # CDP-based UI tests

Environment Variables

Required

• None - the app runs in mock mode by default

Optional

API Configuration:

• ANTHROPIC_API_KEY (optional): Your Anthropic API key. Get one at console.anthropic.com. If not set, runs in mock mode.
• OPENAI_API_KEY (optional): Your OpenAI API key. Set LLM_PROVIDER=openai to activate.
• NAPGPT_MODEL (optional): Model to use (default: claude-sonnet-4-5 for Anthropic, gpt-4o-mini for OpenAI)
• NAPGPT_MAX_TOKENS (optional): Max tokens per response (default: 300)

Rate Limiting (Production — Required for serverless):

• KV_URL (optional): Redis connection URL (e.g., from Upstash)
• KV_REST_API_URL (optional): Upstash Redis REST API URL
• KV_REST_API_TOKEN (optional): Upstash Redis REST API Token

Note: Rate limiting is per-instance only without Redis configured. On serverless platforms like Vercel, each cold-start instance is independent, so a single IP can bypass limits across instances. Configure Redis (KV_* vars) for production.

Feature Configuration:

• NEXT_PUBLIC_NAPGPT_BLANKET_IDLE_MS (optional): Idle time in ms before blanket activates (default: 30000)
• NEXT_PUBLIC_NAPGPT_BLANKET_EFFORT_THRESH (optional): Effort threshold below which blanket activates (default: 20)
• NEXT_PUBLIC_NAPGPT_BEAN_MAX (optional): Maximum coffee beans (default: 10)

API Endpoints

POST /api/chat

Sends a chat message and receives a response from NapGPT.

Request Body:

{
  "messages": [
    { "role": "user", "content": "Hello" }
  ],
  "effort": 50,
  "flags": {
    "dream": false
  }
}

Response:

{
  "reply": "Response text...",
  "meta": {
    "strategy": "short-answer",
    "effort": 50,
    "gaveUp": false,
    "nonSequitur": false
  }
}

Rate Limiting: 10 requests per minute per IP address (100 requests/minute in test mode).

PUT /api/chat

Sets a boost cookie for the next chat request.

Request Body:

{
  "boost": 20
}

GET /api/mode

Returns the current mode (mock or real API).

Response:

{
  "isMock": true
}

Tech Stack

• Next.js 15 (App Router)
• TypeScript
• Tailwind CSS + custom cozy theme
• Framer Motion for animations
• Zustand for state management
• Anthropic SDK (primary) / OpenAI SDK (optional)
• Playwright for e2e testing

License

MIT

---

Note: This is a parody project. NapGPT is intentionally lazy and may not always be helpful. That's the point! 😴