✦ ChaqGPT

A sleek, free AI chat interface powered by the world's best open-source models.

Live Demo · Report Bug · Request Feature

---

📸 Screenshots

Welcome Screen

Math & LaTeX Rendering

Thinking Visualization

Sidebar & Chat Search

---

✨ Features

Category	Details
Multi-Model Access	30+ free models — Qwen3 VL, DeepSeek R1, Llama 3.3 70B, Gemma 3, Mistral, and more
Vision Support	Upload images and get AI analysis with vision-enabled models
Thinking Visualization	Watch reasoning models "think" in real-time with expandable thought chains
Rich Rendering	Full Markdown, LaTeX math (KaTeX), syntax-highlighted code blocks with copy
Chat History	Persistent conversations stored locally — pick up where you left off
Dark & Light Themes	Toggle between themes; follows system preference by default
Temporary Chat	One-off conversations that aren't saved to history
Mobile Optimized	Responsive design with keyboard-aware viewport handling
Streaming Responses	Token-by-token streaming with adaptive 60fps render queue
Zero Framework	Pure HTML/CSS/JS — fast, lightweight, no build step required

---

🛠️ Tech Stack

• Frontend — Vanilla HTML, CSS, JavaScript (no framework overhead)
• Backend — Vercel Serverless Functions (Node.js)
• AI Provider — OpenRouter (free-tier models)
• Rendering — Marked + KaTeX + Highlight.js
• Design — Custom CSS with spring animations, glassmorphism, and safe-area support

---

🚀 Quick Start

Prerequisites

• Node.js 18+
• An OpenRouter API key (free)

Local Development

# Clone the repo
git clone https://github.com/Amer-alsayed/chaqgpt.git
cd chaqgpt

# Install Vercel CLI
npm i -g vercel

# Create .env file (multi-key preferred, single-key fallback supported)
echo OPENROUTER_API_KEYS_JSON='["sk-or-1","sk-or-2","sk-or-3"]' > .env
# Optional fallback
echo OPENROUTER_API_KEY=your_key_here >> .env

# Start the dev server
vercel dev

The app will be available at http://localhost:3000.

---

📦 Deploy to Vercel

1. Push your code to GitHub
2. Import the repo in Vercel
3. Add environment variables:

   Variable	Value
   OPENROUTER_API_KEYS_JSON	JSON array of OpenRouter keys (recommended), e.g. ["sk-or-1","sk-or-2","sk-or-3"]
   OPENROUTER_API_KEY	Single-key fallback (optional when JSON var is present)
   OPENROUTER_KEY_COOLDOWN_RATE_LIMIT_MS	Optional. Default 900000 (15 min)
   OPENROUTER_KEY_COOLDOWN_AUTH_MS	Optional. Default 21600000 (6 h)
   OPENROUTER_KEY_COOLDOWN_TRANSIENT_MS	Optional. Default 30000 (30 s)
   OPENROUTER_MAX_FAILOVER_ATTEMPTS	Optional. Default 10

4. Deploy — that's it!

Tip

The app automatically fetches all available free models from OpenRouter at runtime. No config changes needed when new models are added.

---

🤖 Default Model

New users start with Qwen3 VL 30B A3B Thinking — a powerful vision + reasoning model. Users can switch models anytime via the header dropdown, and their preference is saved locally.

---

📁 Project Structure

chaqgpt/
├── api/
│   ├── chat.js          # Chat completion proxy (streaming)
│   └── models.js        # Fetches & filters free models from OpenRouter
├── assets/
│   ├── css/style.css    # Complete design system
│   └── js/
│       ├── app.js       # Core application logic
│       └── config.js    # Welcome messages & suggestion sets
├── index.html           # Single-page application entry
├── vercel.json          # Vercel deployment config
└── .env                 # API key (not committed)

---

🔒 Security

• API keys are never exposed to the client
• All API calls route through serverless functions
• .env is gitignored by default

---

📄 License

This project is licensed under the MIT License.

---

_{Built with ☕ by Amer Alsayed}