PR Feedback

Author: kderusso

x-pack/plugin/inference/qa/test-service-plugin/src/main/java/org/elasticsearch/xpack/inference/mock/TestDenseInferenceServiceExtension.java

@@ -180,16 +180,14 @@ private List<ChunkedInference> makeChunkedResults(List<ChunkInferenceInput> inpu
             var results = new ArrayList<ChunkedInference>();
             for (ChunkInferenceInput input : inputs) {
                 List<ChunkedInput> chunkedInput = chunkInputs(input);
-                List<TextEmbeddingFloatResults.Chunk> chunks = new ArrayList<>(
-                    chunkedInput.stream()
-                        .map(
-                            c -> new TextEmbeddingFloatResults.Chunk(
-                                makeResults(List.of(c.input()), serviceSettings).embeddings().get(0),
-                                new ChunkedInference.TextOffset(c.startOffset(), c.endOffset())
-                            )
+                List<TextEmbeddingFloatResults.Chunk> chunks = chunkedInput.stream()
+                    .map(
+                        c -> new TextEmbeddingFloatResults.Chunk(
+                            makeResults(List.of(c.input()), serviceSettings).embeddings().get(0),
+                            new ChunkedInference.TextOffset(c.startOffset(), c.endOffset())
                         )
-                        .toList()
-                );
+                    )
+                    .toList();
                 ChunkedInferenceEmbedding chunkedInferenceEmbedding = new ChunkedInferenceEmbedding(chunks);
                 results.add(chunkedInferenceEmbedding);
             }

x-pack/plugin/inference/qa/test-service-plugin/src/main/java/org/elasticsearch/xpack/inference/mock/TestSparseInferenceServiceExtension.java

@@ -170,16 +170,15 @@ private SparseEmbeddingResults makeResults(List<String> input) {
         private List<ChunkedInference> makeChunkedResults(List<ChunkInferenceInput> inputs) {
             List<ChunkedInference> results = new ArrayList<>();
             for (ChunkInferenceInput chunkInferenceInput : inputs) {
-                String input = chunkInferenceInput.input();
                 List<ChunkedInput> chunkedInput = chunkInputs(chunkInferenceInput);
-                List<SparseEmbeddingResults.Chunk> chunks = new ArrayList<>(chunkedInput.stream().map(c -> {
+                List<SparseEmbeddingResults.Chunk> chunks = chunkedInput.stream().map(c -> {
                     var tokens = new ArrayList<WeightedToken>();
                     for (int i = 0; i < 5; i++) {
                         tokens.add(new WeightedToken("feature_" + i, generateEmbedding(c.input(), i)));
                     }
                     var embeddings = new SparseEmbeddingResults.Embedding(tokens, false);
                     return new SparseEmbeddingResults.Chunk(embeddings, new ChunkedInference.TextOffset(c.startOffset(), c.endOffset()));
-                }).toList());
+                }).toList();
                 ChunkedInferenceEmbedding chunkedInferenceEmbedding = new ChunkedInferenceEmbedding(chunks);
                 results.add(chunkedInferenceEmbedding);
             }

x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/chunking/EmbeddingRequestChunker.java

@@ -48,10 +48,7 @@ public String chunkText() {
 
     public record BatchRequest(List<Request> requests) {
         public List<String> inputs() {
-            return requests.stream()
-                .map(r -> new ChunkInferenceInput(r.chunkText(), r.inputs().getFirst().chunkingSettings()))
-                .map(ChunkInferenceInput::input)
-                .collect(Collectors.toList());
+            return requests.stream().map(Request::chunkText).collect(Collectors.toList());
         }
     }
 

x-pack/plugin/inference/src/main/java/org/elasticsearch/xpack/inference/mapper/SemanticTextFieldMapper.java

@@ -317,13 +317,14 @@ private void validateServiceSettings(MinimalServiceSettings settings) {
          * @return A mapper with the copied settings applied
          */
         private SemanticTextFieldMapper copySettings(SemanticTextFieldMapper mapper, MapperMergeContext mapperMergeContext) {
-            SemanticTextFieldMapper returnedMapper;
-            Builder builder = from(mapper);
+            SemanticTextFieldMapper returnedMapper = mapper;
             if (mapper.fieldType().getModelSettings() == null) {
+                Builder builder = from(mapper);
                 builder.setModelSettings(modelSettings.getValue());
+                returnedMapper = builder.build(mapperMergeContext.getMapperBuilderContext());
             }
-            builder.setChunkingSettings(mapper.fieldType().getChunkingSettings());
-            return builder.build(mapperMergeContext.getMapperBuilderContext());
+
+            return returnedMapper;
         }
     }
 

x-pack/plugin/inference/src/test/java/org/elasticsearch/xpack/inference/external/action/SingleInputSenderExecutableActionTests.java

