Implemented ChatCompletion task for Google VertexAI with Gemini Models #128105

leo-hoet · 2025-05-16T18:44:21Z

This PR implements the task type chat_completion for Google Vertex AI in the inference api

cla-checker-service · 2025-05-16T18:44:25Z

💚 CLA has been signed

beltrangs · 2025-05-16T20:43:38Z

❌ Author of the following commits did not sign a Contributor Agreement: 00a6636, 9be2a44

Please, read and sign the above mentioned agreement if you want to contribute to this project

Done, both developers have signed. Thanks!

…for compatibility

elasticsearchmachine · 2025-05-19T19:34:23Z

Pinging @elastic/ml-core (Team:ML)

jonathan-buttner

PR is looking good, here is some initial feedback, I haven't gotten through the whole PR yet.

jonathan-buttner · 2025-05-19T19:42:32Z

docs/changelog/128105.yaml

@@ -0,0 +1,5 @@
+pr: 128105
+summary: "Google VertexAI integration now supports chat_completion task"


Suggested change

summary: "Google VertexAI integration now supports chat_completion task"

summary: "Adding Google VertexAI chat completion integration"

jonathan-buttner · 2025-05-19T19:42:54Z

server/src/main/java/org/elasticsearch/TransportVersions.java

@@ -254,6 +254,7 @@ static TransportVersion def(int id) {
    public static final TransportVersion ESQL_FIELD_ATTRIBUTE_DROP_TYPE = def(9_075_0_00);
    public static final TransportVersion ESQL_TIME_SERIES_SOURCE_STATUS = def(9_076_0_00);
    public static final TransportVersion ESQL_HASH_OPERATOR_STATUS_OUTPUT_TIME = def(9_077_0_00);
+    public static final TransportVersion ML_INFERENCE_VERTEXAI_CHATCOMPLETION_ADDED = def(9_078_0_00);


We'll want to backport this to 8.19. To do that we need to reserve a transport version for 8.19 but in the main branch.

Let's add another transport version similar to what I did here: https://github.com/elastic/elasticsearch/pull/126805/files#diff-85e782e9e33a0f8ca8e99b41c17f9d04e3a7981d435abf44a3aa5d954a47cd8fR175

public static final TransportVersion ML_INFERENCE_VERTEXAI_CHATCOMPLETION_ADDED_8_19 = def(8_841_0_30);

Or whatever the latest version number is (it might be 30, or 31 etc).

jonathan-buttner · 2025-05-19T19:43:57Z

...icsearch/xpack/inference/services/googlevertexai/GoogleVertexAiCompletionRequestManager.java

+import java.util.Objects;
+import java.util.function.Supplier;
+
+public class GoogleVertexAiCompletionRequestManager extends GoogleVertexAiRequestManager {


We're trying to transition away from the request manager pattern to avoid the extra class since all the classes are pretty similar.

Here's an example of how we implemented it for voyageai: #124512

Here's how we do it for chat completions in openai: https://github.com/elastic/elasticsearch/blob/d2be03c946c94943dca8fe5da75a125fa70ddaa6/x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/services/openai/action/OpenAiActionCreator.java

If we could switch to using a generic request manager that'd be great.

I made to swtich to use a generic request and it compiles and works fine. My only fear is that I had to change the base class of GoogleVertexAiModel from Model to RateLimitGroupingModel. Does that have any implication that i am not aware of?

Awesome! No that should be fine. Thanks for making that change.

jonathan-buttner · 2025-05-19T20:02:13Z

...ck/inference/services/googlevertexai/GoogleVertexAiUnifiedChatCompletionResponseHandler.java

+            ) {
+                return ERROR_PARSER.apply(parser, null).orElse(ErrorResponse.UNDEFINED_ERROR);
+            } catch (Exception e) {
+                logger.warn("Failed to parse Google Vertex AI error response body", e);


