@@ -23,33 +23,33 @@ Other versions:
23
23
[[integrate_filtering_support_for_approximate_nearest_neighbor_search]]
24
24
=== Integrate filtering support for approximate nearest neighbor search
25
25
26
- The {ref}/knn-search-api.html[_knn_search endpoint] now has a "filter" option
27
- that allows to return only the nearest documents that satisfy the provided
26
+ The {ref}/knn-search-api.html[_knn_search endpoint] now has a "filter" option
27
+ that allows to return only the nearest documents that satisfy the provided
28
28
filter.
29
29
30
30
[discrete]
31
31
[[nlp-latency-throughput-stats]]
32
32
=== NLP latency and throughput stats
33
33
34
- New statistics are available for the NLP inference to show how quickly it is
34
+ New statistics are available for the NLP inference to show how quickly it is
35
35
working. The three new statistics are:
36
36
37
37
* peak_throughput_per_minute
38
38
* throughput_last_minute
39
39
* average_inference_time_ms_last_minute
40
40
41
- The aim is to provide an indication of whether inference is currently keeping up
42
- with requirements or the cluster needs to be scaled up to meet demand. The
43
- statistics for the last minute give quick feedback to show the effect of scaling
41
+ The aim is to provide an indication of whether inference is currently keeping up
42
+ with requirements or the cluster needs to be scaled up to meet demand. The
43
+ statistics for the last minute give quick feedback to show the effect of scaling
44
44
the {ml} nodes.
45
45
46
46
[discrete]
47
47
[[random-sampler-aggregation]]
48
48
=== Random sampler aggregation
49
49
50
- With the new random sampler aggregation, in technical preview, developers can
51
- exponentially accelerate their aggregations for calculations, with a slight
52
- trade off in accuracy, by randomly sampling documents in a statistically robust
53
- manner. The random sampler aggregation can be used to accelerate any application
50
+ With the new random sampler aggregation, in technical preview, developers can
51
+ exponentially accelerate their aggregations for calculations, with a slight
52
+ trade off in accuracy, by randomly sampling documents in a statistically robust
53
+ manner. The random sampler aggregation can be used to accelerate any application
54
54
that utilizes aggregations for calculations.
55
- // end::notable-highlights[]
55
+ // end::notable-highlights[]
0 commit comments