Concepts v1.3

EDB Postgres® AI (EDB PG AI) brings AI workloads into your Postgres platform and infrastructure — with an architecture designed to power Sovereign AI on hybrid and cloud-native environments.

AI Factory extends Postgres to support:

• Vector search and semantic search
• Data pipelines for embedding generation
• Gen AI application development
• Scalable model serving and inference
• AI-driven governance and observability

AI Factory helps you build and operate trusted, enterprise-grade AI systems using your data, models, and applications.

Before you start

Prerequisites

You’ll get the most out of this section if you have:

• Familiarity with Postgres and database concepts
• Basic understanding of Large Language Models (LLMs) and vector search
• Awareness of Retrieval-Augmented Generation (RAG) techniques
• Experience with AI concepts such as embedding models, pipelines, and inference services

If new to AI Factory, consider starting with:

• AI Factory Terminology
• AI Factory 101 Path

EDB’s vision for AI with Postgres and Hybrid Manager

EDB’s AI Factory strategy makes Postgres a first-class platform for AI workloads and AI-driven applications.

The architecture combines:

• AI in Postgres: vector search, embedding pipelines, and AI functions within the database
• Gen AI apps: conversational and task-driven AI applications with Gen AI Builder
• Scalable inference: GPU-accelerated model serving using Kubernetes and KServe
• Unified control plane: model lifecycle management and observability in Hybrid Manager

This enables Sovereign AI — AI on your data, on your infrastructure, with full governance.

Core AI patterns

Vector search and semantic search

AI Factory makes it easy to implement semantic search, product recommendations, anomaly detection, and similar patterns using vector search.

You can leverage:

• The Vector Engine, providing pgvector-based capabilities
• AI Factory Pipelines to generate embeddings
• Knowledge Bases for integrated vector search across structured and unstructured data

Retrieval-Augmented Generation (RAG)

RAG pipelines enhance LLM responses by combining:

• Domain-specific embeddings from Knowledge Bases
• Configurable search behavior with Retrievers
• Scalable model inference through Model Serving
• Application orchestration via Gen AI Builder

This enables more accurate, grounded, and compliant AI applications.

AI in Database (In-DB AI and Vector Search)

AI Factory transforms Postgres into an AI data platform with:

• Vector similarity search using the Vector Engine
• Embedding pipelines integrated via Pipelines
• Search and retrieval across Knowledge Bases

Future capabilities will extend in-database AI to include model scoring and advanced AI operations.

Model serving and inference

AI Factory provides robust model serving with:

• Scalable inference services for LLMs, embedding models, and vision models
• GPU-accelerated serving via Kubernetes + KServe
• Flexible deployment options through the Model Serving framework
• Lifecycle management and discoverability with the Model Library Explained

These services power both Gen AI Builder agents and Knowledge Bases.

Gen AI application development

Gen AI Builder enables rapid development of intelligent applications that:

• Orchestrate LLMs with Assistants
• Retrieve content using Knowledge Bases and Retrievers
• Enforce behavior via Rulesets
• Execute actions via Tools
• Maintain conversational context through memory settings in Assistants

Explore Gen AI Builder to start building AI-driven apps.

Pipelines and data preparation

AI Factory Pipelines automate:

• Embedding generation
• Chunking, parsing, summarizing content
• Keeping embeddings fresh with Auto-Processing

Explore Preparers for reusable data preparation components.

Architectural principles

Modular architecture

AI Factory is built for:

• Hybrid environments (on-premises, cloud, multi-cloud)
• Seamless integration with Postgres and cloud-native services
• Composable building blocks for AI applications and workflows

Open standards and interoperability

AI Factory components use:

• Open model formats via KServe
• pgvector for vector search in Postgres
• Griptape for framework-agnostic LLM orchestration
• RAG-friendly patterns with flexible data and model integration

Summary

EDB PG AI enables you to:

• Embed AI capabilities directly into Postgres
• Build scalable, production-ready AI applications
• Operate Sovereign AI on your infrastructure with unified governance
• Integrate open-source tools and standards across the AI stack

Next steps:

• AI Factory Terminology
• AI Factory Generic Concepts
• AI Factory Learning Center