We'll likely refactor the error logic for all the services to return whatever is sent even if we can't parse it. If we fail to parse the error response, how about we return a new ErrorResponse but just put the body as the message:

var resultAsString = new String(httpResult.body(), StandardCharsets.UTF_8); return new ErrorResponse(Strings.format("Unable to parse the Google Vertex AI error, response body: [%s]", resultAsString));

Sounds good!

jonathan-buttner · 2025-05-19T20:05:06Z

...ck/inference/services/googlevertexai/GoogleVertexAiUnifiedChatCompletionResponseHandler.java

+            ) {
+                return ERROR_PARSER.apply(parser, null).orElse(ErrorResponse.UNDEFINED_ERROR);
+            } catch (Exception e) {
+                logger.warn("Failed to parse Google Vertex AI error string", e);


Same as comment above.

jonathan-buttner · 2025-05-19T20:25:15Z

...erence/services/googlevertexai/request/GoogleVertexAiUnifiedChatCompletionRequestEntity.java

+            "Role [%s] not supported by Google VertexAI ChatCompletion. Supported roles: [%s, %s]",
+            messageRole,
+            USER_ROLE,
+            MODEL_ROLE


Just a reminder to switch this to assistant.

jonathan-buttner · 2025-05-19T20:29:20Z

...erence/services/googlevertexai/request/GoogleVertexAiUnifiedChatCompletionRequestEntity.java

+
+        try (XContentParser parser = XContentFactory.xContent(XContentType.JSON).createParser(parserConfig, jsonString)) {
+            XContentParser.Token token = parser.nextToken();
+            if (token != XContentParser.Token.START_OBJECT) {


If we omit this check, what is the error that is returned?

We also might be able to leverage the helper method:

ensureExpectedToken(XContentParser.Token.START_OBJECT, token, parser);

VertexAI request expected the arguments to be a map, and in the current specs the function arguments is a string, so I am using that method to convert the data between both. If the check is not there, depending on which case, it can fail with org.elasticsearch.xcontent.XContentParseException: Unrecognized token or return an empty object. I put that check there so we are sure that the string being parsed is an object and not any other json token

Gotcha, I believe ensureExpectedToken(XContentParser.Token.START_OBJECT, token, parser); checks the same thing right? Or is the error message it produces not sufficient?

Will change this to use ensureExpectedToken method

…ntParser

…medWriteablesProvider. Added InferenceSettingsTests

…ADDED to the right location

…ExpectedToken`

jonathan-buttner

Left some more suggestions, thanks for implementing the initial changes!

jonathan-buttner · 2025-05-21T18:54:27Z

server/src/main/java/org/elasticsearch/TransportVersions.java

@@ -254,6 +254,7 @@ static TransportVersion def(int id) {
    public static final TransportVersion ESQL_FIELD_ATTRIBUTE_DROP_TYPE = def(9_075_0_00);
    public static final TransportVersion ESQL_TIME_SERIES_SOURCE_STATUS = def(9_076_0_00);
    public static final TransportVersion ESQL_HASH_OPERATOR_STATUS_OUTPUT_TIME = def(9_077_0_00);
+    public static final TransportVersion ML_INFERENCE_VERTEXAI_CHATCOMPLETION_ADDED_8_19 = def(8_841_0_30);


Sorry I meant we'll need two transport versions. Let's move ML_INFERENCE_VERTEXAI_CHATCOMPLETION_ADDED_8_19 to be with the other 8_19 style versions.

We'll need to create another one called ML_INFERENCE_VERTEXAI_CHATCOMPLETION_ADDED and have it's version be 9_078_0_00 or whatever the latest is.

