Knowledge Bases explained

Knowledge Bases are the semantic core of AI Factory — transforming your organizational content into a structured, queryable, AI-friendly layer.

They enable AI applications to retrieve trusted, relevant information at query time:

• Power semantic search and Retrieval-Augmented Generation (RAG)
• Ground AI Agents and Assistants in your organization’s knowledge
• Support hybrid structured + unstructured queries for advanced AI workflows

Without Knowledge Bases, Gen AI Builder agents and Assistants would lack the ability to retrieve trusted, explainable knowledge.

Before you start

Prerequisites for understanding Knowledge Bases:

• Basic understanding of Embeddings and Vector Search
• Familiarity with AI Factory Data Sources and Libraries
• Awareness of Retrievers, Assistants, and RAG

Suggested path:

• AI Factory Concepts
• AI Factory 101 Path

What is a Knowledge Base?

A Knowledge Base is an indexed store of content optimized for:

• Semantic search — find relevant content by meaning, not keywords
• RAG — enrich LLM outputs with trusted, up-to-date knowledge
• Hybrid queries — combine metadata filtering with semantic search
• Explainability — trace responses back to original content sources

It acts as an optimized layer between your raw content and your AI Agents / Assistants.

Pipeline flow:

Data Sources → Libraries → Knowledge Bases → Retrievers → Assistants → AI Applications

Components:

• Data Sources — Raw content (documents, web pages, APIs)
• Libraries — Processed and structured content collections
• Knowledge Bases — Indexed, query-optimized semantic layer
• AI Applications — Agents, Assistants, APIs, user-facing apps

Why use Knowledge Bases?

Knowledge Bases enable you to:

• Ground AI models in your organization's knowledge
• Provide explainable AI with traceable, source-linked results
• Support RAG pipelines for domain-specific and compliant LLM applications
• Perform semantic search across trusted content
• Support hybrid filtering (structured metadata + semantic vectors)
• Enable multi-source knowledge orchestration (combine multiple Libraries)

How Knowledge Bases work

1. Creation

You create a Knowledge Base and select one or more Libraries as its source:

• Ingests content from Libraries
• Creates vector embeddings for unstructured content
• Indexes structured metadata columns for hybrid queries

2. Querying

Knowledge Bases support:

• Vector search — semantic similarity matching
• Hybrid search — combine vector + structured filters
• RAG — retrieved content injected into LLM prompts at runtime
• Explainability — results linked back to source documents

3. Refresh & maintenance

When source Libraries change:

• Knowledge Bases can be re-synced to reflect latest content
• You can manage versioning and refresh cycles for consistency and governance

Types of Knowledge Bases

AI Factory supports multiple Knowledge Base deployment options:

• Fully managed vector store Cloud-based, optimized for semantic search and scalability

• Self-managed Postgres with vector search Leverage Postgres + pgvector extension for in-database semantic search

• Hybrid Knowledge Base Combine structured metadata with unstructured embeddings for advanced hybrid search patterns

Patterns of use

Retrieval-Augmented Generation (RAG)

• Power AI Agents and Assistants with trusted knowledge
• Inject retrieved context into LLM prompts
• Improve accuracy and reduce hallucination

Knowledge search portals

• Build semantic search experiences on top of Knowledge Bases
• Combine with metadata filtering for enterprise knowledge discovery

Explainable AI workflows

• Ensure responses can be traced back to source content
• Support regulatory, compliance, and audit requirements

Multi-Knowledge Base orchestration

• Use multiple Knowledge Bases with weighted relevance
• Support multiple domains or languages
• Enable personalized or role-specific knowledge pipelines

Best practices

• Carefully design Hybrid Knowledge Bases to align structured + unstructured content
• Organize Libraries thoughtfully — they define your Knowledge Base quality
• Maintain clear governance — version Libraries and Knowledge Bases
• Monitor query performance — tune retrievers and embeddings as needed
• Combine with Rulesets and Memory in Assistants to ensure coherent behavior

Role in Hybrid and Sovereign AI

Knowledge Bases are central to AI Factory’s Sovereign AI vision:

• Content stays in your infrastructure
• AI applications use your knowledge and pipelines
• Knowledge Bases can run on:
• Fully managed vector store
• Self-managed Postgres
• Hybrid Manager (KServe + Postgres + object storage)
• Full observability and control through Hybrid Manager dashboards
• Knowledge governance and versioning are first-class features

Lifecycle of a Knowledge Base

Related topics

• Create a Knowledge Base
• Manage Knowledge Bases
• Best practices for Hybrid Knowledge Bases
• AI Factory Concepts
• Retrieval-Augmented Generation (RAG)
• Embeddings explained
• Structures explained
• Hybrid Manager: Using Gen AI Builder

Next steps

• Explore Retrievers Explained
• Learn about Assistants Explained
• Build a Knowledge Base with the AI Factory 101 Path