Open WebUI with LiteLLM Setup

This project sets up an Open WebUI interface with LiteLLM as a backend proxy for various AI models. It uses Docker Compose to orchestrate the services.

Prerequisites

• Docker and Docker Compose installed on your system
• API keys for the AI models you want to use

Configuration

1. Create a .env file in the project root with the following content:

   MASTER_KEY=your_master_key #required
   ANTHROPIC_API_KEY=your_anthropic_api_key
   OPENAI_API_KEY=your_openai_api_key
   DEEPSEEK_API_KEY=your_deepseek_api_key
   CODESTRAL_API_KEY=your_codestral_api_key
   GROQ_API_KEY=your_groq_api_key
   

   Replace your_*_api_key with your actual API keys.

2. Ensure the config.yml file is present in the project root. This file configures the available models for LiteLLM. Feel free to add more models here!

Usage

1. Start the services:

   docker-compose up -d

2. Access the Open WebUI interface at http://localhost:3000

3. Access LiteLLM OpenAPI page at http://localhost:4000

Services

• Open WebUI: Runs on port 3000, provides the user interface.
• LiteLLM: Runs on port 4000, acts as a proxy for various AI models.

Available Models

The following models are configured in config.yml:

• claude-3-5-sonnet (Anthropic)
• gpt-4o (OpenAI)
• gpt-4o-mini (OpenAI)
• deepseek-coder (DeepSeek)
• deepseek-chat (DeepSeek)
• codestral (Codestral)
• groq-llama-3.1-405b (Groq)
• groq-llama-3.1-70b (Groq)
• groq-llama-3.1-8b (Groq)

Security Note

This setup uses environment variables to manage API keys. Ensure that your .env file is not committed to version control and is properly secured.

Troubleshooting

If you encounter any issues, check the Docker logs for each service: