An LLM playground you can run on your laptop.

• Use any model from OpenAI, Anthropic, Cohere, Forefront, HuggingFace, Aleph Alpha, and llama.cpp.
• Full playground UI, including history, parameter tuning, keyboard shortcuts, and logprops.
• Compare models side-by-side with the same prompt, individually tune model parameters, and retry with different parameters.
• Automatically detects local models in your HuggingFace cache, and lets you install new ones.
• Works OK on your phone.
• Probably won't kill everyone.

git clone https://github.com/keldenl/openplayground # or git@github.com:keldenl/openplayground.git for SSH
cd app && npm install # Install requirements
npx parcel watch src/index.html --no-cache # Run the front end

# Open another terminal / cmd window
cd server && pip3 install -r requirements.txt && cd .. # Install requirements
python3 -m server.app # Run the backend

To run openai requests through gpt-llama.cpp, please make sure you have a server instance of gpt-llama.cpp running. If not, please follow this separate guide.

Run openplayground as specified above, and when setting the API KEY remember to set it as the path to your local model. Set the custom base url to http://localhost:443/v1, replace "443" with the port you are using.

gpt-3.5-turbo will produce results in "chat completion" style, while models like text-davinci-003 will produce results as "text completion" style.

Instigated by Nat Friedman. Initial implementation by Zain Huda as a repl.it bounty. Many features and extensive refactoring by Alex Lourenco.