Got it! Fixed

Which one should GoogleVertexAiChatCompletionServiceSettings.getMinimalSupportedVersion return? Right now it's returning ML_INFERENCE_VERTEXAI_CHATCOMPLETION_ADDED_8_19

For this PR we'll want it to point to the 9 version (not the 8_19) one. For the backport we'll switch it to be the 8.19 version.

jonathan-buttner · 2025-05-21T19:21:33Z

...erence/services/googlevertexai/request/GoogleVertexAiUnifiedChatCompletionRequestEntity.java

+
+        try (XContentParser parser = XContentFactory.xContent(XContentType.JSON).createParser(parserConfig, jsonString)) {
+            XContentParser.Token token = parser.nextToken();
+            if (token != XContentParser.Token.START_OBJECT) {


Gotcha, I believe ensureExpectedToken(XContentParser.Token.START_OBJECT, token, parser); checks the same thing right? Or is the error message it produces not sufficient?

jonathan-buttner · 2025-05-21T19:25:51Z

...erence/services/googlevertexai/request/GoogleVertexAiUnifiedChatCompletionRequestEntity.java

+        builder.startArray(PARTS);
+        for (var systemMessage : systemMessages) {
+            switch (systemMessage.content()) {
+                case UnifiedCompletionRequest.ContentString contentString -> {


We can leave this but just an heads up that when we go to backport the changes to the 8.x branch it's going to complain because that branch isn't on the JDK version that supports this type of switch statement. It might be easier to change it here even though the IDE will complain to avoid having to adjust in the backport. Up to you.

Got it, will change this switch and others that I made for if-else . I think that should work

jonathan-buttner · 2025-05-21T19:29:06Z

...erence/services/googlevertexai/request/GoogleVertexAiUnifiedChatCompletionRequestEntity.java

+            return;
+        }
+
+        builder.startArray(TOOLS);


nit: adding some indentation with scoping via {} could help with understanding the nesting here, optional though.

jonathan-buttner · 2025-05-21T20:45:17Z

...ck/inference/services/googlevertexai/GoogleVertexAiUnifiedChatCompletionResponseHandler.java

+    private static final String ERROR_STATUS_FIELD = "status";
+
+    public GoogleVertexAiUnifiedChatCompletionResponseHandler(String requestType, ResponseParser parseFunction) {
+        super(requestType, parseFunction, GoogleVertexAiErrorResponse::fromResponse, true);


This is an area of the code that we need to refactor. The parseFunction is only used in non-streaming cases. So I think we can actually pass in an empty lambda style function, maybe just one that's defined statically.

Great, so we can also remove GoogleVertexAiChatCompletionResponseEntity since we are only doing streaming responses and that class is not being used, right?

Nevermind, saw you response in another comment. Will delete GoogleVertexAiChatCompletionResponseEntity and refactor the code as you suggested

We can wait to delete it if you like. Until we get a response to you about whether we want to include the completion task type. If we do that, we'll want to implement both streaming and non-stream for completion.

jonathan-buttner · 2025-05-21T21:07:07Z

...k/inference/services/googlevertexai/response/GoogleVertexAiChatCompletionResponseEntity.java

+
+            StringBuilder fullText = new StringBuilder();
+
+            while (parser.nextToken() != XContentParser.Token.END_ARRAY) {


I think we can use XContentParserUtils.parseList here instead.

I removed this class from this PR. If we are doing completion in another PR I will add it there

jonathan-buttner · 2025-05-21T21:09:39Z

...k/inference/services/googlevertexai/response/GoogleVertexAiChatCompletionResponseEntity.java

+            while (parser.nextToken() != XContentParser.Token.END_ARRAY) {
+                ensureExpectedToken(XContentParser.Token.START_OBJECT, parser.currentToken(), parser);
+                Chunk chunk = Chunk.PARSER.apply(parser, null);
+                chunk.extractText().ifPresent(fullText::append);


This class would be used for non-streaming scenarios. If we get multiple entries in the array could those be for separate input values in the originating request?

Like:

{"input": ["text 1", "text 2"]}

Would we get 2 items in the array from the upstream server? If so, I don't think we want to combine the text as we'd want to return a list of 2 items below.

