LearnedKnowledgeFeed

Overview

The LearnedKnowledgeFeed component in the Vantage platform is designed to present auto-learned knowledge entries derived from user interactions and conversations. This component displays these entries along with functionalities to approve, reject, edit, and manage the knowledge base snippets. It enhances the user's ability to monitor AI-learned insights efficiently and ensures that the knowledge base remains relevant and accurate by allowing for manual oversight of entries generated by the AI.

Settings

The LearnedKnowledgeFeed component has one primary configurable setting:

1. canManageOrgSnippets

• Input Type: Boolean
• Description: This setting governs whether a user has permission to manage organization snippets. When set to true, users are granted the ability to approve, reject, and edit snippets. Conversely, if the value is false, the component will restrict users from making changes to the snippets, effectively limiting their interaction to viewing only.
• Default Value: false

How It Works

Upon rendering, the LearnedKnowledgeFeed component performs the following core functions:

1. Data Fetching:

   • It makes an API call to fetch the knowledge entries, filtering them based on their current status (pending, approved, rejected) and collecting up to 50 results at a time.

2. Entry Management:

   • Users can approve, reject, or edit knowledge entries:
     • Approve an Entry: Updates the entry's status to 'approved'.
     • Reject an Entry: Updates the entry's status to 'rejected'.
     • Edit an Entry: Allows users to modify the text of a knowledge entry and automatically sets the status to 'approved' upon saving changes.

3. Display and Interaction:

   • The component presents a UI with filter tabs for entries based on their status and displays counts of pending, approved, and rejected items.
   • It features loading indicators during data fetching and updates, ensuring users are always aware of the component's state.

4. Aggregated Actions:

   • A button is provided to approve all pending entries simultaneously, which is useful for bulk management of knowledge snippets.

Data Expectations

The LearnedKnowledgeFeed component expects the following data structure from the API:

• Entries: An array of objects representing the learned knowledge instances. Each entry should contain:

  • id: Unique identifier for the entry.
  • text: The content of the knowledge snippet.
  • status: Indicates whether the entry is 'pending', 'approved', or 'rejected'.
  • snippetId: Identifier of the knowledge snippet from which the entry was generated.
  • snippetCategory: Category of the knowledge snippet, affecting how it is displayed (e.g., terminology, rules).
  • source: The origin of the snippet, providing context (e.g., AI chat, manual).
  • confidence: Indicates the confidence level of the learned entry (high, medium, low).

• Counts: An object containing:

  • pending: Number of entries currently awaiting approval.
  • approved: Number of entries that have been approved.
  • rejected: Number of entries that have been reviewed and rejected.
  • total: Total number of entries retrieved.

AI Integrations

The LearnedKnowledgeFeed leverages AI capabilities by automatically learning and generating knowledge entries from user interactions. The AI continuously refines its understanding based on feedback from approved, rejected, and edited snippets, creating a dynamic learning environment that improves over time.

Billing Impacts

While the use of the LearnedKnowledgeFeed itself does not directly affect billing, it is essential to consider the following aspects in relation to potential costs:

1. API Usage: Frequent API calls to fetch snippets and update statuses could result in increased usage of API credits if the platform implements a charge based on API call volume.

2. Data Storage: The number and size of stored user interactions and learned snippets may influence storage costs, especially in high-traffic environments.

3. User Licenses: Depending on the level of access and functionality available, the number of users who can manage snippets may impact licensing fees.

Use Cases & Examples

Use Case 1: Knowledge Management in Customer Support

Scenario: A customer support team utilizes the LearnedKnowledgeFeed to manage AI-generated insights from customer interactions. Customer inquiries are logged, and the AI learns common responses and solutions.

Use Case 2: Continuous Improvement in Product Training

Scenario: In a product training environment, educators use the LearnedKnowledgeFeed to review AI-generated notes and responses provided to trainees, ensuring that the context and accuracy of information are maintained.

Use Case 3: Content Moderation for AI-Generated Suggestions

Scenario: Content managers review AI-generated suggestions that are displayed via the LearnedKnowledgeFeed. They can approve well-structured content and reject entries that do not align with brand standards.

Example Configuration

To configure the LearnedKnowledgeFeed for a customer support implementation:

<LearnedKnowledgeFeed canManageOrgSnippets={true} />

In this configuration, the customer support team members can approve, reject, and edit AI-learned responses, enabling them to refine and maintain a high-quality knowledge base as they interact with users. This setup allows for real-time insights into customer needs while ensuring that the AI continues to learn and adapt based on feedback.