Prompt Template Editor Guide

Introduction

The Prompt Template Editor is a powerful tool that allows you to control how different pieces of information are organized and formatted when sending prompts to language models. This guide will help you understand the editor’s features and provide best practices for creating effective prompts.

Understanding Composite Prompts

In most specialized or domain-specific generative AI applications, effective prompting involves much more than simply sending the user’s latest query to the LLM. A complete prompt typically combines multiple elements:

• System instructions that define the LLM’s role, behavior and style of response
• The prior conversation history to maintain context across interactions
• Retrieved knowledge from databases or documents to be used in answering the user’s query
• User-provided rich content like uploaded files or images
• The current user query
• Optional formatting and structural elements

We refer to this comprehensive assembly as a “composite prompt”—a carefully structured combination of all elements needed for the LLM to perform its task effectively. The quality and organization of this composite prompt directly impacts the quality of the AI’s response.

The Prompt Template Editor gives you precise control over how this composite prompt is assembled, allowing you to:

1. Determine which elements to include
2. Define their order and formatting
3. Control token allocation across different sections
4. Manage conversation history and context retention

What is a Prompt Template?

A prompt template is a structured definition that controls:

• How conversation history is managed
• Which pieces of information are included in the prompt
• How to format and organize that information
• How to allocate token budgets to different sections

Unlike simple text templates, HyperFlow’s prompt templates give you precise control over the structure and composition of prompts, allowing you to build more effective LLM interactions. The template serves as a blueprint for how the system will gather and structure information into a composite prompt before sending it to the LLM.