Hi Team! I’m evaluating whether to use flat documents vs. reference (joined) documents in Typesense for an

asset

collection that references multiple other collections like

location

,

group

,

manufacturer

, and

label

. I understand that

include_fields

with

reference

is useful for enriching search results, but I'm trying to estimate the search performance impact when referencing multiple documents. If a flat asset document takes X ms to search, how much additional latency is typically introduced per reference join? For example, if I reference 3–4 collections, what kind of search latency increase (roughly) should I expect compared to a single flat collection? Additionally, is there a recommended upper limit on the number of referenced documents per collection before it starts to noticeably degrade search performance? Thanks in advance for your insights!

References will require additional disk lookups to join fields from referred collections, whose link documents have to be fetched and parsed from disk. The exact latency has to be measured with actual data: it's tough to give an estimate because it depends on size of documents.