fix(llm): address PR review feedback

- Bedrock extractText concatenates all text blocks (avoid truncating multi-block Converse output)
- drop unused applicationConfig param from getEmbeddingConfigurationMessage
- regenerate TS types for the llmConfiguration.embeddings / naturalLanguageSearch schema change

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: pmbrull

openmetadata-service/src/main/java/org/openmetadata/service/jdbi3/SystemRepository.java

@@ -730,7 +730,7 @@ StepValidation getEmbeddingsValidation(OpenMetadataApplicationConfig application
           .withPassed(false);
     }
 
-    String configMessage = getEmbeddingConfigurationMessage(applicationConfig);
+    String configMessage = getEmbeddingConfigurationMessage();
 
     if (searchRepository.getVectorIndexService() == null) {
       return retryInitAndReportError(
@@ -884,7 +884,7 @@ private StepValidation validateEmbeddingGeneration(
         .withPassed(true);
   }
 
-  private String getEmbeddingConfigurationMessage(OpenMetadataApplicationConfig applicationConfig) {
+  private String getEmbeddingConfigurationMessage() {
     try {
       LLMConfiguration llmConfig = LlmConfigHolder.get();
       LLMEmbeddingsConfig embeddings = llmConfig != null ? llmConfig.getEmbeddings() : null;

openmetadata-service/src/main/java/org/openmetadata/service/llm/BedrockCompletionClient.java

@@ -118,16 +118,15 @@ static CompletionResult toResult(ConverseResponse response) {
   }
 
   private static String extractText(ConverseResponse response) {
-    String text = "";
+    StringBuilder text = new StringBuilder();
     if (response.output() != null && response.output().message() != null) {
       for (ContentBlock block : response.output().message().content()) {
         if (block.text() != null) {
-          text = block.text();
-          break;
+          text.append(block.text());
         }
       }
     }
-    return text;
+    return text.toString();
   }
 
   @Override

openmetadata-ui/src/main/resources/ui/src/generated/configuration/elasticSearchConfiguration.ts

@@ -118,18 +118,6 @@ export interface Aws {
  * Configuration for natural language search capabilities
  */
 export interface NaturalLanguageSearch {
-    /**
-     * AWS Bedrock configuration for natural language processing
-     */
-    bedrock?: Bedrock;
-    /**
-     * Embedding generation using Deep Java Library (DJL)
-     */
-    djl?: Djl;
-    /**
-     * The provider to use for generating vector embeddings (e.g., bedrock, openai, google, djl).
-     */
-    embeddingProvider?: string;
     /**
      * Enable or disable natural language search
      */
@@ -138,10 +126,6 @@ export interface NaturalLanguageSearch {
      * NLQ filter extractor cache and prompt tuning.
      */
     filterExtractor?: FilterExtractor;
-    /**
-     * Google Gemini configuration for embedding generation via the Generative Language API.
-     */
-    google?: Google;
     /**
      * Hybrid search runtime tuning combining BM25 keyword and KNN semantic queries.
      */
@@ -150,16 +134,6 @@ export interface NaturalLanguageSearch {
      * Weight for BM25 keyword search results in hybrid RRF pipeline (0.0-1.0)
      */
     keywordWeight?: number;
-    /**
-     * Maximum number of concurrent embedding and NLQ provider requests. Controls the semaphore
-     * used to throttle calls to the providers and prevent overwhelming HTTP/2 connection limits.
-     */
-    maxConcurrentRequests?: number;
-    /**
-     * OpenAI configuration for embedding generation. Supports both OpenAI and Azure OpenAI
-     * endpoints.
-     */
-    openai?: Openai;
     /**
      * Fully qualified class name of the NLQService implementation to use
      */
@@ -174,93 +148,6 @@ export interface NaturalLanguageSearch {
     semanticWeight?: number;
 }
 
