• 11 Jun 2024
  • 2 Minutes to read
  • Contributors
  • Dark
  • PDF


  • Dark
  • PDF

Article summary

Feature restriction

The feature you are trying to access is only available to administrators with specific permissions. If you believe you should have access, please contact your account manager or the support team for assistance.

Capture d’écran 2023-10-16 à 12.01.04

In order to create a chatbot, you need to access the admin page and follow these steps:

Make sure that the company and users have been created correctly.

Create chat settings

A chat setting is automatically generated upon the creation of a company.

When a new company is created, a chat setting is automatically established. This ensures that all necessary configurations and preferences for chat functionalities are in place, facilitating seamless settings.

Chat settings:

The Chat Configuration Management functionality in the admin interface allows administrators to customize the behavior of the chat feature through various settings. These include:

  • Default Model Selection: Administrators can specify which AI model is the default choice in the chat interface for their company.
  • Behavior Customization: Define the prompt that influences the chat behavior for all users within the company.
  • Response Handling for Unmatched Queries: Configure how the chat should respond when a user asks a question that does not match any document in the database.
    • Option 1: The chat can still provide an answer, specifying that the response is not sourced from the provided documents.
    • Option 2: The chat can be set to provide no answer if no relevant document is found.
  • Customize the message displayed when the chat cannot find an answer in the document database and is configured not to respond in such cases.
  • Single Chat Settings Association: Each company can be associated with only one set of chat settings.
  • Document Association: View the documents linked to the chat functionality.
  • Bug Report Email Configuration: Define the email address to which bug reports will be sent when users encounter issues and submit bug reports.

Chat messages:

Displays the list of user inputs/queries submitted through the chat interface.

Chat responses flags:

Lists messages flagged by users when they are submitted a potential bugs or issues. This helps in identifying and addressing user-reported problems.

Chat responses:

Provides a detailed view of interactions between a user and the chat system, including technical information not visible from the front end. Details include:

  • Date/Time: The date and time when the response was generated.
  • Standalone Question: How the system rephrased the user’s question.
  • Text: The content of the response generated by the model
  • User query: The original question posed by the user.
  • Chunk Response Links: The document extracts used by Paradigm to generate the response.
    • distance: score of similarity between rephrased query and chunk
    • certainty: result of LLM filtering, the higher the better, chunks below a certain threshold won't be included in the response generation

Chat sessions:

Lists all chat sessions generated by users. Each session includes multiple chat responses and tracks the entire conversation flow.

Document extracts:

Details the excerpts from documents that are used when a user asks a question with the "search in my document" option selected. The model responds by providing relevant document excerpts to justify the generated answer.

Feedback categories:

When a user provides feedback on a generated response by clicking like or dislike, a pop-up invites them to specify their satisfaction level through checkboxes. Feedback categories allow customization of the items presented in the pop-up. These categories apply to all companies.

Response feedbacks:

enables viewing of user reactions to various responses, such as:

  • Regenerate: The user requested to regenerate the response.
  • Copy: The user copied the response.
  • Like/Dislike: The user liked or disliked the response.

Was this article helpful?