Create a Knowledge Base in Gen AI Builder

Suggest edits

Who is this for

Platform users configuring Knowledge Bases in Gen AI Builder in Hybrid Manager AI Factory. Typical users include developers, AI architects, and business owners managing Knowledge Bases for production AI applications.

What you will accomplish

You will create a Knowledge Base by selecting one or more Libraries as its content source.

You can configure:

  • Fully-managed vector stores
  • Self-managed Postgres vector stores
  • Hybrid Knowledge Bases combining structured and unstructured data

You will prepare your Knowledge Base for use in:

  • AI Assistants
  • RAG pipelines
  • Semantic search applications

Why use Knowledge Bases

Knowledge Bases provide the queryable foundation for Gen AI applications:

  • Power semantic search and Retrieval-Augmented Generation (RAG).
  • Enable LLM-based agents to retrieve content grounded in your data.
  • Combine structured metadata with rich unstructured text for advanced use cases.
  • Support refreshable pipelines — when source Libraries are updated, Knowledge Bases can be re-synced.

For background, see:

To see how Knowledge Bases are used in Hybrid Manager deployments, visit:

Complexity and time to complete

  • Complexity: Low to moderate (depends on Knowledge Base type selected)
  • Estimated time: 5–15 minutes

When to create a Knowledge Base

  • When preparing content for use in Assistants.
  • To enable RAG pipelines.
  • To deliver semantic search experiences.
  • To combine structured filters with AI-driven search (Hybrid KB).
  • After creating and validating Libraries with your source content.

How to create a Knowledge Base

1. Navigate to Knowledge Bases

  • In Gen AI Builder UI, go to Knowledge Bases.

2. Create a new Knowledge Base

  • Click Create Knowledge Base.

3. Configure common fields

Name (required)

  • Provide a clear, unique name.
  • Example: Company Policy KB

Description (optional)

  • Describe the Knowledge Base’s purpose or intended use.

4. Select Knowledge Base type

Gen AI Builder supports three Knowledge Base types:

PG.AI - Fully-Managed Vector Store

  • Create a vector-based Knowledge Base with a fully managed backend.
  • Optimized for unstructured text and general-purpose semantic search.

Configure:

  • Libraries: Select one or more Libraries.
  • Example: HR Policy Documents Library, Employee Handbook Library

Example

  • Name: Company Policy KB
  • Libraries selected: HR Policy Documents Library, Employee Handbook Library

PG.AI Database - Self-Managed

  • Connect to your own Postgres-compatible database with AI/vector extensions.
  • Useful when you want full control over your vector store (ex: HCP-managed Postgres + AIDB).

Configure:

  • Connection String: Full connection string to your database.
  • Example: postgresql://user:password@host:port/dbname
  • Password: Database password.
  • Libraries: Select one or more Libraries.

Example

  • Name: Tech Docs Self-Managed KB
  • Connection String: postgresql://user:password@host:5432/techdocs
  • Password: **********
  • Library selected: API Documentation Library

PG.AI Hybrid Database

  • Build a Hybrid Knowledge Base with both structured filters and semantic search.
  • Ideal for catalogs, customer profiles, and multi-attribute search experiences.

Configure:

  • Libraries: Select one or more Libraries (typically structured data sources such as CSV, Google Sheet).
  • Structured Columns:
  • Add columns used for filtering.
  • Define name, optional description, data type (Text, Number, Date, Boolean).
  • Unstructured Columns:
  • Add columns used for semantic search embeddings.
  • Define name and optional description.

Example

  • Name: Product Catalog Hybrid KB
  • Library selected: Product CSV Library
  • Structured Columns:
  • SKU (Text)
  • Price (Number)
  • Unstructured Columns:
  • ProductName
  • Features

5. Finalize creation

  • Click Create.
  • The system will build and populate your Knowledge Base from the selected Libraries.
  • Progress is visible in the Knowledge Bases list view.

Example scenario

You want to build a Hybrid Knowledge Base to power product search and filtering in your AI Assistant.

Example configuration:

  • Name: Product Catalog Hybrid KB
  • Library selected: Product CSV Library
  • Structured Columns:
  • SKU (Text)
  • Price (Number)
  • Unstructured Columns:
  • ProductName
  • Features

Troubleshooting

Missing Libraries

  • Verify that your source Libraries are created and populated.
  • Hybrid Knowledge Bases require Libraries with defined schemas.

Connection errors (Self-Managed)

  • Validate your database connection string and credentials.
  • Ensure required extensions (pgvector, AIDB) are installed.
  • Confirm that your database is reachable from Gen AI Builder.

Incomplete indexing (Hybrid KB)

  • Verify correct columns were selected for structured and unstructured indexing.
  • Column names must match the Library schema.

Related topics

← Prev

Create an Assistant in Gen AI Builder

↑ Up

Gen AI How-To Guides

Next →

Create a Retriever in Gen AI Builder

Could this page be better? Report a problem or suggest an addition!