-/**
- * AWS Bedrock configuration for natural language processing
- */
-export interface Bedrock {
-    /**
-     * AWS credentials configuration for Bedrock service
-     */
-    awsConfig?: AWSBaseConfig;
-    /**
-     * Dimension of the embedding vector
-     */
-    embeddingDimension?: number;
-    /**
-     * Bedrock embedding model identifier to use for vector search
-     */
-    embeddingModelId?: string;
-    /**
-     * Maximum tokens the Bedrock model is allowed to generate.
-     */
-    maxTokens?: number;
-    /**
-     * Bedrock model identifier to use for query transformation
-     */
-    modelId?: string;
-    /**
-     * Sampling temperature for Bedrock requests.
-     */
-    temperature?: number;
-    /**
-     * Bedrock InvokeModel API call timeout in seconds.
-     */
-    timeoutSeconds?: number;
-}
-
-/**
- * AWS credentials configuration for Bedrock service
- *
- * Base AWS configuration for authentication. Supports static credentials, IAM roles, and
- * default credential provider chain.
- */
-export interface AWSBaseConfig {
-    /**
-     * AWS Access Key ID. Falls back to default credential provider chain if not set.
-     */
-    accessKeyId?: string;
-    /**
-     * ARN of IAM role to assume for cross-account access.
-     */
-    assumeRoleArn?: string;
-    /**
-     * Session name for assumed role.
-     */
-    assumeRoleSessionName?: string;
-    /**
-     * Enable AWS IAM authentication. When enabled, uses the default credential provider chain
-     * (environment variables, instance profile, etc.). Defaults to false for backward
-     * compatibility.
-     */
-    enabled?: boolean;
-    /**
-     * Custom endpoint URL for AWS-compatible services (MinIO, LocalStack).
-     */
-    endpointUrl?: string;
-    /**
-     * AWS Region (e.g., us-east-1). Required when AWS authentication is enabled.
-     */
-    region?: string;
-    /**
-     * AWS Secret Access Key. Falls back to default credential provider chain if not set.
-     */
-    secretAccessKey?: string;
-    /**
-     * AWS Session Token for temporary credentials.
-     */
-    sessionToken?: string;
-}
-
-/**
- * Embedding generation using Deep Java Library (DJL)
- */
-export interface Djl {
-    /**
-     * DJL model name for embedding generation
-     */
-    embeddingModel?: string;
-}
-
 /**
  * NLQ filter extractor cache and prompt tuning.
  */
@@ -277,40 +164,23 @@ export interface FilterExtractor {
      * Max sample values shown per filter category in the system prompt.
      */
     maxSampleValues?: number;
-}
-
-/**
- * Google Gemini configuration for embedding generation via the Generative Language API.
- */
-export interface Google {
-    /**
-     * API key from Google AI Studio for authenticating with the Generative Language API.
-     */
-    apiKey?: string;
     /**
-     * Dimension of the embedding vector, sent to Google as `outputDimensionality`. For
-     * `gemini-embedding-001` valid values are 768, 1536, or 3072. For `text-embedding-004` use
-     * 768.
+     * Maximum tokens the model may generate for NLQ filter extraction.
      */
-    embeddingDimension?: number;
+    maxTokens?: number;
     /**
-     * Gemini embedding model identifier (e.g., gemini-embedding-001, text-embedding-004).
+     * Optional model override for NLQ filter extraction. Leave empty to use the model from
+     * llmConfiguration.
      */
-    embeddingModelId?: string;
+    modelId?: string;
     /**
-     * Optional override for the full embedding endpoint URL. Must be the complete URL including
-     * the model and `:embedContent` action (e.g.
-     * `https://generativelanguage.googleapis.com/v1beta/models/text-embedding-004:embedContent`),
-     * not just a base URL. Leave empty to use the default Generative Language API endpoint,
-     * which is constructed from `embeddingModelId`. The `key` query parameter is appended
-     * automatically.
+     * Sampling temperature for NLQ filter extraction.
      */
-    endpoint?: string;
+    temperature?: number;
     /**
-     * Gemini chat model identifier for query transformation (e.g., gemini-2.5-flash,
-     * gemini-1.5-flash).
+     * Per-call timeout in seconds for NLQ filter extraction completion.
      */
-    modelId?: string;
+    timeoutSeconds?: number;
 }
 
 /**
@@ -339,54 +209,6 @@ export interface HybridSearch {
     semanticScoreThreshold?: number;
 }
 
