# ai-router 🚀

Ultra-light AI API Router - Zero deps, single file, auto failover

---

Why ai-router?

Have you ever:

• 🤯 An AI API goes down and your app crashes with it
• 😤 Free API quota is exhausted, no auto-fallback to paid
• 🔄 Want to use multiple AI providers without rewriting integration for each
• 📦 Just want to make an API call without installing 500MB of dependencies

ai-router is your answer. One file, zero dependencies, deploy anywhere.

---

Quick Start

Install

pip install git+https://github.com/longongzi/ai-router.git

Or just download the single file:

curl -O https://raw.githubusercontent.com/longongzi/ai-router/main/ai_router.py

Start

export DEEPSEEK_API_KEY="sk-your-key-here"
ai-router serve

Use

OpenAI-compatible API:

from openai import OpenAI
client = OpenAI(base_url="http://localhost:8080/v1", api_key="-")
response = client.chat.completions.create(
    model="gpt-4o-mini",  # auto-mapped to deepseek-chat
    messages=[{"role": "user", "content": "Hello!"}],
)

Or curl:

curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "gpt-4o-mini", "messages": [{"role": "user", "content": "Hello"}]}'

---

Features

Feature	Description
🪶 Zero deps	Pure Python stdlib, runs anywhere
🔄 Auto failover	Primary down? Switch to backup
🎯 Model aliases	Drop-in replacement for any OpenAI SDK
🤝 Multi-provider	DeepSeek, OpenClaw, any OpenAI-compatible API
⚡ Streaming	Native SSE streaming
🐳 Single file	One .py, no node_modules, no venv hell
📦 CLI tools	Built-in CLI for testing

---

Why ai-router?

vs	ai-router	Alternative
LiteLLM	1 file, 0 deps	~50MB deps
OpenRouter	Self-hosted	Hosted, privacy concerns
Glue code	One solution	Per-project wrappers

---

Roadmap

• Local model support (llama.cpp / Ollama)
• Load balancing
• Request caching
• Rate limiting
• Docker image

---

Sponsor

Every sponsorship fuels continued development 🙏

---

MIT License © 2026 longongzi

Built with ❤️ and ☕