I removed this class from this PR since the chat completion only supports streaming. If we are doing completion in another PR I will add it there

jonathan-buttner · 2025-05-21T21:12:08Z

...st/java/org/elasticsearch/xpack/inference/services/elastic/ElasticInferenceServiceTests.java

+                assertThat(
+                    httpRequest.getBody().toString(),
+                    equalTo(
+                        "{\"messages\":[{\"content\":\"Hello\",\"role\":\"user\"}],\"n\":1,\"stream\":true,\"stream_options\":{\"include_usage\":true},\"model\":\"gemini-2.0-flash-001\"}"


Let's use XContentHelper.stripWhitespace() for things like this. That way we can create more readable multiline string in this file and strip the white space when we compare it for equality.

jonathan-buttner · 2025-05-21T21:13:03Z

...st/java/org/elasticsearch/xpack/inference/services/elastic/ElasticInferenceServiceTests.java

@@ -1299,6 +1299,64 @@ private InferenceEventsAssertion testUnifiedStream(int responseCode, String resp
        }
    }

+    public void testUnifiedCompletionInfer_WithGoogleVertexAiModel() throws IOException {


Let's move this to the google vertex service test file.

Yeah sorry, this test got slipped in when we were testing some things. It's not necessary so I will remove it

jonathan-buttner · 2025-05-21T21:14:23Z

...st/java/org/elasticsearch/xpack/inference/services/elastic/ElasticInferenceServiceTests.java

+                );
+            } finally {
+                // Clean up the thread context
+                threadPool.getThreadContext().stashContext();


Why do we need to stash the context here? Typically we terminate the thread pool after tests: https://github.com/elastic/elasticsearch/blob/main/x-pack/plugin/inference/src/test/java/org/elasticsearch/xpack/inference/services/openai/OpenAiServiceTests.java#L119

Ah I see, that's in some of our other tests. I don't believe we need that. Let me know if the test starts failing after we remove it though.

… new one for ML_INFERENCE_VERTEXAI_CHATCOMPLETION_ADDDED

…mproved indentation via `{}`

…e around it.

jonathan-buttner

Thanks for the changes, left one more question

jonathan-buttner · 2025-05-27T18:55:48Z

...main/java/org/elasticsearch/xpack/inference/services/googlevertexai/GoogleVertexAiModel.java

+    @Override
+    public int rateLimitGroupingHash() {
+        // In VertexAI rate limiting is scoped to the project and the model. URI already has this information so we are using that
+        return Objects.hash(uri);


Just to clarify, it's not based on the service account key information too?

Can you add a link to the docs that indicates this?

Great! Will do. https://ai.google.dev/gemini-api/docs/rate-limits

Rate limits are applied per project, not per API key.

Also on the VertexAI quotas https://cloud.google.com/vertex-ai/docs/quotas#request_quotas

The following quotas apply to Vertex AI requests for a given project and supported region...

Some resources may not be affected by the region, but I choose to be conservative and go with a safe default

jonathan-buttner · 2025-05-27T19:02:53Z

...icsearch/xpack/inference/services/googlevertexai/GoogleVertexAiCompletionRequestManager.java

+import java.util.Objects;
+import java.util.function.Supplier;
+
+public class GoogleVertexAiCompletionRequestManager extends GoogleVertexAiRequestManager {


Awesome! No that should be fine. Thanks for making that change.

# Conflicts: # server/src/main/java/org/elasticsearch/TransportVersions.java

jonathan-buttner

Thanks for the changes!

jonathan-buttner · 2025-05-28T15:23:28Z

@elasticsearchmachine test this please

jonathan-buttner · 2025-05-28T15:48:25Z

@elasticsearchmachine run

jonathan-buttner · 2025-05-28T15:50:31Z

@elasticmachine test this please

jonathan-buttner · 2025-05-28T16:05:52Z

...erence/services/googlevertexai/request/GoogleVertexAiUnifiedChatCompletionRequestEntity.java

+    }
+
+    private String messageRoleToGoogleVertexAiSupportedRole(String messageRole) {
+        var messageRoleLowered = messageRole.toLowerCase();


Looks like CI is complaining about using the default locale here:

> Task :x-pack:plugin:inference:forbiddenApisMain | Forbidden method invocation: java.lang.String#toLowerCase() [Uses default locale] | in org.elasticsearch.xpack.inference.services.googlevertexai.request.GoogleVertexAiUnifiedChatCompletionRequestEntity (GoogleVertexAiUnifiedChatCompletionRequestEntity.java:73) | Scanned 890 class file(s) for forbidden API invocations (in 0.78s), 1 error(s). |

