Configurable Database Chunking for various types of Dataset Processing - T3Planet Docs

Overview

When AI Search reads a lot of content, it works in small portions (chunks) instead of all at once. This keeps your server from running out of memory or stopping halfway. The default is 1000 records per portion. Most websites never need to change it.

Configuration Steps

Go to AI Universe → AI Foundation → AI Features.
Open the Training card.
Change Chunk Size if needed (see the table below).
Optional: change Batch Size and Retention Days.
Click Save Changes.

Training drawer in AI Foundation with Chunk Size, Batch Size and Retention Days

Recommendations

Your website	Chunk Size
Small (fewer than 10,000 records)	1000 (default) or 500–800
Medium (10,000–50,000 records)	1000–1500
Large (more than 50,000 records)	1500–2000 – ask your hosting to watch memory use

Training stops with a memory error? Make the Chunk Size smaller. Training is very slow and your server has plenty of memory? Make it bigger.