ES|QL Add initial support for semantic_text field type #113920

ioanatia · 2024-10-02T07:48:21Z

Support is added behind a feature flag. We could not simply use EsqlCapabilities since that's not available in esql-core.
Right now we have no support in existing functions.
I followed the approach for adding initial support for date_nanos which was also added behind a feature flag which allowed for incremental progress, rather than adding support for everything in one PR: #110205

With this PR we will return semantic_text fields as part of the results and it will also allow us to refer to semantic_text fields in the match function (to run semantic search):

{
  "columns": [
    {
      "name": "semantic_text_field",
      "type": "semantic_text"
    }
  ],
  "values": [
    [
      "these are not the droids you're looking for. He's free to go around"
    ],
    [
      "all we have to decide is what to do with the time that is given to us"
    ],
    [
      "I have no idea what I'm doing, but I know I'm doing it really, really well."
    ]
  ]
}

realtes: #115103

elasticsearchmachine · 2024-10-02T07:49:52Z

Hi @ioanatia, I've created a changelog YAML for you.

elasticsearchmachine · 2024-10-07T17:02:44Z

Pinging @elastic/es-search-relevance (Team:Search Relevance)

ioanatia · 2024-10-07T17:10:03Z

...a/server/multi-clusters/src/javaRestTest/java/org/elasticsearch/xpack/esql/ccq/Clusters.java

@@ -43,6 +44,7 @@ public static ElasticsearchCluster localCluster(ElasticsearchCluster remoteClust
            .setting("cluster.remote.connections_per_cluster", "1")
            .shared(true)
            .setting("cluster.routing.rebalance.enable", "none")
+            .plugin("inference-service-test")


this plugin is used for testing semantic_text - it's an inference service that can create sparse or dense embeddings - that have no actual "semantic" meaning since they are not using a model, but they are supposed to be deterministic.

ioanatia · 2024-10-07T17:10:39Z

...ugin/esql/qa/testFixtures/src/main/java/org/elasticsearch/xpack/esql/CsvTestsDataLoader.java

+        Request request = new Request("PUT", "_inference/sparse_embedding/test_sparse_inference");
+        request.setJsonEntity("""
+                  {
+                   "service": "test_service",


this is the inference service from the inference-service-test test plugin

This one's probably worth javadoc to explain that it's for the semantic text fields.

It kinda surprises me that this is the first capability conditional in the test infra. How expensive is this service endpoint? Conditionally creating it is of course fine, but the test infra would be simpler if it was unconditionally registered.

The test endpoint is quite lightweight. We could always register it.

javadoc was added

nik9000

Makes sense to me. @not-napoleon might want to take a look because this is beginning to follow your lead.

nik9000 · 2024-10-07T18:05:08Z

.../multi-node/src/javaRestTest/java/org/elasticsearch/xpack/esql/qa/multi_node/EsqlSpecIT.java

@@ -14,7 +14,7 @@

 public class EsqlSpecIT extends EsqlSpecTestCase {
    @ClassRule
-    public static ElasticsearchCluster cluster = Clusters.testCluster(spec -> {});
+    public static ElasticsearchCluster cluster = Clusters.testCluster(spec -> spec.plugin("inference-service-test"));


Will everything need this plugin? should it be in testCluster?

I think only this one needs it - it's the only one for multi_node that extends from EsqlSpecTestCase that loads the CSV data sets.

nik9000 · 2024-10-07T18:06:24Z

...ugin/esql/qa/testFixtures/src/main/java/org/elasticsearch/xpack/esql/CsvTestsDataLoader.java

+        Request request = new Request("PUT", "_inference/sparse_embedding/test_sparse_inference");
+        request.setJsonEntity("""
+                  {
+                   "service": "test_service",


This one's probably worth javadoc to explain that it's for the semantic text fields.

x-pack/plugin/esql/qa/testFixtures/src/main/resources/semantic_text.csv-spec

...nference/src/main/java/org/elasticsearch/xpack/inference/mapper/SemanticTextFieldMapper.java

