# Genkit Hugging Face Plugin
This Genkit plugin provides integration with Hugging Face's Inference API and
Inference Providers, giving access to 1,000,000+ models through a unified interface.
## Installation
```bash
pip install genkit-plugin-huggingface
```
Or with uv:
```bash
uv add genkit-plugin-huggingface
```
## Setup
Set your Hugging Face token:
```bash
export HF_TOKEN=your-token
```
Get your token from: https://huggingface.co/settings/tokens
## Usage
```python
from genkit import Genkit
from genkit.plugins.huggingface import HuggingFace
ai = Genkit(
plugins=[HuggingFace()],
model='huggingface/meta-llama/Llama-3.3-70B-Instruct',
)
response = await ai.generate(prompt='Hello, Hugging Face!')
print(response.text)
```
## Popular Models
| Model | Description |
|-------|-------------|
| `huggingface/meta-llama/Llama-3.3-70B-Instruct` | Meta's latest Llama model |
| `huggingface/mistralai/Mistral-Small-24B-Instruct-2501` | Mistral's efficient model |
| `huggingface/Qwen/Qwen2.5-72B-Instruct` | Alibaba's multilingual model |
| `huggingface/deepseek-ai/DeepSeek-R1` | DeepSeek's reasoning model |
| `huggingface/google/gemma-2-27b-it` | Google's open Gemma model |
| `huggingface/microsoft/phi-4` | Microsoft's compact Phi model |
## Features
- **Text Generation**: Standard chat completions
- **Streaming**: Real-time token streaming
- **Inference Providers**: Access 17+ providers (Cerebras, Groq, Together, etc.)
- **Any Model**: Use any of 1M+ models on Hugging Face Hub
## Configuration
```python
from genkit.plugins.huggingface import HuggingFace
# With explicit token
ai = Genkit(plugins=[HuggingFace(token='your-token')])
# With specific inference provider
ai = Genkit(plugins=[HuggingFace(provider='cerebras')])
```
## Inference Providers
Hugging Face supports routing requests to different inference providers for
better performance or cost optimization:
- Cerebras
- Groq
- Together
- Fireworks
- Replicate
- And more...
```python
# Use a specific provider
ai = Genkit(plugins=[HuggingFace(provider='groq')])
```
## Environment Variables
| Variable | Description | Required |
|----------|-------------|----------|
| `HF_TOKEN` | Your Hugging Face API token | Yes |
## Notes
- Free tier has rate limits; consider [HF Pro](https://huggingface.co/pricing) for higher limits
- Some models require accepting terms on huggingface.co first
- Model IDs use the format `owner/model-name` (e.g., `meta-llama/Llama-3.3-70B-Instruct`)
## Links
- [Hugging Face Hub](https://huggingface.co/)
- [Inference API Documentation](https://huggingface.co/docs/api-inference/)
- [Inference Providers](https://huggingface.co/docs/inference-providers/)
- [Genkit Documentation](https://genkit.dev/)
