pi-routerai

RouterAI provider extension for Pi — connect your terminal coding agent to hundreds of LLMs via RouterAI's OpenAI-compatible API.

Features

• Dynamic model discovery — Fetches live model list from https://routerai.ru/api/v1/models
• Actual context windows & pricing — Uses context_length and pricing from API
• Capability detection — Vision (image input modality) and reasoning (reasoning parameter) detected automatically
• OpenAI-compatible API — Uses openai-completions streaming (works with all Pi features)
• Embeddings tool — Generate embeddings via /v1/embeddings for RAG and similarity search
• Direct chat tool — Send one-off completions for model comparison or testing
• Authentication — Supports Pi's native ~/.pi/agent/auth.json and environment variables

Supported Models (examples)

Provider	Model ID	Capabilities
OpenAI	openai/gpt-5.2-codex	Reasoning, Vision, 400K ctx
Anthropic	anthropic/claude-sonnet-4.6	Reasoning, Vision
Google	google/gemini-3.1-pro-preview	Reasoning, Vision
DeepSeek	deepseek/deepseek-v3.2	Text only
Qwen	qwen/qwen3-coder-next	Reasoning
Z.ai	z-ai/glm-5	Reasoning, Vision
MiniMax	minimax/minimax-m2.5	Reasoning

Full list is fetched dynamically from RouterAI API.

Installation

# Install via pi
pi install npm:pi-routerai

# Or install locally
pi install npm:pi-routerai -l

Setup

1. Get an API key from routerai.ru (sign up and get API key)
2. Start Pi and run:

   /routerai-login
   

   Paste your API key when prompted. It is stored in Pi's ~/.pi/agent/auth.json.
3. Select a model with /model → pick any routerai/* model

Environment Variables

Variable	Default	Description
ROUTERAI_API_KEY	—	API key (fallback if not in auth.json)
ROUTERAI_BASE_URL	https://routerai.ru/api/v1	Override endpoint
ROUTERAI_TIMEOUT	30000	Model discovery timeout in ms

Usage

Select a model

/model

Then pick any routerai/* model. Examples:

• routerai/openai/gpt-5.2-codex — OpenAI's coding model
• routerai/anthropic/claude-sonnet-4.6 — Anthropic's Claude
• routerai/google/gemini-3.1-pro-preview — Google Gemini

Commands

Command	Description
/routerai-status	Check API key status and model count
/routerai-list	Pretty-print all models with 🧠/🖼️/💬 badges
/routerai-login	Save API key to Pi's auth storage

Tools (LLM-callable)

Tool	Purpose
routerai_list_models	Filter models by family, vision, or reasoning
routerai_embeddings	Generate embeddings via /v1/embeddings
routerai_chat	Direct chat completion via /v1/chat/completions
routerai_model_info	Get detailed metadata for a specific model

Quick Examples

# Check what models are available
Use routerai_list_models to show all available models

# Get embeddings for a document
Use routerai_embeddings with model "openai/text-embedding-3-small" and input "The quick brown fox"

# Compare model outputs
Use routerai_chat with model "openai/gpt-5.2-codex" and messages [{role: "user", content: "Hello"}]
Use routerai_chat with model "anthropic/claude-sonnet-4.6" and messages [{role: "user", content: "Hello"}]

API Compatibility

This extension uses RouterAI's OpenAI-compatible API (/v1/chat/completions), which supports:

• Chat completions with streaming
• Vision (multimodal) inputs
• Tool calling
• JSON mode
• Reasoning/thinking control
• Embeddings (/v1/embeddings)

License

MIT