pi-xiaomi-mimo-provider

Xiaomi Mimo provider extension for the pi coding agent. Registers Xiaomi Mimo models using the OpenAI-compatible API.

Models

Model	Context (default)	Max Context	Max Output	Reasoning	Input	Input $/MTok	Output $/MTok
MiMo V2 Pro	128K	1M	16K	✅	Text, Image	$1.00	$3.00
MiMo V2 Omni	128K	1M	16K	✅	Text, Image	$1.00	$3.00
MiMo V2 TTS	128K	1M	16K	❌	Text	Free	Free

Installation

# From GitHub
pi install git:github.com/ulusoyomer/pi-xiaomi-mimo-provider

# From npm
pi install npm:pi-xiaomi-mimo-provider

Setup

After installing, authenticate with your Xiaomi Mimo API key:

/login xiaomi-mimo

Get your API key at: https://platform.xiaomimimo.com/#/console/plan-manage

Alternative: Environment Variable

You can also set your API key via environment variable:

export XIAOMI_MIMO_API_KEY="your-api-key-here"

Or add it to ~/.pi/agent/models.json:

{
  "providers": {
    "xiaomi-mimo": {
      "baseUrl": "https://token-plan-sgp.xiaomimimo.com/v1",
      "api": "openai-completions",
      "apiKey": "YOUR_API_KEY",
      "models": [
        { "id": "mimo-v2-pro" },
        { "id": "mimo-v2-omni" },
        { "id": "mimo-v2-tts" }
      ]
    }
  }
}

Usage

Switch to a Xiaomi Mimo model:

/model mimo-v2-pro

API Endpoints

Protocol	Base URL
OpenAI Compatible	https://token-plan-sgp.xiaomimimo.com/v1
Anthropic Compatible	https://token-plan-sgp.xiaomimimo.com/anthropic

This extension uses the OpenAI-compatible endpoint.

Context Window

The default context window is 128K (131072 tokens) to keep credit usage low. You can increase or change it at any time.

Runtime Command

Use /mimo-context to view or change the context window without restarting:

/mimo-context           # Show current value
/mimo-context 256k      # Set to 256K
/mimo-context 512k      # Set to 512K
/mimo-context 1m        # Set to 1M (maximum)
/mimo-context 131072    # Set with a plain number

Environment Variable

Set the context window at startup via environment variable:

export XIAOMI_MIMO_CONTEXT_WINDOW="256k"

Supported formats: 64k, 128k, 256k, 512k, 1m, or a plain number (max 1048576).

Reasoning Control

Toggle reasoning on/off with pi's thinking commands:

/thinking off       # Reasoning OFF
/thinking minimal   # Minimal reasoning
/thinking low       # Low reasoning
/thinking medium    # Medium reasoning
/thinking high      # Max reasoning

Applies to mimo-v2-pro and mimo-v2-omni only. mimo-v2-tts does not support reasoning.

Notes

• The API returns reasoning_content in responses, which pi handles natively for reasoning models
• Default context window is 128K tokens (131072) — configurable via XIAOMI_MIMO_CONTEXT_WINDOW env var or /mimo-context command
• Maximum supported context window is 1M tokens (1048576)
• Cost values based on official Xiaomi pricing: $1/$3 per MTok (up to 256K context), $2/$6 (256K-1M context)
• Cache write is currently free per Xiaomi's pricing page

License

MIT