I think we can use Locale.ROOT instead.

jonathan-buttner · 2025-05-28T17:33:15Z

@elasticmachine test this please

jonathan-buttner · 2025-05-28T18:01:39Z

...ence/services/googlevertexai/completion/GoogleVertexAIChatCompletionServiceSettingsTest.java

@@ -0,0 +1,39 @@
+/*


From CI:

Caused by: org.gradle.api.GradleException: Following test classes do not match naming convention to use suffix 'Tests': -- | org.elasticsearch.xpack.inference.services.googlevertexai.completion.GoogleVertexAIChatCompletionServiceSettingsTest

The file name needs to be: GoogleVertexAIChatCompletionServiceSettingsTests (trailing s).

…en api

jonathan-buttner · 2025-05-28T20:55:13Z

@elasticmachine test this please

jonathan-buttner

Looks like we have a failing test:


REPRODUCE WITH: ./gradlew ":x-pack:plugin:inference:qa:inference-service-tests:javaRestTest" --tests "org.elasticsearch.xpack.inference.InferenceGetServicesIT.testGetServicesWithChatCompletionTaskType" -Dtests.seed=54CCE93B3DEDB87E -Dtests.locale=su-Latn -Dtests.timezone=Asia/Aqtobe -Druntime.java=24
--
  |  
  | InferenceGetServicesIT > testGetServicesWithChatCompletionTaskType FAILED
  | java.lang.AssertionError:
  | Expected: <6>
  | but: was <7>
  | at __randomizedtesting.SeedInfo.seed([54CCE93B3DEDB87E:DC4FE40BF96FF6C7]:0)
  | at org.hamcrest.MatcherAssert.assertThat(MatcherAssert.java:20)
  | at org.hamcrest.MatcherAssert.assertThat(MatcherAssert.java:6)
  | at org.elasticsearch.test.ESTestCase.assertThat(ESTestCase.java:2653)
  | at org.elasticsearch.xpack.inference.InferenceGetServicesIT.testGetServicesWithChatCompletionTaskType(InferenceGetServicesIT.java:154)

I think we just need to bump the value.

lhoet-google · 2025-05-29T12:43:10Z

Looks like we have a failing test:


REPRODUCE WITH: ./gradlew ":x-pack:plugin:inference:qa:inference-service-tests:javaRestTest" --tests "org.elasticsearch.xpack.inference.InferenceGetServicesIT.testGetServicesWithChatCompletionTaskType" -Dtests.seed=54CCE93B3DEDB87E -Dtests.locale=su-Latn -Dtests.timezone=Asia/Aqtobe -Druntime.java=24
--
  |  
  | InferenceGetServicesIT > testGetServicesWithChatCompletionTaskType FAILED
  | java.lang.AssertionError:
  | Expected: <6>
  | but: was <7>
  | at __randomizedtesting.SeedInfo.seed([54CCE93B3DEDB87E:DC4FE40BF96FF6C7]:0)
  | at org.hamcrest.MatcherAssert.assertThat(MatcherAssert.java:20)
  | at org.hamcrest.MatcherAssert.assertThat(MatcherAssert.java:6)
  | at org.elasticsearch.test.ESTestCase.assertThat(ESTestCase.java:2653)
  | at org.elasticsearch.xpack.inference.InferenceGetServicesIT.testGetServicesWithChatCompletionTaskType(InferenceGetServicesIT.java:154)

I think we just need to bump the value.

Working on it. If I do ./gradlew checkPart2 that should run the whole failing pipeline, right?

lhoet-google · 2025-05-29T14:20:53Z

@elasticmachine test this please

jonathan-buttner · 2025-05-29T15:02:37Z

Working on it. If I do ./gradlew checkPart2 that should run the whole failing pipeline, right?

