LiteLLM Test Scripts Setup Instructions

This document provides instructions for setting up and running the LiteLLM test scripts.

Prerequisites

• Python 3.7 or later
• pip (Python package manager)

Setup Steps

1. Create a project directory

   mkdir litellm-tests
   cd litellm-tests

2. Create a virtual environment (recommended)

   python -m venv venv
   
   # Windows
   venv\Scripts\activate
   
   # macOS/Linux
   source venv/bin/activate

3. Install required packages

   pip install litellm python-dotenv

4. Create a .env file

   Create a file named .env in your project directory and add your API keys, add keys and model providers as needed:

   OPENAI_API_KEY=your_openai_api_key
   ANTHROPIC_API_KEY=your_anthropic_api_key
   # GOOGLE_API_KEY=your_google_api_key
   

5. Create a directory for test files

   mkdir docs

6. Add test files (optional)

   • For long context tests: Add a long text file (e.g., a transcript) in the docs folder
   • For image tests: Add an image file (e.g., a JPEG) in the docs folder

Running the Test Scripts

Place each of the test scripts in your project directory and run them individually:

python basic-completion.py
python system-message-test.py
python long-context-test.py
python streaming-test.py
python tool-calling-test.py
python image-input-test.py

Modifying the Scripts

• To test different models, update the models list in each script
• Adjust the parameters (e.g., temperature, max_tokens) according to your requirements
• For the long context test, update the file path in the script to your text file

Troubleshooting

• If you encounter API key errors, check that your .env file is correctly formatted and the keys are valid
• For context length errors, try using a smaller text file or a model with a larger context window
• For image input errors, ensure your image file exists and is in a supported format (JPEG, PNG)

Additional Features to Test

The scripts cover the main features mentioned in the transcript:

• Basic completion
• System messages
• Long context
• Streaming
• Tool calling
• Image input

If you want to test additional features:

• Modify parameters (temperature, top_p, etc.)
• Test with different prompts
• Try different model versions