Every AI app should include evals! With PromptFlow, we ensure not only the functional correctness of our code but also the precision and reliability of AI inferences, preventing issues like hallucinations. For more information about PromptFlow, visit the PromptFlow documentation and the GitHub repository.

To set up Azure Pipelines for this project:

1. Create a Variable Group:

   • The pipeline depends on a specific variable group named promptflow-evals.
   • Ensure this group includes all necessary variables, such as DEPLOYMENT_NAME, AZURE_OPENAI_ENDPOINT, and AZURE_OPENAI_API_KEY.

2. Configure the Pipeline:

   • Modify the azure-pipelines.yaml file according to your project's needs.
   • This file outlines the build and test processes, integral to integrating PromptFlow evaluations into your CI/CD workflow.

3. Run the Pipeline:

   • Trigger the pipeline through Azure Pipelines.
   • Upon completion, the results of the PromptFlow evaluations will be accessible as an artifact.
   • Review this artifact for a detailed report of the evaluations.

VS Code, complemented by the PromptFlow extension, offers an intuitive environment for working with PromptFlow:

1. Install Required Tools:

   • Download and install the latest stable version of VS Code.
   • Add the Python extension for VS Code.
   • Install the PromptFlow extension for VS Code.

2. Set Up Your Environment:

   • Choose the appropriate Python interpreter in VS Code.
   • Use the visual editor to open flow.dag.yaml for an enhanced flow editing and testing experience.