Using models in Gen AI Builder Innovation Release

This documentation covers the current Innovation Release of EDB Postgres AI. You may also want the docs for the current LTS version.

Hub quick links: Gen AI how-tos — Model Serving — Model Library

How models are consumed

In Hybrid Manager, Gen AI Builder lets you build assistants that use model endpoints served from the Model Library.

When creating or editing an Assistant, you select which model to use (see Create an Assistant).
Models available here are governed by Hybrid Manager: pulled from the Model Library, deployed to your project, and exposed via internal endpoints.
This ensures that model calls stay within your environment — no external API calls are made by default.

Knowledge Bases in Gen AI Builder can be populated with embeddings generated by Pipelines:

Pipelines ingest and prepare documents into vector indexes (see Vector Engine concepts).
Knowledge Bases reference these pipelines, making them queryable for Retrieval-Augmented Generation (RAG).
Assistants then combine model responses with Knowledge Base retrievals to ground outputs in your organization’s data.

When Gen AI Builder runs inside Hybrid Manager:

Service discovery and endpoints are automatically managed by the platform.
You do not need to configure external URLs; models appear directly in the Assistant creation UI.
Any required environment variables (for service routing or authentication) are injected by HM.

This means you focus on building assistants — Hybrid Manager takes care of wiring models, data, and observability together.

Use these how-tos from the hub to start building:

In Hybrid Manager, models and data are co-located:

AI Factory in Hybrid Manager