ChrisHegarty

LGTM

carlosdelest

semantic_text support, yay! 💯

...nference/src/main/java/org/elasticsearch/xpack/inference/mapper/SemanticTextFieldMapper.java

carlosdelest · 2024-10-08T11:55:35Z

...ugin/esql/qa/testFixtures/src/main/java/org/elasticsearch/xpack/esql/CsvTestsDataLoader.java

+        Request request = new Request("PUT", "_inference/sparse_embedding/test_sparse_inference");
+        request.setJsonEntity("""
+                  {
+                   "service": "test_service",


The test endpoint is quite lightweight. We could always register it.

ioanatia · 2024-10-08T14:29:39Z

There are some issues with loading this plugin for multi-cluster and mixed-versions when one of the cluster nodes in on 8.16.0:

       Caused by:
        java.lang.RuntimeException: Execution of elasticsearch-plugin failed with exit code 70
            at org.elasticsearch.test.cluster.local.AbstractLocalClusterFactory$Node.runToolScript(AbstractLocalClusterFactory.java:795)
            at org.elasticsearch.test.cluster.local.AbstractLocalClusterFactory$Node.installPlugins(AbstractLocalClusterFactory.java:633)
            at org.elasticsearch.test.cluster.local.AbstractLocalClusterFactory$Node.start(AbstractLocalClusterFactory.java:159)
            at org.elasticsearch.test.cluster.local.DefaultLocalClusterHandle.lambda$start$0(DefaultLocalClusterHandle.java:78)
            at java.base/java.util.stream.ForEachOps$ForEachOp$OfRef.accept(ForEachOps.java:184)
            at java.base/java.util.AbstractList$RandomAccessSpliterator.forEachRemaining(AbstractList.java:722)
            at java.base/java.util.stream.AbstractPipeline.copyInto(AbstractPipeline.java:556)
            at java.base/java.util.stream.ForEachOps$ForEachTask.compute(ForEachOps.java:291)
            at java.base/java.util.concurrent.CountedCompleter.exec(CountedCompleter.java:759)
            at java.base/java.util.concurrent.ForkJoinTask.doExec(ForkJoinTask.java:507)
            at java.base/java.util.concurrent.ForkJoinPool$WorkQueue.topLevelExec(ForkJoinPool.java:1491)
            at java.base/java.util.concurrent.ForkJoinPool.scan(ForkJoinPool.java:2073)
            at java.base/java.util.concurrent.ForkJoinPool.runWorker(ForkJoinPool.java:2035)
            at java.base/java.util.concurrent.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:187)

multi-node and single-node should continue to work just fine.
For now I conditionally disabled loading the plugin in multi-cluster and mixed-versions when one of the cluster nodes is on 8.x. Once we get a fix for the plugin, we can remove this check, but I don't have that much familiarity with the plugin and don't know how to fix it. I reached out to the team that owns it.

The checks CsvTestsDataLoader have also been changed - they no longer check for SEMANTIC_TEXT_TYPE ESQL capability, but rather check whether a dataset requires the test inference endpoint. This kind of makes more sense than to check for an ES|QL capability, we are not using ES|QL APIs to load the data, but the create index and bulk APIs.
If the test inference endpoint does not exist and a dataset requires it, we simply do not load the dataset.
In EsqlSpecTestCase I added another check for shouldSkipTest - in case SEMANTIC_TEXT_TYPE is required we also check that we can use the test inference service. This makes sense because if the test inference service we have no data with semantic_text fields.

ChrisHegarty · 2024-10-19T11:33:18Z

Ok, mixed node testing will be avoided until the following issue is resolved as a follow up. #115166. (given that mixed mode will only be relevant when this PR is backported)

ChrisHegarty · 2024-10-19T11:38:48Z

@elasticmachine update branch

ChrisHegarty

This PR has had a number of review iterations. It's a great first step towards support for semantic search in ES|QL. LGTM 👍

ChrisHegarty · 2024-10-19T21:48:54Z

@elasticmachine update branch

ChrisHegarty · 2024-10-20T20:45:03Z

@elasticmachine update branch

ioanatia · 2024-10-21T11:27:39Z

@elasticmachine test this please

ioanatia · 2024-10-21T12:23:17Z

@elasticmachine update branch

ioanatia · 2024-10-21T13:40:18Z

Thank you folks, especially @fang-xing-esql and @ChrisHegarty for your help on this PR!
All the comment have been addressed, CI is green, the PR has a final approval from Chris after we addressed the comments, so it looks good to go 🚀

elasticsearchmachine · 2024-10-21T13:44:10Z

💔 Backport failed

The backport operation could not be completed due to the following error:

An unexpected error occurred when attempting to backport this PR.

