1)

add_field

is not a valid payload. You want to follow the payload structure mentioned in this code block exactly: https://typesense.org/docs/27.1/api/collections.html#modifying-an-existing-field 2) The 4K limit only applies to query parameters for GET requests. If you use the

multi_search

endpoint (even for a single search), that uses POST and the filter goes in the POST body, and that does not have the 4K limit. It sounds like the OS might be killing the Typesense process due to memory constraints. You want to increase the amount of RAM and observe RAM usage as you send these requests. 3) Filtering and searching require some transient amounts of memory to process as they execute, but it's hard to exactly come up with a formula for that because the exact usage depends many factors like search parameters, shape of data. Same goes for CPU usage as well. So the best way to know how much RAM / CPU you need for search would be to run a benchmark with search queries that are representative of your production search patterns and size your nodes based on that. On a side note, we recently fixed an issue around filtering taking up too much RAM. This fix is in v27.1, so I would recommend using that version.