title: Inference Playground
emoji: 🔋
colorFrom: blue
colorTo: pink
sdk: docker
pinned: false
app_port: 3000

This application provides a user interface to interact with various large language models, leveraging the @huggingface/inference library. It allows you to easily test and compare models hosted on Hugging Face, connect to different third-party Inference Providers, and even configure your own custom OpenAI-compatible endpoints.

Local Setup

TL;DR: After cloning, run pnpm i && pnpm run dev --open

Prerequisites

Before you begin, ensure you have the following installed:

• Node.js: Version 20 or later is recommended.
• pnpm: Install it globally via npm install -g pnpm.
• Hugging Face Account & Token: You'll need a free Hugging Face account and an access token to interact with models. Generate a token with at least read permissions from hf.co/settings/tokens.

Follow these steps to get the Inference Playground running on your local machine:

1. Clone the Repository:

   git clone https://github.com/huggingface/inference-playground.git
   cd inference-playground
   

2. Install Dependencies:

   pnpm install
   

3. Start the Development Server:

   pnpm run dev
   

4. Access the Playground:

   • Open your web browser and navigate to http://localhost:5173 (or the port indicated in your terminal).

Features

• Model Interaction: Chat with a wide range of models available through Hugging Face Inference.
• Provider Support: Connect to various third-party inference providers (like Together, Fireworks, Replicate, etc.).
• Custom Endpoints: Add and use your own OpenAI-compatible API endpoints.
• Comparison View: Run prompts against two different models or configurations side-by-side.
• Configuration: Adjust generation parameters like temperature, max tokens, and top-p.
• Session Management: Save and load your conversation setups using Projects and Checkpoints.
• Code Snippets: Generate code snippets for various languages to replicate your inference calls.
• Organization Billing: Specify an organization to bill usage to for Team and Enterprise accounts.

Organization Billing

For Team and Enterprise Hugging Face Hub organizations, you can centralize billing for all users by specifying an organization to bill usage to. This feature allows:

• Centralized Billing: All inference requests can be billed to your organization instead of individual user accounts
• Usage Tracking: Track inference usage across your organization from the organization's billing page
• Spending Controls: Organization administrators can set spending limits and manage provider access

How to Use Organization Billing

1. In the UI: Navigate to the settings panel and enter your organization name in the "Billing Organization" field
2. In Code Snippets: Generated code examples will automatically include the billing organization parameter
3. API Integration: The playground will include the X-HF-Bill-To header in API requests when an organization is specified

Requirements

• You must be a member of a Team or Enterprise Hugging Face Hub organization
• The organization must have billing enabled
• You need appropriate permissions to bill usage to the organization

For more information about organization billing, see the Hugging Face documentation.

We hope you find the Inference Playground useful for exploring and experimenting with language models!