We are prototyping the type sense for global search for the internal product which was using SQL server. We have search across 4 different domain entities like device, user application and content . We have a single text box when entered we are searching across all the tables to get the details . Now we are planning to redesign it using typesense. What is the best way to design the document structure. Is de normalise better or have multiple documents and do multi search

If you need to visually show your users that these results are from Entity X and these other results are from another Entity Y, then you want to put them in separate collections and do a multi-search.

Will that having a scaling issue in terms of response time?

Sending multiple searches in a single search multi_search request has less overhead, and fanning out searches across multiple collections is actually more performant