pi-mimo-provider

Xiaomi MiMo LLM provider for the pi coding agent.

Registers MiMo models (V2-Flash, V2-Pro, V2-Omni, 7B-RL) using the OpenAI-compatible chat completions API at https://api.xiaomimimo.com/v1.

Models

Model	Context	Reasoning	Images	Cost (in/out per 1M tok)
MiMo-V2-Flash	256k	✅	❌	$0.10 / $0.30
MiMo-V2-Pro	1M	✅	❌	$1.00 / $3.00
MiMo-V2-Omni	1M	✅	✅	TBD
MiMo-7B-RL	32k	✅	❌	Free

Setup

1. Install

bun add pi-mimo-provider

2. Set your API key

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

3. Use as pi extension

Add to ~/.pi/agent/settings.json:

{
  "packages": ["pi-mimo-provider/extension"]
}

Or via CLI flag:

pi --extension pi-mimo-provider/extension

4. Or use programmatically

import { createMiMoProvider } from "pi-mimo-provider";

// Register with pi
pi.registerProvider("xiaomi-mimo", createMiMoProvider());

// Or with custom options
pi.registerProvider("xiaomi-mimo", createMiMoProvider({
  apiKey: "sk-...",
  baseUrl: "https://custom.endpoint.com/v1",
  headers: { "X-Region": "cn" },
}));

API

createMiMoProvider(options?)

Creates a ProviderConfig for registration with pi.

Options:

Option	Type	Default	Description
apiKey	string	${XIAOMI_MIMO_API_KEY}	API key (supports pi env var syntax)
baseUrl	string	https://api.xiaomimimo.com/v1	API base URL
headers	Record<string, string>	—	Custom HTTP headers

Exports

// Model configs
export const MIMO_V2_FLASH: MiMoModelConfig;
export const MIMO_V2_PRO: MiMoModelConfig;
export const MIMO_V2_OMNI: MiMoModelConfig;
export const MIMO_7B_RL: MiMoModelConfig;
export const MIMO_MODELS: MiMoModelConfig[];

// Constants
export const MIMO_BASE_URL: string;
export const MIMO_API_KEY_ENV: string;

// Factory
export function createMiMoProvider(options?: MiMoProviderOptions): ProviderConfig;

Development

bun install        # Install deps
bun run typecheck  # Type check
bun run test       # Run tests
bun run build      # Build dist/
bun run check      # All of the above

License

MIT