Add Google Model Garden's Anthropic support to Inference Plugin#134080
Merged
jonathan-buttner merged 64 commits intoelastic:mainfrom Sep 29, 2025
Merged
Conversation
elasticsearchmachine
added
v9.2.0
external-contributor
…ntegration # Conflicts: # server/src/main/java/org/elasticsearch/TransportVersions.java # x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/services/googlevertexai/action/GoogleVertexAiActionCreator.java # x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/services/googlevertexai/request/completion/GoogleVertexAiUnifiedChatCompletionRequest.java # x-pack/plugin/inference/src/test/java/org/elasticsearch/xpack/inference/services/googlevertexai/action/GoogleVertexAiUnifiedChatCompletionActionTests.java # x-pack/plugin/inference/src/test/java/org/elasticsearch/xpack/inference/services/googlevertexai/completion/GoogleVertexAiChatCompletionModelTests.java # x-pack/plugin/inference/src/test/java/org/elasticsearch/xpack/inference/services/googlevertexai/request/completion/GoogleVertexAiUnifiedChatCompletionRequestTests.java
…ntegration # Conflicts: # server/src/main/java/org/elasticsearch/TransportVersions.java
…al parameters based on transport version
…ntegration # Conflicts: # server/src/main/java/org/elasticsearch/TransportVersions.java
… to support new content block types and improve parsing logic
… parser and add unit tests for response validation
…ate response parsing and error handling
…ity to validate serialization of user fields
…n model and update related tests
elasticsearchmachine
added
the
needs:triage
…ntegration # Conflicts: # server/src/main/java/org/elasticsearch/TransportVersions.java
Contributor
Author
|
Hello @jonathan-buttner @dan-rubinstein |
| public static final TransportVersion ESQL_DOCUMENTS_FOUND_AND_VALUES_LOADED_8_19 = def(8_841_0_61); | ||
| public static final TransportVersion ESQL_PROFILE_INCLUDE_PLAN_8_19 = def(8_841_0_62); | ||
| public static final TransportVersion INITIAL_ELASTICSEARCH_8_19_4 = def(8_841_0_68); | ||
| public static final TransportVersion ML_INFERENCE_GOOGLE_MODEL_GARDEN_ADDED_8_19 = def(8_841_0_69); |
Contributor
Author
There was a problem hiding this comment.
Let me know if this needs to be removed. I haven't seen backports in a while. But Google Vertex AI is there for quite some time, so probably we'd require one.
Contributor
There was a problem hiding this comment.
Let's remove this, we won't be backporting the changes.
Contributor
Author
There was a problem hiding this comment.
Removed.
…ntegration # Conflicts: # server/src/main/java/org/elasticsearch/TransportVersions.java
…ntegration # Conflicts: # server/src/main/java/org/elasticsearch/TransportVersions.java # x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/services/googlevertexai/completion/GoogleVertexAiChatCompletionServiceSettings.java
Contributor
Author
|
@jonathan-buttner your comments are addressed. Could you please take a look at the PR once more? |
…ntegration # Conflicts: # server/src/main/resources/transport/upper_bounds/9.2.csv
Contributor
jonathan-buttner
left a comment
jonathan-buttner
left a comment
There was a problem hiding this comment.
Looking good, couple more changes
| } | ||
|
|
||
| public GoogleVertexAiChatCompletionTaskSettings(StreamInput in) throws IOException { | ||
| thinkingConfig = new ThinkingConfig(in); | ||
| TransportVersion version = in.getTransportVersion(); | ||
| if (GoogleVertexAiUtils.supportsModelGarden(version)) { | ||
| maxTokens = Objects.requireNonNullElse(in.readOptionalInt(), DEFAULT_MAX_TOKENS); | ||
| maxTokens = in.readOptionalInt(); |
Contributor
There was a problem hiding this comment.
Can we use readOptionalVInt?
Contributor
Author
There was a problem hiding this comment.
Good thinking. Done.
| @@ -124,7 +124,9 @@ public TransportVersion getMinimalSupportedVersion() { | |||
| @Override | |||
| public void writeTo(StreamOutput out) throws IOException { | |||
| thinkingConfig.writeTo(out); | |||
| out.writeOptionalInt(maxTokens); | |||
| if (GoogleVertexAiUtils.supportsModelGarden(out.getTransportVersion())) { | |||
| out.writeOptionalInt(maxTokens); | |||
Contributor
There was a problem hiding this comment.
Let's use writeOptionalVInt
Contributor
Author
There was a problem hiding this comment.
Done.
