Download the PHP package symfony/ai-hugging-face-platform without Composer

HuggingFace Bridge

The HuggingFace bridge provides integration with Hugging Face Inference Providers, enabling access to thousands of pre-trained models for various AI tasks including natural language processing, computer vision, audio processing, and more.

HuggingFace Documentation

• Inference Providers overview
• Tasks reference
• Authentication / access tokens

Features

• Multi-Provider Support: Access models through multiple inference providers (HuggingFace Inference, Cerebras, Cohere, Groq, Together, and others)
• Comprehensive Task Support: Support for 40+ different AI tasks including:
  • Chat completion and text generation
  • Image classification, object detection, and segmentation
  • Automatic speech recognition and audio classification
  • Text classification, translation, summarization
  • Feature extraction and embeddings
  • And many more...
• Model Discovery: Built-in API client to discover and query available models
• Flexible Input/Output: Support for text, images, audio, and binary data
• Type-Safe Results: Structured result objects for each task type

Installation

Install the symfony/ai-hugging-face-platform package:

Quick Start

Initialize the Platform

Create a Hugging Face access token with the Make calls to Inference Providers permission and use it as the API key.

The HTTP client is optional; a default client is used when none is provided. Choose a provider that serves the model and task you want to use.

Chat Completion Example

Image Classification Example

Feature Extraction (Embeddings)

Providers

The bridge supports multiple inference providers. For a complete list of available providers and their constants, see the Provider interface.

Specify a provider when creating the platform:

Supported Tasks

The bridge supports 40+ AI tasks across multiple categories including:

• Natural Language Processing: Chat completion, text generation, classification, translation, summarization, embeddings, and more
• Computer Vision: Image classification, object detection, segmentation, depth estimation, and more
• Audio: Speech recognition, audio classification, text-to-speech, and more
• Multimodal: Visual question answering, document understanding, and more

For the complete list of supported tasks and their constants, see the Task interface.

Model Discovery

The bridge includes commands to discover available models without using inference credits. They help to research and use models, and can be registered in a Symfony Console application.

Command-Line Interface

The HuggingFace bridge provides two console commands for model discovery:

List Models

List available models with optional filtering:

Options:

• --provider, -p: Filter by inference provider (e.g., hf-inference)
• --task, -t: Filter by task type (e.g., text-generation)
• --search, -s: Search term to filter models
• --warm, -w: Only list models with warm inference status (served by an inference provider)

Examples:

Get Model Information

Retrieve detailed information about a specific model:

Examples:

Output includes:

• Model ID, downloads, and community likes
• Task type (pipeline tag)
• Whether the model has warm inference status (served by an inference provider); an absent status does not indicate a cold model
• Inference provider mappings and availability

Options and Configuration

When invoking the platform, you can pass task-specific options:

You can also override the provider per request: