This example demonstrates how to use LaunchDarkly's create_judge method to evaluate specific input/output pairs directly, without an associated chat session.
- Python 3.10 or higher
- Poetry installed
- A LaunchDarkly account and SDK key
- API keys for the provider you want to use (OpenAI, Bedrock, or Gemini)
-
Create the following config in your LaunchDarkly project. You can use a different key by setting the environment variable in your
.env.- Create a Judge Config for evaluation. Default key:
sample-judge.
- Create a Judge Config for evaluation. Default key:
-
Copy
.env.exampleto.envand fill in your keys:cp .env.example .env
-
Install the required dependencies:
poetry install
poetry run judge