@@ -64,7 +64,7 @@ public void testOneInputIsValid() {
 
     public void testMoreThanOneInput() {
         var badInput = mock(EmbeddingsInput.class);
-        var input = List.of(new ChunkInferenceInput("one", null), new ChunkInferenceInput("two", null));
+        var input = List.of(new ChunkInferenceInput("one"), new ChunkInferenceInput("two"));
         when(badInput.getInputs()).thenReturn(input);
         when(badInput.inputSize()).thenReturn(input.size());
         var actualException = new AtomicReference<Exception>();

x-pack/plugin/inference/src/test/java/org/elasticsearch/xpack/inference/queries/SemanticQueryBuilderTests.java

@@ -60,7 +60,6 @@
 import org.elasticsearch.xpack.core.ml.search.WeightedToken;
 import org.elasticsearch.xpack.inference.InferencePlugin;
 import org.elasticsearch.xpack.inference.mapper.SemanticTextField;
-import org.elasticsearch.xpack.inference.mapper.SemanticTextFieldTests;
 import org.junit.Before;
 import org.junit.BeforeClass;
 
@@ -370,12 +369,7 @@ private static SourceToParse buildSemanticTextFieldWithInferenceResults(
                 useLegacyFormat,
                 SEMANTIC_TEXT_FIELD,
                 null,
-                new SemanticTextField.InferenceResult(
-                    INFERENCE_ID,
-                    modelSettings,
-                    SemanticTextFieldTests.generateRandomChunkingSettings(),
-                    Map.of(SEMANTIC_TEXT_FIELD, List.of())
-                ),
+                new SemanticTextField.InferenceResult(INFERENCE_ID, modelSettings, null, Map.of(SEMANTIC_TEXT_FIELD, List.of())),
                 XContentType.JSON
             );
 

x-pack/plugin/inference/src/yamlRestTest/resources/rest-api-spec/test/inference/25_semantic_text_field_mapping_chunking.yml

@@ -274,13 +274,14 @@ setup:
             fields:
               inference_field:
                 type: "semantic"
-                number_of_fragments: 2
+                number_of_fragments: 3
 
   - match: { hits.total.value: 1 }
   - match: { hits.hits.0._id: "doc_4" }
-  - length: { hits.hits.0.highlight.inference_field: 2 }
+  - length: { hits.hits.0.highlight.inference_field: 3 }
   - match: { hits.hits.0.highlight.inference_field.0: "Elasticsearch is an open source, distributed, RESTful, search engine which" }
   - match: { hits.hits.0.highlight.inference_field.1: " which is built on top of Lucene internally and enjoys" }
+  - match: { hits.hits.0.highlight.inference_field.2: " enjoys all the features it provides." }
 
 ---
 "We respect multiple semantic_text fields with different chunking configurations":

x-pack/plugin/inference/src/yamlRestTest/resources/rest-api-spec/test/inference/25_semantic_text_field_mapping_chunking_bwc.yml

@@ -286,13 +286,14 @@ setup:
             fields:
               inference_field:
                 type: "semantic"
-                number_of_fragments: 2
+                number_of_fragments: 3
 
   - match: { hits.total.value: 1 }
   - match: { hits.hits.0._id: "doc_4" }
-  - length: { hits.hits.0.highlight.inference_field: 2 }
+  - length: { hits.hits.0.highlight.inference_field: 3 }
   - match: { hits.hits.0.highlight.inference_field.0: "Elasticsearch is an open source, distributed, RESTful, search engine which" }
   - match: { hits.hits.0.highlight.inference_field.1: " which is built on top of Lucene internally and enjoys" }
+  - match: { hits.hits.0.highlight.inference_field.2: " enjoys all the features it provides." }
 
 ---
 "We respect multiple semantic_text fields with different chunking configurations":