Skip to content

Prompt Management

Saved prompts represent a sophisticated approach to standardizing AI interactions across your organization. Rather than repeatedly crafting complex instructions for recurring tasks, saved prompts establish reliable workflows that ensure consistent quality and efficiency.

Why Save Instructions?

Saved prompts offer multiple benefits:

  • Quality & Consistency: Ensure uniform formatting and style across all outputs
  • Efficiency & Time Savings: Save significant time by not reconstructing complex instructions repeatedly
  • Completeness & Safety: Include all necessary parameters and requirements, reducing errors

Creating a Simple Instruction

Let's start with a simple example: you frequently want to summarize texts in three bullet points.

  1. Click Workspace in the left sidebar, then click "Prompts"
  2. Click the plus icon in the top right to create a new prompt
  3. Fill in the key fields:
  4. Title: A name you'll recognize, like "Summary in 3 Points"
  5. Command: A short command to invoke the instruction in chat, like /summary
  6. Prompt Content: The actual instruction to the AI
    • Example: Summarize the following text in three clear and concise bullet points. Focus on the core messages.

When you type /summary in a chat, your saved instruction will be automatically inserted.

Next Level: Interactive Workflows with Variables

This is where things get really powerful. You can make your instructions interactive with variables (placeholders). When you use such a prompt, a form opens asking for the needed information, which is then automatically inserted into your instruction.

Example: A workflow for summarizing articles

  1. Create a new prompt with the command /article
  2. Enter the following instruction:
Please summarize the following article in three key messages.

Article:
{{article_text|textarea:placeholder="Insert the complete article text here..."}}

In Open WebUI, you can use variables with the syntax {{variable_name|input_type}} to create interactive forms. For example, {{article_text|textarea}} would create a text area field for users to input article content.

When you type /article in chat, here's what happens: 1. A pop-up window appears asking you to insert the article text 2. You copy the article into this field and click "Save" 3. The system builds a complete instruction from this and sends it to the AI

Your Advantage: Reproducible and Error-Free Processes

Using variables creates genuine workflows for the future. These are not just faster but also reproducible and robust, as no important details or instructions can be forgotten.

Advanced Tips

Prompt Categories Your Organization Might Need

Content Creation Prompts: - Document summaries - Report generation - Email composition - Creative writing frameworks

Analysis and Review Prompts: - Code reviews - Document quality assessments - Data analysis procedures - Risk evaluation frameworks

Administrative Prompts: - Meeting minutes creation - Action item tracking - Compliance verification - Process documentation

Best Practices

  • Be specific about expected output formats
  • Include helpful examples
  • Define clear scope boundaries
  • Test thoroughly across different scenarios
  • Use intuitive variable naming
  • Keep prompts accessible and practical

Troubleshooting Common Issues

  • Poor output quality: Needs clearer instructions or better examples
  • User confusion: Improved guidance or interface design needed
  • Optimization: Use A/B testing, gather feedback, and refine iteratively

Team Collaboration

Enhance prompt value through: - Standardized sharing across teams for consistency - Collaborative development to capture diverse expertise - Regular review sessions and shared best practices libraries

The formalization of instructions from informal commands into saved, standardized workflows represents a fundamental shift in how organizations interact with AI systems. This transformation enables consistent, predictable outcomes while reducing errors and improving quality across all AI-supported processes.