You can use sqren/backport to manually backport by running backport --upstream elastic/elasticsearch --pr 113920

* Add initial support for semantic_text field type * Update docs/changelog/113920.yaml * More tests and fixes * Use mock inference service * Fix tests * Spotless * Fix mixed-cluster and multi-clusters tests * sort * Attempt another fix for bwc tests * Spotless * Fix merge * Attempt another fix * Don't load the inference-service-test plugin for mixed versions/clusters * Add more tests, address review comments * trivial * revert * post-merge fix block loader * post-merge fix compile * add mixed version testing * whitespace * fix MultiClusterSpecIT * add more fields to mapping * Revert mixed version testing * whitespace --------- Co-authored-by: ChrisHegarty <[email protected]> Co-authored-by: Elastic Machine <[email protected]>

…5256) * Add initial support for semantic_text field type * Update docs/changelog/113920.yaml * More tests and fixes * Use mock inference service * Fix tests * Spotless * Fix mixed-cluster and multi-clusters tests * sort * Attempt another fix for bwc tests * Spotless * Fix merge * Attempt another fix * Don't load the inference-service-test plugin for mixed versions/clusters * Add more tests, address review comments * trivial * revert * post-merge fix block loader * post-merge fix compile * add mixed version testing * whitespace * fix MultiClusterSpecIT * add more fields to mapping * Revert mixed version testing * whitespace --------- Co-authored-by: ChrisHegarty <[email protected]> Co-authored-by: Elastic Machine <[email protected]>

* Add initial support for semantic_text field type * Update docs/changelog/113920.yaml * More tests and fixes * Use mock inference service * Fix tests * Spotless * Fix mixed-cluster and multi-clusters tests * sort * Attempt another fix for bwc tests * Spotless * Fix merge * Attempt another fix * Don't load the inference-service-test plugin for mixed versions/clusters * Add more tests, address review comments * trivial * revert * post-merge fix block loader * post-merge fix compile * add mixed version testing * whitespace * fix MultiClusterSpecIT * add more fields to mapping * Revert mixed version testing * whitespace --------- Co-authored-by: ChrisHegarty <[email protected]> Co-authored-by: Elastic Machine <[email protected]>

ioanatia added >enhancement :Search Relevance/Search Catch all for Search Relevance labels Oct 2, 2024

elasticsearchmachine added the v9.0.0 label Oct 2, 2024

ioanatia and others added 3 commits October 4, 2024 14:57

Add initial support for semantic_text field type

4e3ce86

Update docs/changelog/113920.yaml

ce76729

More tests and fixes

36e4c7d

ioanatia force-pushed the semantic_text_support branch from 8537aaa to 36e4c7d Compare October 4, 2024 13:07

ioanatia added 4 commits October 4, 2024 20:36

Use mock inference service

c2a67e5

Fix tests

07f023a

Spotless

6ce6c7d

Fix mixed-cluster and multi-clusters tests

9954098

ioanatia marked this pull request as ready for review October 7, 2024 17:02

ioanatia requested a review from a team as a code owner October 7, 2024 17:02

elasticsearchmachine added the Team:Search Relevance Meta label for the Search Relevance team in Elasticsearch label Oct 7, 2024

ioanatia added the Team:Analytics Meta label for analytical engine team (ESQL/Aggs/Geo) label Oct 7, 2024

elasticsearchmachine removed the Team:Analytics Meta label for analytical engine team (ESQL/Aggs/Geo) label Oct 7, 2024

ioanatia commented Oct 7, 2024

View reviewed changes

nik9000 reviewed Oct 7, 2024

View reviewed changes

mark-vieira approved these changes Oct 7, 2024

View reviewed changes

costin requested a review from fang-xing-esql October 7, 2024 23:11

tteofili approved these changes Oct 8, 2024

View reviewed changes

...nference/src/main/java/org/elasticsearch/xpack/inference/mapper/SemanticTextFieldMapper.java Outdated Show resolved Hide resolved