-/**
- * OpenAI configuration for embedding generation. Supports both OpenAI and Azure OpenAI
- * endpoints.
- */
-export interface Openai {
-    /**
-     * API key for authenticating with OpenAI or Azure OpenAI.
-     */
-    apiKey?: string;
-    /**
-     * Azure OpenAI API version. Only used with Azure OpenAI.
-     */
-    apiVersion?: string;
-    /**
-     * Azure OpenAI deployment name. Required when using Azure OpenAI.
-     */
-    deploymentName?: string;
-    /**
-     * Dimension of the embedding vector. Default is 1536 for text-embedding-3-small.
-     */
-    embeddingDimension?: number;
-    /**
-     * OpenAI embedding model identifier (e.g., text-embedding-3-small, text-embedding-ada-002).
-     */
-    embeddingModelId?: string;
-    /**
-     * Custom endpoint URL. For Azure OpenAI, use the Azure resource endpoint (e.g.,
-     * https://your-resource.openai.azure.com). Leave empty for standard OpenAI API.
-     */
-    endpoint?: string;
-    /**
-     * Maximum tokens the OpenAI model is allowed to generate.
-     */
-    maxTokens?: number;
-    /**
-     * OpenAI model identifier to use for query transformation (chat completions).
-     */
-    modelId?: string;
-    /**
-     * Sampling temperature for OpenAI requests.
-     */
-    temperature?: number;
-    /**
-     * OpenAI HTTP request and connect timeout in seconds.
-     */
-    timeoutSeconds?: number;
-}
-
 /**
  * This schema defines the language options available for search index mappings.
  */

openmetadata-ui/src/main/resources/ui/src/generated/configuration/llmConfiguration.ts

@@ -16,17 +16,24 @@
  */
 export interface LlmConfiguration {
     anthropic?: Anthropic;
-    bedrock?:   Bedrock;
+    bedrock?:   LlmConfigurationBedrock;
+    /**
+     * Vector embedding configuration for semantic search. Credentials are reused from the
+     * sibling provider blocks (bedrock.awsConfig, openai.apiKey/endpoint,
+     * google.apiKey/endpoint); this section selects the embedding provider and per-provider
+     * embedding model. The embedding provider may differ from the chat completion provider.
+     */
+    embeddings?: Embeddings;
     /**
      * Whether LLM completion features are enabled.
      */
     enabled?: boolean;
-    google?:  Google;
+    google?:  LlmConfigurationGoogle;
     /**
      * Maximum number of concurrent LLM completion requests.
      */
     maxConcurrentRequests?: number;
-    openai?:                Openai;
+    openai?:                LlmConfigurationOpenai;
     provider?:              LlmProvider;
 }
 
@@ -39,7 +46,7 @@ export interface Anthropic {
     timeoutSeconds?: number;
 }
 
-export interface Bedrock {
+export interface LlmConfigurationBedrock {
     /**
      * AWS credentials configuration for Bedrock service.
      */
@@ -93,7 +100,57 @@ export interface AWSBaseConfig {
     sessionToken?: string;
 }
 
-export interface Google {
+/**
+ * Vector embedding configuration for semantic search. Credentials are reused from the
+ * sibling provider blocks (bedrock.awsConfig, openai.apiKey/endpoint,
+ * google.apiKey/endpoint); this section selects the embedding provider and per-provider
+ * embedding model. The embedding provider may differ from the chat completion provider.
+ */
+export interface Embeddings {
+    bedrock?: EmbeddingsBedrock;
+    djl?:     Djl;
+    google?:  EmbeddingsGoogle;
+    /**
+     * Maximum number of concurrent embedding requests.
+     */
+    maxConcurrentRequests?: number;
+    openai?:                EmbeddingsOpenai;
+    /**
+     * Embedding provider to use for semantic search vectors.
+     */
+    provider?: Provider;
+}
+
+export interface EmbeddingsBedrock {
+    embeddingDimension?: number;
+    embeddingModelId?:   string;
+}
+
+export interface Djl {
+    embeddingModel?: string;
+}
+
+export interface EmbeddingsGoogle {
+    embeddingDimension?: number;
+    embeddingModelId?:   string;
+}
+
+export interface EmbeddingsOpenai {
+    embeddingDimension?: number;
+    embeddingModelId?:   string;
+}
+
+/**
+ * Embedding provider to use for semantic search vectors.
+ */
+export enum Provider {
+    Bedrock = "bedrock",
+    Djl = "djl",
+    Google = "google",
+    Openai = "openai",
+}
+
+export interface LlmConfigurationGoogle {
     apiKey?:         string;
     endpoint?:       string;
     maxTokens?:      number;
@@ -102,7 +159,7 @@ export interface Google {
     timeoutSeconds?: number;
 }
 
-export interface Openai {
+export interface LlmConfigurationOpenai {
     apiKey?:         string;
     apiVersion?:     string;
     deploymentName?: string;