Zepiru

← Back to Chat

About Zepiru

A powerful multi-model AI chat platform that lets you compare responses from different AI models side by side and watch them have conversations with each other.

Note: This application was developed by a human with zero prior web development experience, using collaborative AI tools including Claude Code, Gemini CLI, Cursor IDE, and Claude 4 Sonnet. Even this about page was generated by an LLM as part of the development process.

🤖 Multi-Model Chat

Compare responses from up to 2 AI models simultaneously:

  • Free Models: OpenRouter Auto, Llama 4 Maverick, Mistral Small 3.2, DeepSeek Chat
  • Premium Models: Claude Sonnet 4, Gemini 2.5 Flash, GPT-4.1 Mini, Grok 4
  • Easy model switching with visual toggles
  • Real-time response comparison
  • Split-screen view for side-by-side analysis

🎭 Dual Model Mode

Select two models and compare their responses to the same prompt:

  • Automatic UI switching between single and dual modes
  • Parallel processing for faster responses
  • Independent model responses (no cross-talk)
  • Clean model attribution with headers
  • Seamless mode transitions

🔐 Authentication

Secure access to premium models with zero password risk:

  • Google OAuth integration - no passwords stored
  • Nothing to hack - we don't keep usernames or passwords
  • Perfectly safe to use - Google handles all security
  • Premium model access for authenticated users
  • Anonymous usage with free models
  • User profile management
  • Session persistence across browser sessions

⚡ Real-time Performance

Lightning-fast responses with edge computing:

  • Global CDN distribution for minimal latency
  • Parallel API calls for dual models
  • Edge computing with Cloudflare Workers
  • Optimized database queries
  • Instant UI updates and transitions
  • Responsive design for all devices
  • Automatic scaling with zero server management

🎨 Clean Interface

Apple-inspired minimalist design:

  • Intuitive sidebar for model selection
  • Clear visual indicators for active models
  • Responsive design for all screen sizes
  • Dark mode support
  • Focus on content, not clutter

⚡ Performance

Built for speed and reliability:

  • Edge computing with Cloudflare Workers
  • Global CDN distribution
  • Parallel API calls for dual models
  • Optimized database queries
  • Minimal latency for real-time chat

🛠️ Technology Stack

Built with modern web technologies for speed and reliability:

Cloudflare Workers
OpenRouter API
Cloudflare D1 Database
Vanilla JavaScript
Modern CSS
Edge Computing
Google OAuth
SQLite (D1)

🏗️ Architecture Overview

Frontend: Single-page application built with vanilla JavaScript, HTML5, and CSS3. No frameworks or build tools required - pure web standards for maximum compatibility and performance.

Backend: Serverless architecture using Cloudflare Workers, providing global edge computing with automatic scaling and zero server management.

Database: Cloudflare D1 (SQLite) for persistent storage of users, sessions, and chat messages with automatic backups and global distribution.

API Integration: OpenRouter.ai as the primary LLM provider, offering access to 100+ models from providers like OpenAI, Anthropic, Google, and Meta.

Authentication: Google OAuth 2.0 for secure user authentication without storing passwords.

🔧 How It Works

Model Selection: Choose up to 2 models from the sidebar. Free models are available to all users, while premium models require Google authentication.

Dual Mode: When two models are selected, the interface automatically switches to split-screen mode, displaying responses side-by-side for easy comparison.

Message Flow: Your messages are sent to OpenRouter's API, which routes them to the selected AI models. Responses are streamed back and displayed with model attribution.

Data Storage: Chat sessions and messages are stored in Cloudflare D1 with minimal metadata (timestamp, model used, IP for rate limiting).

Session Management: Each chat session is assigned a unique identifier, allowing for conversation persistence and potential future features like chat history.

🔒 Privacy & Security

Data Collection: We only store chat messages, timestamps, and basic session info. No personal data beyond IP addresses for rate limiting.

API Security: All AI model requests go through our Cloudflare Worker, keeping API keys secure on the server side.

User Privacy: Google OAuth provides secure authentication without sharing sensitive information. User data is minimal and focused on functionality.

Open Source: The codebase is transparent and can be audited for security and privacy practices.

🤝 Collaborative Development Process

Human-AI Collaboration: This application was developed through a unique collaboration between a human developer (with no prior web development experience) and multiple AI tools:

Development Workflow: The typical development process involves iterative cycles of human direction, AI code generation, testing, and refinement. This collaborative approach enabled rapid development despite the human developer's lack of prior experience.

Version Control: Code changes are managed through Git with automated deployment to Cloudflare Workers, demonstrating how AI tools can handle complex deployment workflows.

📊 Database Schema

The application uses a simple but effective database structure:

-- Users table for Google OAuth CREATE TABLE users ( id TEXT PRIMARY KEY, email TEXT, name TEXT, picture TEXT, created_at INTEGER, last_login INTEGER ); -- Chat sessions CREATE TABLE chat_sessions ( id TEXT PRIMARY KEY, user_id TEXT, created_at INTEGER, last_active INTEGER, user_ip TEXT, user_agent TEXT ); -- Chat messages CREATE TABLE chat_messages ( id TEXT PRIMARY KEY, session_id TEXT, message_type TEXT, content TEXT, model_id TEXT, model_name TEXT, timestamp INTEGER );

Ready to Start Chatting?

Experience multi-model AI conversations and see how different models approach the same questions.

Start Chatting →