ChrisHegarty approved these changes Oct 8, 2024

View reviewed changes

ChrisHegarty added 2 commits October 8, 2024 12:10

sort

60615e7

Merge branch 'main' into semantic_text_support

e70acaa

carlosdelest approved these changes Oct 8, 2024

View reviewed changes

Attempt another fix for bwc tests

1aec2d1

Revert mixed version testing

19fc260

whitespace

3f847a2

Merge branch 'main' into semantic_text_support

79bef4d

ChrisHegarty added the test-release Trigger CI checks against release build label Oct 19, 2024

ChrisHegarty approved these changes Oct 19, 2024

View reviewed changes

ChrisHegarty changed the title ~~Add initial support for semantic_text field type~~ ES|QL Add initial support for semantic_text field type Oct 19, 2024

Merge branch 'main' into semantic_text_support

198f635

elastic deleted a comment from elasticmachine Oct 20, 2024

elasticmachine and others added 2 commits October 20, 2024 22:45

Merge branch 'main' into semantic_text_support

fec65c4

Merge branch 'main' into semantic_text_support

b38aa6c

ioanatia removed the test-release Trigger CI checks against release build label Oct 21, 2024

Merge branch 'main' into semantic_text_support

9027cfd

ioanatia added auto-backport Automatically create backport pull requests when merged v8.17.0 labels Oct 21, 2024

ioanatia merged commit fd43adc into elastic:main Oct 21, 2024
16 checks passed

elasticsearchmachine added the backport pending label Oct 21, 2024

ioanatia deleted the semantic_text_support branch October 21, 2024 13:44

ioanatia mentioned this pull request Oct 21, 2024

ES|QL Add initial support for semantic_text field type (#113920) #115256

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

ES|QL Add initial support for semantic_text field type #113920

ES|QL Add initial support for semantic_text field type #113920

ioanatia commented Oct 2, 2024 •

edited by ChrisHegarty

Loading

elasticsearchmachine commented Oct 2, 2024

elasticsearchmachine commented Oct 7, 2024

ioanatia Oct 7, 2024

ioanatia Oct 7, 2024

nik9000 Oct 7, 2024

ChrisHegarty Oct 8, 2024

carlosdelest Oct 8, 2024

ioanatia Oct 17, 2024

nik9000 left a comment

nik9000 Oct 7, 2024

ioanatia Oct 8, 2024

nik9000 Oct 7, 2024

ChrisHegarty left a comment

carlosdelest left a comment

carlosdelest Oct 8, 2024

ioanatia commented Oct 8, 2024

ChrisHegarty commented Oct 19, 2024

ChrisHegarty commented Oct 19, 2024

ChrisHegarty left a comment •

edited

Loading

ChrisHegarty commented Oct 19, 2024

ChrisHegarty commented Oct 20, 2024

ioanatia commented Oct 21, 2024

ioanatia commented Oct 21, 2024

ioanatia commented Oct 21, 2024

elasticsearchmachine commented Oct 21, 2024

ES|QL Add initial support for semantic_text field type #113920

ES|QL Add initial support for semantic_text field type #113920

Conversation

ioanatia commented Oct 2, 2024 • edited by ChrisHegarty Loading

elasticsearchmachine commented Oct 2, 2024

elasticsearchmachine commented Oct 7, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

nik9000 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ChrisHegarty left a comment

Choose a reason for hiding this comment

carlosdelest left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ioanatia commented Oct 8, 2024

ChrisHegarty commented Oct 19, 2024

ChrisHegarty commented Oct 19, 2024

ChrisHegarty left a comment • edited Loading

Choose a reason for hiding this comment

ChrisHegarty commented Oct 19, 2024

ChrisHegarty commented Oct 20, 2024

ioanatia commented Oct 21, 2024

ioanatia commented Oct 21, 2024

ioanatia commented Oct 21, 2024

elasticsearchmachine commented Oct 21, 2024

💔 Backport failed

ioanatia commented Oct 2, 2024 •

edited by ChrisHegarty

Loading

ChrisHegarty left a comment •

edited

Loading