Pi OpenRouter Extension

OpenRouter provider extension for Pi Coding Agent.

Features

• 🚀 Access 20+ LLM models through OpenRouter's unified API
• 🔄 OpenAI-compatible streaming
• 🤖 Supports Claude, GPT-4, Gemini, Llama, Mistral, DeepSeek, Perplexity
• 🆓 Free models available (Gemini 2.0 Flash)
• 📦 One-command installation via npm
• ⚡ Auto-installs to ~/.pi/agent/extensions/

Installation

Quick Install (Recommended)

npm install -g @verioussmith/pi-openrouter

That's it! The extension automatically installs to ~/.pi/agent/extensions/openrouter.ts

Alternative Methods

See INSTALL.md for manual installation options.

Configuration

Set your OpenRouter API key in ~/.zshrc or environment:

export OPENROUTER_API_KEY="sk-or-v1-YOUR_KEY_HERE"

Get your key at: https://openrouter.ai/settings/keys

Usage

# Start pi
pi

# List available models
/model

# Select an OpenRouter model
/model openrouter/google/gemini-2.0-flash-exp:free

# Or any other model
/model openrouter/anthropic/claude-3.5-sonnet
/model openrouter/openai/gpt-4o
/model openrouter/deepseek/deepseek-r1

Available Models

Free Tier

• google/gemini-2.0-flash-exp:free - Gemini 2.0 Flash Experimental
• google/gemini-2.0-flash-thinking-exp:free - Gemini 2.0 Flash with reasoning

Anthropic

• anthropic/claude-3.5-sonnet - Claude 3.5 Sonnet
• anthropic/claude-3-opus - Claude 3 Opus
• anthropic/claude-3-haiku - Claude 3 Haiku

OpenAI

• openai/gpt-4o - GPT-4o
• openai/gpt-4o-mini - GPT-4o Mini
• openai/o1 - O1 (reasoning model)
• openai/o1-mini - O1 Mini

Google

• google/gemini-pro-1.5 - Gemini Pro 1.5

Meta

• meta-llama/llama-3.3-70b-instruct - Llama 3.3 70B
• meta-llama/llama-3.1-405b-instruct - Llama 3.1 405B

Mistral

• mistralai/mistral-large - Mistral Large
• mistralai/mistral-small - Mistral Small

DeepSeek

• deepseek/deepseek-chat - DeepSeek Chat
• deepseek/deepseek-r1 - DeepSeek R1 (reasoning)

Perplexity

• perplexity/sonar-pro - Sonar Pro

See all models: https://openrouter.ai/models

Adding More Models

Edit openrouter.ts and add to the models array:

{
  id: "provider/model-name",
  name: "Display Name (OpenRouter)",
  reasoning: false, // true for reasoning models
  input: ["text", "image"], // or just ["text"]
  cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
  contextWindow: 128000,
  maxTokens: 4096,
}

Costs are in $/million tokens. Check https://openrouter.ai/models for pricing.

Development

# Clone
git clone https://github.com/verioussmith/pi-openrouter-extension.git
cd pi-openrouter-extension

# Install in Pi (development mode)
ln -s $(pwd)/openrouter.ts ~/.pi/agent/extensions/openrouter.ts

# Test
pi
/model openrouter/google/gemini-2.0-flash-exp:free

Publishing Updates

# Update version in package.json
npm version patch  # or minor, major

# Commit and push
git add . && git commit -m "chore: bump version" && git push

# Publish to npm (requires OTP)
npm publish --access public --otp=YOUR_OTP

Package Links

• npm: https://www.npmjs.com/package/@verioussmith/pi-openrouter
• GitHub: https://github.com/verioussmith/pi-openrouter-extension
• Issues: https://github.com/verioussmith/pi-openrouter-extension/issues

References

• Pi Coding Agent
• Pi Custom Provider Docs
• OpenRouter API
• OpenRouter Models

Contributing

Contributions welcome! Please:

1. Fork the repo
2. Create a feature branch
3. Submit a pull request

Author

Verious Smith (@verioussmith)

• Website: https://philoveracity.com
• Email: [email protected]

License

MIT - see LICENSE file for details