relayai 1.0.0

Claude thinks. Gemini speaks. Save up to 80% on AI API costs.

⚡ RelayAI

Claude thinks. Gemini speaks. Save up to 80% on AI API costs.

RelayAI is an open-source terminal AI tool (like Claude Code) where Claude plans and reasons, then passes the answer to Gemini to output — dramatically reducing your Claude API costs by offloading output tokens to Gemini's free tier.

User Input → Claude (thinks & reasons) → Gemini (presents output) → Terminal

---

💡 Why RelayAI?

Claude charges $15/million output tokens. Output tokens are 5x more expensive than input.

RelayAI flips this: Claude produces a concise internal plan/answer, then Gemini (free tier) handles the expensive output step.

	Normal Claude	RelayAI
Per query (avg)	$0.031	$0.006
100 queries/day	~$95/mo	~$18/mo
1000 queries/day	~$945/mo	~$180/mo
Savings		~80%

---

🚀 Install

pip install relayai

Or from source:

git clone https://github.com/yourusername/relayai
cd relayai
pip install -e .

---

⚙️ Setup

relayai login

You'll be asked to provide:

1. Claude API Key — from console.anthropic.com
2. Gemini access — either:
   • Gemini API Key (from ai.google.dev — free)
   • Google Account Login (OAuth — no key needed)

---

🧑‍💻 Usage

# Ask a single question
relayai "write a binary search in python"

# Fix code / debug
relayai "fix this bug in my FastAPI route"

# Pipe files
cat main.py | relayai "review this code"

# Interactive chat mode
relayai --chat

# Show Claude's internal reasoning (debug)
relayai --verbose "explain transformers"

# Check your config
relayai status

---

🔁 How It Works

┌─────────────┐
│  User Input │
└──────┬──────┘
       │
       ▼
┌─────────────────────────────────────┐
│  Claude Sonnet (via your API key)   │
│  • Understands the full request     │
│  • Reasons step by step             │
│  • Produces complete answer         │
│  • Uses minimal output tokens       │
└──────────────┬──────────────────────┘
               │ Claude's answer becomes
               │ Gemini's INPUT (free)
               ▼
┌─────────────────────────────────────┐
│  Gemini Flash (free tier)           │
│  • Receives Claude's answer         │
│  • Presents it clearly              │
│  • All output tokens are FREE       │
└──────────────┬──────────────────────┘
               │
               ▼
┌─────────────┐
│  Terminal   │
│  Output     │
└─────────────┘

---

📂 Project Structure

relayai/
├── relayai/
│   ├── __init__.py
│   ├── cli.py            # Terminal commands
│   ├── auth.py           # Login & credential management
│   ├── claude_client.py  # Claude API (thinker)
│   ├── gemini_client.py  # Gemini API (speaker)
│   └── pipeline.py       # The bridge logic
├── setup.py
├── requirements.txt
└── README.md

---

🤝 Contributing

PRs welcome! This is fully open source. Feel free to:

• Add streaming support
• Add more model options
• Improve the OAuth flow
• Build a config TUI

---

📄 License

MIT — free to use, modify, and distribute.