Concepts

Knowledge Library Articles

Knowledge Library Articles are designed to streamline the management and retrieval of large text documents within your organization.
By categorizing these articles into specific "rooms" and "shelves," you can efficiently organize and access the information when building AI models using retrieval augmented generation RAG

Basic Principle

Organizational Structure

Our system uses an intuitive structure to help you categorize and locate articles.

The structure consists of:

Library: The overarching container for all your knowledge articles. This is automatically set to your organization.
Rooms: Broad categories representing different departments or areas within your organization.
Shelves: Subcategories within each room, further organizing the documents.
Articles: The actual documents stored within each shelf.

As you can see, the structure is hierarchical, with each level providing more specific information about the location of the article.

It is similar to how you would think about retrieving a book from a library, starting with the library, then moving to a specific room, shelf, and finally the book itself.

For example, consider the following organizational structure:

• Client Services Room:

  • Client Contracts
    • Active Clients
    • Past Clients
  • Support Documents
    • User Guides
    • Troubleshooting Manuals
    • FAQs
  • Service Reports
    • Monthly Reports
    • Annual Reviews
    • Client Feedback

• Administrative Room:

  • Human Resources
    • Employee Records
    • Payroll Documents
    • Benefits and Compensation
  • Finance
    • Accounts Payable
    • Accounts Receivable
    • Financial Statements
  • Tax Documents
    • Legal
    • Contracts
    • Compliance Documents
    • Legal Correspondence

When setting up your articles, you simply specify the room and shelf where the article belongs.
This categorization aids in more precise retrieval of documents, enabling AI models to access relevant information quickly.

Integration with Stubber Flows and Assistants

More documentation on how to integrate knowledge libraries into flows and models will be added soon.

Tasks

• Search Knowledge Library Task
  • Can be used to search your libraries, and the results can then be fed into the GPT Chat Task to generate responses.
• Add Knowledge Library Article Task
  • Can be used to add articles to your knowledge library
  • Coming soon

Managing Articles

See Managing Articles for more information on creating and managing articles in your Knowledge Library.

Document Embedding and RAG Overview

Document Embedding

Document embedding transforms textual data into numerical vectors that capture the semantic meaning of the text. This allows AI models to retrieve relevant information efficiently.
In your Knowledge Library, embeddings are generated for each article (once synced), enabling precise and context-aware searches rather than basic keyword matching.

Retrieval-Augmented Generation (RAG)

RAG leverages document embeddings to enhance the AI's ability to generate responses to queries. When a query is made, the system retrieves relevant articles using embeddings and then generates a response based on both the query and the retrieved content. This ensures responses are accurate and contextually relevant to your organization's specific knowledge.

Benefits

• Accurate Responses: Embeddings and RAG improve the relevance and precision of AI-generated responses.
• Enhanced AI Output: RAG allows the AI to provide more informed and context-specific answers.
• Cost reduction: RAG retrieves relevant data only when needed, avoiding unnecessary context window overload, reducing overall cost.