Number of "best" hits that will be kept for navigation and full hits. It limits the number of hits that can be used for navigation and full hits fetch. Reducing this parameter can slightly increase performance in some situations (many slices, IO-bound). The "nhits" value in answer is at most equal to this limit. 0 means no limit. Note: Heapsort executor is used if this setting is defined in ranked mode. If not in ranked mode, this setting is ignored.

Maximum number of hits that a single index slice can fetch.

Unused setting

Each time the slice heap receives this number of hits, it will be flushed to the main heap. This parameter only impacts performance. If value is 0, the system will pick up an appropriate value. Note: Only used in unranked and heapsort. Not used in localsort.

Maximum time spent in the "hits matching" part of the search internal "query" part of the search. After this delay, hits matching is interrupted, partial results will be returned. 0 means no limit.

When this total number of hits matching the query have been found (on any slice), the query is interrupted. These hits are ranked and the "best" ones are selected for navigation and full hits. Reducing this parameter improves performance but it means that a potentially "better" hit located further in the index will not be found. Above this limit, the "nmatches" value in answer will be estimated. 0 means no limit.

The query is interrupted if at least this percentage of active slices finished fetching hits.

After the query is interrupted, the slices have a small time period to flush their heaps. Hits that arrive to the main heap after this time period are dropped.

Maximum time for the entire search (hits matching, navigation and hits computation). After this delay, the search query is aborted. 0 means no limit.

Additional time after maxTotalTime before forcing the full hits processor to stop. Note that this elapsed time is only checked between two full hits.

Only fetch a sample of the matches. This is useful to improve search performance while still getting a good hits distribution.

Set this optio to True when you plan to use consecutive queries, then the dedup will occur.

If this value is not zero, when doing expand queries without using the "expand=true" mode, the slices will only consider the first maxExploredPaths paths they find for deciding which DIDs are contained in the graph.

If this value is not zero, when doing attrgraph queries, each document that need to be postfiltered will have its Fullhit rewritten once said number of documents to postfilter has been reached.

Optional specific per content group limits.