This is the command: ./gradlew ":x-pack:plugin:inference:qa:inference-service-tests:javaRestTest" --tests "org.elasticsearch.xpack.inference.InferenceGetServicesIT.testGetServicesWithChatCompletionTaskType" -Dtests.seed=54CCE93B3DEDB87E -Dtests.locale=su-Latn -Dtests.timezone=Asia/Aqtobe -Druntime.java=24

That'll run it the same way that CI did. Or if you want to run all the rest tests I think this would work: ./gradlew ":x-pack:plugin:inference:qa:inference-service-tests:javaRestTest"

jonathan-buttner · 2025-05-29T15:03:30Z

@elasticmachine test this please

elasticsearchmachine · 2025-05-29T17:36:40Z

💔 Backport failed

Status	Branch	Result
❌	8.19	Commit could not be cherrypicked due to conflicts

You can use sqren/backport to manually backport by running backport --upstream elastic/elasticsearch --pr 128105

elastic#128105) * Implemented ChatCompletion task for Google VertexAI with Gemini Models * changelog * System Instruction bugfix * Mapping role assistant -> model in vertex ai chat completion request for compatibility * GoogleVertexAI chat completion using SSE events. Removed JsonArrayEventParser * Removed buffer from GoogleVertexAiUnifiedStreamingProcessor * Casting inference inputs with `castoTo` * Registered GoogleVertexAiChatCompletionServiceSettings in InferenceNamedWriteablesProvider. Added InferenceSettingsTests * Changed transport version to 8_19 for vertexai chatcompletion * Fix to transport version. Moved ML_INFERENCE_VERTEXAI_CHATCOMPLETION_ADDED to the right location * VertexAI Chat completion request entity jsonStringToMap using `ensureExpectedToken` * Fixed TransportVersions. Left vertexAi chat completion 8_19 and added new one for ML_INFERENCE_VERTEXAI_CHATCOMPLETION_ADDDED * Refactor switch statements by if-else for older java compatibility. Improved indentation via `{}` * Removed GoogleVertexAiChatCompletionResponseEntity and refactored code around it. * Removed redundant test `testUnifiedCompletionInfer_WithGoogleVertexAiModel` * Returning whole body when fail to parse response from VertexAI * Refactor use GenericRequestManager instead of GoogleVertexAiCompletionRequestManager * Refactored to constructorArg for mandatory args in GoogleVertexAiUnifiedStreamingProcessor * Changed transport version in GoogleVertexAiChatCompletionServiceSettings * Bugfix in tool calling with role tool * GoogleVertexAiModel added documentation info on rateLimitGroupingHash * [CI] Auto commit changes from spotless * Fix: using Locale.ROOT when calling toLowerCase * Fix: Renamed test class to match convention & modified use of forbidden api * Fix: Failing test in InferenceServicesIT --------- Co-authored-by: lhoet <[email protected]> Co-authored-by: Jonathan Buttner <[email protected]> Co-authored-by: elasticsearchmachine <[email protected]>

Implemented ChatCompletion task for Google VertexAI with Gemini Models

00a6636

elasticsearchmachine added v9.1.0 needs:triage Requires assignment of a team area label external-contributor Pull request authored by a developer outside the Elasticsearch team labels May 16, 2025

changelog

9be2a44

lhoet-google added 2 commits May 19, 2025 16:02

System Instruction bugfix

c2387e8

Mapping role assistant -> model in vertex ai chat completion request …

50770ea

…for compatibility

jonathan-buttner added the v8.19.0 label May 19, 2025

jonathan-buttner self-assigned this May 19, 2025

jonathan-buttner added >enhancement :ml Machine learning Team:ML Meta label for the ML team auto-backport Automatically create backport pull requests when merged and removed needs:triage Requires assignment of a team area label labels May 19, 2025

