Agent Evals

Test Agents using datasets to evaluate performance across examples

After testing in the playground, evaluate your Agents across multiple test cases to ensure consistent performance using the test runs.

1. Create a Dataset

   Add test cases by creating a Dataset. For this example, we’ll use a Dataset of product images to generate descriptions.

   
2. Build your Agent

   Create an Agent that processes your test examples. In this case, the agent generates product descriptions, translates them to multiple languages, and formats them to match specific requirements.

   
3. Start a test run

   Open the test configuration by clicking Test in the top right corner.

4. Configure your test

   Select your dataset and add Evaluators to measure the quality of outputs.

   

   You can use create and use Presets for your test runs to save time and avoid repeating the same configuration.

5. Review results

   Monitor the test run to analyze the performance of your Prompt Chain across all inputs.