@pandada8/pi-axonhub

Pi extension that discovers AxonHub models from /v1/models and /v1/models?include=all, enriches them with cached metadata from https://models.dev/api.json, and registers them as the axonhub provider.

AxonHub models are cached at ~/.cache/pi/axonhub-models.json for one day. models.dev metadata is cached at ~/.cache/pi/models-dev-api.json for one day. If no API key is configured, the extension does not register the provider.

Usage

Install the package into Pi's global settings:

pi install npm:@pandada8/pi-axonhub

This writes the package to ~/.pi/agent/settings.json under packages. You can also edit it manually:

{
  "packages": ["npm:@pandada8/pi-axonhub"]
}

Configure AxonHub and run Pi:

export AXONHUB_BASE_URL=http://localhost:8090
export AXONHUB_API_KEY=ah-your-api-key
pi

You can also store the key in ~/.pi/agent/auth.json:

{
  "axonhub": {
    "type": "api_key",
    "key": "ah-your-api-key"
  }
}

When using auth.json, AXONHUB_API_KEY is not required. AXONHUB_BASE_URL is optional and defaults to http://localhost:8090.

For local development, point Pi directly at this checkout:

pi -e /path/to/pi-axonhub

OpenAI-compatible models are sent to AxonHub /v1. Anthropic-owned models are sent to AxonHub /anthropic. Gemini-owned models are sent to AxonHub /gemini.