Introduction

Understanding how to effectively use a prompt builder can significantly enhance your interactions with machine learning models. In this article, we'll break down the key features of the Prompt Builder in Paradigm admin, designed to help users create, manage, and evaluate prompts.

Builder Workflows

When users start building a new prompt via the task builder, they follow a structured workflow that guides them through various stages. This workflow ensures that you don't miss any crucial steps in creating a well-structured prompt. All the stages of user workflow are recorded and available for review.

Prompt Templates

A Prompt Template is essentially a saved prompt that users can reuse and share within your company. This feature is especially useful if users have prompts that work well and want to use them repeatedly or share them with colleagues. The Prompt Templates section lets you view all saved and shared prompts, making it easy to manage your prompts.

Prompt Evaluations

Evaluating the effectiveness of prompts is crucial for continuous improvement. For each prompt usage, user can provide feedback by giving a thumbs up or down and selecting the success criteria that the result has met. These evaluations are recorded and accessible, allowing you to see how well your prompts are performing and identify areas for improvement.

Prompt Template Variables

Each Prompt Template comes with variables, which are context fields that you must provide for the model to generate a contextualized response. Think of these variables as the information the model needs to understand the context of your prompt better. This section lists all the context fields associated with your prompt templates, ensuring you provide all necessary information for accurate responses.

Success Criteria

Success criteria are values generated by the model that help users both describe the needs more effectively to improve the prompt and evaluate the quality of the response with objective measures. The Success Criteria section lists all the criteria associated with the prompt templates, helping user set clear goals and benchmarks for your prompts.

Conclusion

By understanding and utilizing these features of the Prompt Builder in Paradigm, you can create more effective and well-evaluated prompts, leading to better interactions with your machine learning models. Whether you're managing workflows, saving and sharing templates, evaluating prompt effectiveness, handling variables, or setting success criteria, these tools provide a comprehensive system to enhance your prompt-building process.