lhoet-google mentioned this pull request May 19, 2025

Vertexai chatcompletion leo-hoet/elasticsearch#1

Closed

jonathan-buttner requested changes May 19, 2025

View reviewed changes

lhoet-google added 7 commits May 20, 2025 11:43

GoogleVertexAI chat completion using SSE events. Removed JsonArrayEve…

42cbbe2

…ntParser

Removed buffer from GoogleVertexAiUnifiedStreamingProcessor

fe8e336

Casting inference inputs with castoTo

7c24f93

Registered GoogleVertexAiChatCompletionServiceSettings in InferenceNa…

2140d05

…medWriteablesProvider. Added InferenceSettingsTests

Changed transport version to 8_19 for vertexai chatcompletion

42dd376

Fix to transport version. Moved ML_INFERENCE_VERTEXAI_CHATCOMPLETION_…

0863316

…ADDED to the right location

VertexAI Chat completion request entity jsonStringToMap using `ensure…

f080e96

…ExpectedToken`

jonathan-buttner requested changes May 21, 2025

View reviewed changes

lhoet-google added 3 commits May 22, 2025 14:08

Fixed TransportVersions. Left vertexAi chat completion 8_19 and added…

8f6648f

… new one for ML_INFERENCE_VERTEXAI_CHATCOMPLETION_ADDDED

Refactor switch statements by if-else for older java compatibility. I…

848dc7a

…mproved indentation via `{}`

Removed GoogleVertexAiChatCompletionResponseEntity and refactored cod…

59862c6

…e around it.

jonathan-buttner reviewed May 27, 2025

View reviewed changes

leo-hoet added 2 commits May 27, 2025 17:33

GoogleVertexAiModel added documentation info on rateLimitGroupingHash

1ead8c5

Merge branch 'main' into google-vertexai-chatcompletion

ad9f0e1

# Conflicts: # server/src/main/java/org/elasticsearch/TransportVersions.java

jonathan-buttner approved these changes May 28, 2025

View reviewed changes

Merge branch 'main' into google-vertexai-chatcompletion

f4057f3

[CI] Auto commit changes from spotless

38b9ca4

jonathan-buttner reviewed May 28, 2025

View reviewed changes

Fix: using Locale.ROOT when calling toLowerCase

2e8dbee

jonathan-buttner reviewed May 28, 2025

View reviewed changes

Fix: Renamed test class to match convention & modified use of forbidd…

ddd19c5

…en api

jonathan-buttner reviewed May 29, 2025

View reviewed changes

leo-hoet added 2 commits May 29, 2025 09:56

Fix: Failing test in InferenceServicesIT

88a2780

Merge branch 'main' into google-vertexai-chatcompletion

b841e4e

jonathan-buttner approved these changes May 29, 2025

View reviewed changes

jonathan-buttner merged commit 107daf3 into elastic:main May 29, 2025
18 of 19 checks passed

elasticsearchmachine added the backport pending label May 29, 2025

This was referenced May 29, 2025

[8.19] [ML] Add Google VertexAi Chat completion integration #128632

Merged

Updated inference vertexai specification completion and chatcompletion elastic/elasticsearch-specification#4439

Merged

		@@ -0,0 +1,5 @@
		pr: 128105
		summary: "Google VertexAI integration now supports chat_completion task"

	summary: "Google VertexAI integration now supports chat_completion task"
	summary: "Adding Google VertexAI chat completion integration"


		StringBuilder fullText = new StringBuilder();

		while (parser.nextToken() != XContentParser.Token.END_ARRAY) {

Implemented ChatCompletion task for Google VertexAI with Gemini Models #128105

Implemented ChatCompletion task for Google VertexAI with Gemini Models #128105

Uh oh!

Conversation

leo-hoet commented May 16, 2025

Uh oh!

cla-checker-service bot commented May 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

beltrangs commented May 16, 2025

Uh oh!

elasticsearchmachine commented May 19, 2025

Uh oh!

jonathan-buttner left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jonathan-buttner left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lhoet-google May 22, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lhoet-google May 22, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cla-checker-service bot commented May 16, 2025 •

edited

Loading

jonathan-buttner left a comment •

edited

Loading

lhoet-google May 22, 2025 •

edited

Loading

lhoet-google May 22, 2025 •

edited

Loading

jonathan-buttner left a comment •

edited

Loading