DonalEvans
reviewed
Sep 26, 2025
Comment on lines
189
to
204
| delta = new StreamingUnifiedChatCompletionResults.ChatCompletionChunk.Choice.Delta( | ||
| null, | ||
| null, | ||
| null, | ||
| List.of( | ||
| new StreamingUnifiedChatCompletionResults.ChatCompletionChunk.Choice.Delta.ToolCall( | ||
| 0, | ||
| id, | ||
| new StreamingUnifiedChatCompletionResults.ChatCompletionChunk.Choice.Delta.ToolCall.Function( | ||
| input != null ? input.toString() : null, | ||
| name | ||
| ), | ||
| null | ||
| ) | ||
| ) | ||
| ); |
Contributor
There was a problem hiding this comment.
For readability, this might be better as:
var function = new StreamingUnifiedChatCompletionResults.ChatCompletionChunk.Choice.Delta.ToolCall.Function(
input != null ? input.toString() : null,
name
);
var toolCall = new StreamingUnifiedChatCompletionResults.ChatCompletionChunk.Choice.Delta.ToolCall(0, id, function, null);
delta = new StreamingUnifiedChatCompletionResults.ChatCompletionChunk.Choice.Delta(null, null, null, List.of(toolCall));
Similar changes can be made in the parseContentBlockDelta() method.
Contributor
Author
There was a problem hiding this comment.
Thanks. Done!
…ntegration # Conflicts: # server/src/main/resources/transport/upper_bounds/9.2.csv
…CompletionStreamingProcessor readability
Jan-Kazlouski-elastic
requested review from
DonalEvans and
jonathan-buttner
…ntegration # Conflicts: # server/src/main/resources/transport/upper_bounds/9.2.csv
Contributor
Author
|
Your comments are addressed. Could you please review the fixes? |
jonathan-buttner
approved these changes
Sep 29, 2025
Contributor
jonathan-buttner
left a comment
jonathan-buttner
left a comment
There was a problem hiding this comment.
Thanks for the changes!
Contributor
Author
Create Completion EndpointNo Provider No URLs: Google Provider With URLs: Google Provider No URLs: No URLs: Both URLs: Only Non-Streaming URL: Only Streaming URL: No Task Parameters: Not Found: Perform Non-Streaming CompletionNon-Streaming Both URLs Non-Streaming Only Non-Streaming URL Non-Streaming Only Streaming URL Non-Streaming Without Task Settings Perform Streaming CompletionStreaming Both URLs Streaming Only Non-Streaming URL Streaming Only Streaming URL Streaming Without Task Settings |
Contributor
Author
Create Chat Completion EndpointNo Provider No URLs: Google Provider With URLs: Google Provider No URLs: No URLs: Both URLs: Only Non-Streaming URL: Only Streaming URL: No Task Parameters: Not Found: Testing of Performing Streaming Chat Completion is done and it is confirmed to be successful. |
Contributor
Author
Perform Chat CompletionBoth URLs Both URLs With Max Tokens in RQ Only Non-Streaming URL Only Non-Streaming URL With Max Tokens in RQ Only Streaming URL Only Streaming URL With Max Tokens in RQ Both URLs No task settings on creation Both URLs No task settings on creation With Max Tokens in RQ |
Contributor
Author
|
Regression Tests for Google Vertex AI. Create Completion endpointSuccess No model_id Perform Non-Streaming CompletionPerform Streaming CompletionCreate Chat Completion endpointPerform Chat Completion |
Contributor
Author
|
@jonathan-buttner |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Update of the existing Google Vertex AI inference provider integration allowing performing completion (both streaming and non-streaming) and chat_completion (only streaming) of Anthropic provider models withing Google Model Garden.
Changes were tested locally against next anthropic models:
Create Completion Endpoint
Success:
With max_tokens in task settings:
Unknown Provider:
No Provider + No Google Vertex AI parameters:
No URL + No Streaming URL + No Google Vertex AI parameters:
URL + No Streaming URL (URL is default for both streaming/non-streaming):
No URL + Streaming URL (Streaming URL is default for both streaming/non-streaming):
Not Found:
Perform Completion
Success Non Streaming:
Success Streaming:
Success Non Streaming with task_settings max_tokens:
Success Streaming with task_settings max_tokens:
Create Chat Completion Endpoint
Success:
Success with task_settings max_tokens:
Unknown Provider:
No url/streaming_url:
Not found:
No streaming_url (url is default for both streaming/non-streaming):
No url (steraming_url is default for both streaming/non-streaming):
Perform Chat Completion
Basic:
Complex