██╗   ██╗███████╗██████╗ ██╗███╗   ██╗ █████╗
██║   ██║██╔════╝██╔══██╗██║████╗  ██║██╔══██╗
██║   ██║█████╗  ██████╔╝██║██╔██╗ ██║███████║
╚██╗ ██╔╝██╔══╝  ██╔══██╗██║██║╚██╗██║██╔══██║
 ╚████╔╝ ███████╗██║  ██║██║██║ ╚████║██║  ██║
  ╚═══╝  ╚══════╝╚═╝  ╚═╝╚═╝╚═╝  ╚═══╝╚═╝  ╚═╝

Meet Verina! An intelligent search engine for you

Quick answers? Search. Curious about something? Chat. Tough research? Agent.

⚡ Fast Mode Quick answers in 5 seconds

🧠 Deep Mode Extended thinking & multi-perspective analysis

💬 Chat Mode Interactive conversation with context

Sustained 18-minute research session generating structured analysis and actionable recommendations

Search → Read → Chat We seamlessly fuse search and chat UI. Search first, click sources to read, then continue chatting with human-in-the-loop interaction.

Agent Mode for Deep Research For complex questions, activate autonomous analysis with efficient tools and external file system for optimized context management.

Lightning Fast Powered by Gemini 2.5 Flash + Exa API for blazing search speed with exceptional quality.

Open Source & Local Perfect alternative to ChatGPT and Perplexity. Build your own browser. BYOK with OpenRouter API (1M free requests/month).

Easy CLI Setup Beginners can launch instantly with one command.

Prerequisites:

• Install Docker Desktop and make sure it's running
• Node.js 18+ installed

1️⃣ Install

npm install -g verina

2️⃣ Configure API Keys

verina init

You'll be prompted for OpenRouter and Exa API keys.

3️⃣ Start

verina

Your browser opens at http://localhost:3000. That's it!

⚠️ First launch takes 5-10 minutes to download Docker images. Subsequent starts take 10-20 seconds.

📖 See CLI documentation for more commands.

git clone https://github.com/YangLi-leo/Verina.git
cd Verina
cp config/.env.example config/.env.development
# Edit config/.env.development with your API keys
verina dev

Your browser opens at http://localhost:3000 with hot reload enabled.

Requirements: Docker Desktop + OpenRouter + Exa keys

🚧 Alpha Release: The frontend optimization needs improvement and may contain bugs. Your feedback helps us improve!

Known Limitations:

• No concurrent sessions: Switching between search/chat sessions will automatically stop the current ongoing task. This prevents message mixing and ensures data consistency.
• To maintain multiple conversations simultaneously, please wait for the current task to complete before switching.

Planned Features:

• Multi-session support with independent concurrent execution
• Enhanced state management for better parallel processing

Please report any issues to GitHub Issues.

Frontend: Next.js 15, React 19, TypeScript Backend: FastAPI, Exa API, OpenRouter Infrastructure: Docker, local file storage

Apache 2.0 - see LICENSE for details.

Copyright 2025 Li Yang