Starting the UI

Reviewing model outputs by hand is an important part of evaluating quality. EP makes this process simple by offering an easy-to-use, locally hosted UI you can set up in minutes.

To start the UI, simply run the following command and open your browser to http://localhost:8000.

CLI

ep logs

Ensure the eval-protocol package is installed so the ep CLI is available.

uv

# Add to your project (installs the `ep` CLI in your environment)
uv add eval-protocol

# Verify installation
uv run ep --version

pip

# Install or upgrade
pip install -U eval-protocol

# Verify installation
ep --version

Once you navigate to the UI at http://localhost:8000, you will see a table of evaluation rows that you can click to inspect.

Whenever you run an @evaluation_test—whether from the VSCode Test Explorer/Debugger or from the CLI via pytest—the UI automatically shows running tests and you can watch rollouts live in the chat interface. When a test finishes, detailed evaluation results appear to the right of the chat. Tests are stored under a SQLite database on your local device at .eval_protocol/logs.db in the root of your Python project.

To run your tests in CLI, you use the pytest command directly.

Run with uv virtual environment

# Run your tests (UI will reflect live rollouts)
uv run pytest

Run with your current Python environment

# Run your tests (UI will reflect live rollouts)
pytest

You can also run tests in your IDE. Once you have your tests running, you can open the UI at http://localhost:8000 to monitor rollouts live.

Next Steps

Checkout the Table View and Pivot View for more information on how to use the UI.