Dluxchat
  • Getting Started
    • Introduction to Dluxchat
    • Creating an Agent
    • Assigning an Identity to Your Agent
    • Educating your Agent through the Knowledge Hub
    • Embedding your Agent to a Website
  • Training Concepts
    • What is the Knowledge Hub
    • What is Real-time training
    • What is One-time training
    • What is Interactive training
    • What is Discover
  • Training How-to
    • Train your Agent with Websites
    • Train your Agent with a Knowledge Base
    • Train your Agent with Google Sheets
    • Train your Agent with Google Docs
    • Train your Agent with PDFs
    • Train your Agent with Youtube Videos
  • Agent
    • Giving an Identity
    • Setting Language Preferences
    • Setting up your AI API Keys
    • Setting Rate Limits
    • Human Takeover
    • Quality Control Systems
    • Finding your Agent ID
    • Collecting Data
  • Chatbot (Web Widget)
    • Chatbot Manager
    • Connecting an Agent
    • Greeting Editor
    • Using Instant Embed
    • GDPR Settings
  • Quality Audit
    • Tabular Quality Audit
    • Sitemap Quality Audit
  • Other
    • Knowledge Hub V2 Update
    • Widget V2 Update
Powered by GitBook
On this page
  • Prevent Hallucinations
  • Prevent Unrelated Conversations
  • Prevent Offensive Messages
  • Side effects
  1. Agent

Quality Control Systems

👋 Hello there! At Dluxchat, we place a high priority on the quality of conversations your website visitors have with our AI chatbot.

PreviousHuman TakeoverNextFinding your Agent ID

Last updated 1 year ago

To ensure this, we've carefully crafted three quality control systems dedicated to enhancing the user experience.

To access it, just click on this .

Below are the three systems that help enhance the quality of the conversation:

Prevent Hallucinations

The initial system addresses instances of low confidence. Essentially, the AI evaluates the user's query and compares it with the content in your Knowledge Hub.

By performing this comparison, the bot calculates a confidence score, indicating how effectively it can respond to the query.

In situations where the confidence score is low, rather than potentially providing an incorrect answer, the bot indicates uncertainty about the query. To prevent leaving the user without assistance, it usually redirects them to your contact form.

Prevent Unrelated Conversations

The second system holds a crucial role in upholding conversation relevance. The AI evaluates whether the query aligns with the information available in your Knowledge Hub.

If the user's query is unrelated to your business or the content in the Knowledge Hub, the chatbot politely informs the user that it cannot assist with the specific query.

Prevent Offensive Messages

The third and last system we employ ensures that the conversation remains professional and respectful.

By leveraging OpenAI's Content Moderation API, the chatbot examines the user's query.

If the query appears to contain rude or offensive content, or violates OpenAI's Terms of Service, the chatbot promptly responds with "I can't help with that" to uphold decency in communication.

Side effects

A couple of considerations regarding these Quality Control Systems: activating them doubles the token usage per conversation.

Additionally, due to the extra evaluations these systems conduct, responses may take an additional second.


I hope this provides you with a clear understanding of our Quality Control Systems here at Dluxchat. If you have any further questions, feel free to reach out at any time!

link