Just a side project, y'know?
Evaluate multiple LLMs using the command-line. Feed the results into your system of choice, another LLM, or your preferred evaluations framework. Built on the wonderful llm.
Write results to a JSON file for you to consume or review with other tooling.
uv run shootout.pyBy default it will look for a model_config.toml file (see the example included) with your models and API keys. You can pass --config some_other_file.toml to pass it an alternative configuration.
To use models with an llm plugin, install the relevant plugin library for llm to use:
uv pip install llm-anthropic llm-cloudflare llm-deepseekReview the llm plugin directory for the list of supported plugins.
Apache 2.0 licensed. See the LICENSE file for details.