[EIS] Dense Text Embedding task type integration #129847
There are no files selected for viewing
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
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
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
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
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
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,103 @@ | ||
| /* | ||
| * Copyright Elasticsearch B.V. and/or licensed to Elasticsearch B.V. under one | ||
| * or more contributor license agreements. Licensed under the Elastic License | ||
| * 2.0; you may not use this file except in compliance with the Elastic License | ||
| * 2.0. | ||
| */ | ||
|
|
||
| package org.elasticsearch.xpack.inference.external.response.elastic; | ||
|
|
||
| import org.elasticsearch.common.xcontent.XContentParserUtils; | ||
| import org.elasticsearch.xcontent.ConstructingObjectParser; | ||
| import org.elasticsearch.xcontent.ParseField; | ||
| import org.elasticsearch.xcontent.XContentFactory; | ||
| import org.elasticsearch.xcontent.XContentParserConfiguration; | ||
| import org.elasticsearch.xcontent.XContentType; | ||
| import org.elasticsearch.xpack.core.inference.results.TextEmbeddingFloatResults; | ||
| import org.elasticsearch.xpack.inference.external.http.HttpResult; | ||
| import org.elasticsearch.xpack.inference.external.request.Request; | ||
|
|
||
| import java.io.IOException; | ||
| import java.util.List; | ||
|
|
||
| import static org.elasticsearch.xcontent.ConstructingObjectParser.constructorArg; | ||
|
|
||
| public class ElasticInferenceServiceDenseTextEmbeddingsResponseEntity { | ||
|
|
||
| /** | ||
| * Parses the Elastic Inference Service Dense Text Embeddings response. | ||
| * | ||
| * For a request like: | ||
| * | ||
| * <pre> | ||
| * <code> | ||
| * { | ||
| * "inputs": ["Embed this text", "Embed this text, too"] | ||
| * } | ||
| * </code> | ||
| * </pre> | ||
| * | ||
| * The response would look like: | ||
| * | ||
| * <pre> | ||
| * <code> | ||
| * { | ||
| * "data": [ | ||
| * [ | ||
| * 2.1259406, | ||
| * 1.7073475, | ||
| * 0.9020516 | ||
| * ], | ||
| * (...) | ||
| * ], | ||
|
There was a problem hiding this comment. I vaguely remembered Tim's thread on this a couple weeks ago, but should we revisit the response format? Looking at OpenAI, Alibaba, and Mixedbread as quick references, it looks like they return a list of objects. I don't have a strong preference, but just wanted to bring this up since we might be differing from others here and wanted to confirm that this is what we want. There was a problem hiding this comment. Answered in the thread |
||
| * "meta": { | ||
| * "usage": {...} | ||
| * } | ||
| * } | ||
| * </code> | ||
| * </pre> | ||
| */ | ||
| public static TextEmbeddingFloatResults fromResponse(Request request, HttpResult response) throws IOException { | ||
| try (var p = XContentFactory.xContent(XContentType.JSON).createParser(XContentParserConfiguration.EMPTY, response.body())) { | ||
| return EmbeddingFloatResult.PARSER.apply(p, null).toTextEmbeddingFloatResults(); | ||
| } | ||
| } | ||
|
|
||
| public record EmbeddingFloatResult(List<EmbeddingFloatResultEntry> embeddingResults) { | ||
| @SuppressWarnings("unchecked") | ||
| public static final ConstructingObjectParser<EmbeddingFloatResult, Void> PARSER = new ConstructingObjectParser<>( | ||
| EmbeddingFloatResult.class.getSimpleName(), | ||
| true, | ||
| args -> new EmbeddingFloatResult((List<EmbeddingFloatResultEntry>) args[0]) | ||
| ); | ||
|
|
||
| static { | ||
| // Custom field declaration to handle array of arrays format | ||
| PARSER.declareField(constructorArg(), (parser, context) -> { | ||
| return XContentParserUtils.parseList(parser, (p, index) -> { | ||
| List<Float> embedding = XContentParserUtils.parseList(p, (innerParser, innerIndex) -> innerParser.floatValue()); | ||
| return EmbeddingFloatResultEntry.fromFloatArray(embedding); | ||
| }); | ||
| }, new ParseField("data"), org.elasticsearch.xcontent.ObjectParser.ValueType.OBJECT_ARRAY); | ||
| } | ||
|
|
||
| public TextEmbeddingFloatResults toTextEmbeddingFloatResults() { | ||
| return new TextEmbeddingFloatResults( | ||
| embeddingResults.stream().map(entry -> TextEmbeddingFloatResults.Embedding.of(entry.embedding)).toList() | ||
| ); | ||
| } | ||
| } | ||
|
|
||
| /** | ||
| * Represents a single embedding entry in the response. | ||
| * For the Elastic Inference Service, each entry is just an array of floats (no wrapper object). | ||
| * This is a simpler wrapper that just holds the float array. | ||
| */ | ||
| public record EmbeddingFloatResultEntry(List<Float> embedding) { | ||
| public static EmbeddingFloatResultEntry fromFloatArray(List<Float> floats) { | ||
| return new EmbeddingFloatResultEntry(floats); | ||
| } | ||
| } | ||
|
|
||
| private ElasticInferenceServiceDenseTextEmbeddingsResponseEntity() {} | ||
| } | ||
Oops, something went wrong.
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.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Not a blocker, but can you explain why the
MinimalServiceSettingsdiffer from other task types?There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think it's just about the different purposes models/tasks:
ElementType. Some models also allow you to specify a target number of dimensions (f.e. when using Matryoshka embeddings, therefore we need to specify the number of dimensions. Also vector embeddings can be compared using different similarity measures, therefore we need to specify the similarity measure.There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Makes sense! Thanks for the background :)