# @royalrat-pack/pi-extension-nvidia-build-provider

[![npm version](https://img.shields.io/npm/v/@royalrat-pack/pi-extension-nvidia-build-provider.svg)](https://www.npmjs.com/package/@royalrat-pack/pi-extension-nvidia-build-provider)
[![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](./LICENSE)

A Pi extension that adds NVIDIA Build API models as a first-class provider with native `/login` support and `/model` selection.

## What this extension does

- Registers provider: `nvidia-build`
- Adds login option in Pi: **NVIDIA Build API**
- Stores API key in Pi auth storage after one-time paste
- Enables model selection through `/model`
- Supports model auto-discovery via NVIDIA `/v1/models`

## Installation

### Option 1: Install from npm (recommended)

```bash
pi install npm:@royalrat-pack/pi-extension-nvidia-build-provider
```

### Option 2: Install from GitHub

```bash
pi install git:github.com/debarch777/pi-extension-nvidia-build-provider
```

### Option 3: Local development install

```bash
pi install /absolute/path/to/pi-extension-nvidia-build-provider
```

## Usage

Inside Pi:

```text
/login
# select: NVIDIA Build API
# paste API key once

/model
# select nvidia-build/<model-id>
```

## Optional environment variables

- `NVIDIA_BASE_URL` (default: `https://integrate.api.nvidia.com/v1`)
- `NVIDIA_MODELS` (comma-separated model IDs; skips auto-discovery)
- `NVIDIA_INCLUDE_ALL_MODELS=1` (includes non-chat model IDs too)
- `NVIDIA_API_KEY` or `NVCF_API_KEY` (optional startup fallback)

## Package

- npm: `@royalrat-pack/pi-extension-nvidia-build-provider`
- latest published: `0.1.0`

## Development

```bash
git clone https://github.com/debarch777/pi-extension-nvidia-build-provider.git
cd pi-extension-nvidia-build-provider
pi -e .
```

## Security note

Only install Pi extensions from sources you trust. Extensions execute with your user permissions.

## License

MIT