Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

description: >- Check whether input consists of any text from Deny list, and prevent being sent to LLM.

Simple Prompt Moderation

Simple Prompt Moderation Node

The Simple Prompt Moderation node provides customizable content filtering by checking input text against user-defined denied phrases or instructions, preventing potentially harmful or unwanted content from being processed.

Parameters

​

Inputs

  1. Deny List (Required)

  • Type: string

  • Description: List of denied phrases or instructions (one per line)

  • Example:

    ignore previous instructions
do not follow the directions
you must ignore all previous instructions
  1. Chat Model (Optional)

  • Type: BaseChatModel

  • Description: Language model to detect semantic similarities with denied phrases

  1. Error Message (Optional)

  • Type: string

  • Default: “Cannot Process! Input violates content moderation policies.”

  • Description: Custom error message to display when moderation fails

Functionality

The Simple Prompt Moderation node provides content filtering through the following features:

  1. Pattern Matching

  • Exact match detection against deny list

  • Case-insensitive comparison

  • Line-by-line analysis

  1. Semantic Analysis (when Chat Model is provided)

  • Similarity detection using LLM

  • Context-aware filtering

  • Flexible matching capabilities

  1. Customization Options

  • User-defined deny lists

  • Configurable error messages

  • Optional LLM integration

Use Cases

  1. Prompt Injection Prevention

  • Block attempts to override system instructions

  • Prevent prompt manipulation

  • Maintain system integrity

  1. Content Filtering

  • Filter specific keywords or phrases

  • Implement custom content policies

  • Control user input quality

  1. Safety Enforcement

  • Prevent harmful instructions

  • Block unwanted commands

  • Maintain usage boundaries

​

Integration Notes

  • Position the node early in your workflow to filter inputs

  • Consider combining with other moderation nodes for layered protection

  • Monitor and update deny lists regularly

  • Test thoroughly with various input patterns

​

Best Practices

  1. Deny List Management

  • Keep deny lists up to date

  • Use specific, clear patterns

  • Document denied phrases

  • Regular expression support for complex patterns

  1. Error Handling

  • Provide clear error messages

  • Log moderation events

  • Implement appropriate fallbacks

  1. Performance Optimization

  • Balance deny list size with performance

  • Consider caching for frequent patterns

  • Monitor LLM usage when enabled

  1. Maintenance

  • Regular deny list reviews

  • Update patterns based on new threats

  • Monitor false positive/negative rates

  • Adjust sensitivity as needed

{% hint style="info" %} This section is a work in progress. We appreciate any help you can provide in completing this section. Please check our Contribution Guide to get started. {% endhint %}