Human feedback
The human feedback section allows the Admins and AI engineers roles to review the feedback provided by the users when interacting with the AI Chat agents if the Allow feedback option is enabled in the Agent settings.
This section collects all feedback coming from the various chats and displays them in a single place.
Main view
The section shows all feedback provided by the users when interacting with the AI agents.
Filters
Service: The type of feedback that was provided to the AI agents. Below the full list of options:
- Multi language
- Documentation
- Reasoning
- Images extraction
- Charts generation
- Live data
Additionally ToothFairyAI by default does not show the feedback that has been resolved. The Show resolved checkbox can be ticked to display the resolved feedback as well.
Feedback details
Below the various feedback fields that are displayed in this section:
ID: The unique identifier for the feedback.
Username: The user who provided the feedback.
Feedback: The type feedback provided by the user (Positive or Negative).
Type: The main type of feedback provided by the user.
Sub Type: The sub type of feedback provided by the user.
Comment: Additional info provided by the user to better explain the feedback.
Resolution: Comments field accessible only by the Admins and AI engineers roles to provide a resolution to the feedback provided by the user.
Mark as resolved: Allows the Admins and AI engineers roles to mark the feedback as resolved for better tracking.
Autotune from resolution: When turned on, the feedback resolution will be used to fine-tune the AI model automatically adjusting the conversation based on the feedback provided by the user. The user must include in the resolution at least 10 characters to provide enough context for the AI model to generate the training dataset.
Actions
Go to chat: Hyperlink to the chat where the feedback was provided.
Create a training document: Allows the Admins and AI engineers roles to create a training document based on the feedback provided by the user.
Only Enterprise workspaces can leverage the fine-tuning feature from human feedback. While Starter and Pro can only create the training documents and datasets