tessl/maven-org-springframework-ai--spring-ai-model

Core model interfaces and abstractions for Spring AI framework providing portable API for chat, embeddings, images, audio, and tool calling across multiple AI providers

Overview

Eval results

Files

Document Enrichment and Transformation

Name: tessl/maven-org-springframework-ai--spring-ai-model
Author: tessl

AI-powered document enrichment utilities for extracting keywords and generating summaries that are added as metadata to documents. These transformers enhance documents with AI-generated metadata for improved retrieval and understanding.

Capabilities

Keyword Metadata Enricher

Extracts keywords from documents using AI and adds them as metadata.

public class KeywordMetadataEnricher implements DocumentTransformer {
    /**
     * Constant for the metadata key where keywords are stored.
     */
    public static final String EXCERPT_KEYWORDS_METADATA_KEY = "excerpt_keywords";

    /**
     * Placeholder for context in keyword templates.
     */
    public static final String CONTEXT_STR_PLACEHOLDER = "context_str";

    /**
     * Default template for keyword extraction.
     */
    public static final String KEYWORDS_TEMPLATE = "{context_str}. Give %s unique keywords for this document. Format as comma separated. Keywords: ";

    /**
     * Construct a KeywordMetadataEnricher with keyword count.
     *
     * @param chatModel the chat model to use for keyword extraction
     * @param keywordCount the number of keywords to extract
     */
    public KeywordMetadataEnricher(ChatModel chatModel, int keywordCount);

    /**
     * Construct a KeywordMetadataEnricher with custom template.
     *
     * @param chatModel the chat model to use for keyword extraction
     * @param keywordsTemplate the custom prompt template for extraction
     */
    public KeywordMetadataEnricher(ChatModel chatModel, PromptTemplate keywordsTemplate);

    /**
     * Apply keyword extraction to documents.
     *
     * @param documents the documents to enrich
     * @return the enriched documents with keywords in metadata
     */
    @Override
    List<Document> apply(List<Document> documents);

    /**
     * Create a builder for KeywordMetadataEnricher.
     *
     * @param chatModel the chat model to use
     * @return a new builder
     */
    public static Builder builder(ChatModel chatModel);
}

KeywordMetadataEnricher.Builder

public static final class Builder {
    /**
     * Construct a builder with chat model.
     *
     * @param chatModel the chat model
     */
    public Builder(ChatModel chatModel);

    /**
     * Set the number of keywords to extract.
     *
     * @param keywordCount the number of keywords (must be >= 1)
     * @return this builder
     */
    public Builder keywordCount(int keywordCount);

    /**
     * Set a custom prompt template for keyword extraction.
     *
     * @param keywordsTemplate the custom template
     * @return this builder
     */
    public Builder keywordsTemplate(PromptTemplate keywordsTemplate);

    /**
     * Build the KeywordMetadataEnricher.
     *
     * @return the configured enricher
     */
    public KeywordMetadataEnricher build();
}

Summary Metadata Enricher

Generates summaries for documents and adds them as metadata, optionally including summaries of adjacent documents.

public class SummaryMetadataEnricher implements DocumentTransformer {
    /**
     * Default template for summary extraction.
     */
    public static final String DEFAULT_SUMMARY_EXTRACT_TEMPLATE = """
        Here is the content of the section:
        {context_str}

        Summarize the key topics and entities of the section.

        Summary:""";

    /**
     * Construct a SummaryMetadataEnricher with summary types.
     *
     * @param chatModel the chat model to use for summarization
     * @param summaryTypes the types of summaries to generate (CURRENT, PREVIOUS, NEXT)
     */
    public SummaryMetadataEnricher(ChatModel chatModel, List<SummaryType> summaryTypes);

    /**
     * Construct a SummaryMetadataEnricher with full configuration.
     *
     * @param chatModel the chat model to use for summarization
     * @param summaryTypes the types of summaries to generate
     * @param summaryTemplate the template for summary generation
     * @param metadataMode the metadata mode for document formatting
     */
    public SummaryMetadataEnricher(
        ChatModel chatModel,
        List<SummaryType> summaryTypes,
        String summaryTemplate,
        MetadataMode metadataMode
    );

    /**
     * Apply summary generation to documents.
     *
     * @param documents the documents to enrich
     * @return the enriched documents with summaries in metadata
     */
    @Override
    List<Document> apply(List<Document> documents);
}

SummaryType Enum

public enum SummaryType {
    /**
     * Include summary of the previous document.
     * Stored in metadata as "prev_section_summary".
     */
    PREVIOUS,

    /**
     * Include summary of the current document.
     * Stored in metadata as "section_summary".
     */
    CURRENT,

    /**
     * Include summary of the next document.
     * Stored in metadata as "next_section_summary".
     */
    NEXT
}

MetadataMode Enum

Controls how metadata is included when formatting document content.

public enum MetadataMode {
    /**
     * Include all metadata fields when formatting document content.
     */
    ALL,

    /**
     * Include only metadata relevant for embedding operations.
     */
    EMBED,

    /**
     * Include only metadata relevant for inference operations.
     */
    INFERENCE,

    /**
     * Do not include any metadata in formatted content.
     */
    NONE
}

Usage Examples

Basic Keyword Extraction

import org.springframework.ai.model.transformer.KeywordMetadataEnricher;
import org.springframework.ai.chat.model.ChatModel;
import org.springframework.ai.document.Document;

@Service
public class DocumentEnrichmentService {
    private final ChatModel chatModel;

    public List<Document> enrichWithKeywords(List<Document> documents) {
        // Extract 5 keywords from each document
        KeywordMetadataEnricher enricher = new KeywordMetadataEnricher(chatModel, 5);

        // Apply enrichment
        List<Document> enriched = enricher.apply(documents);

        // Access keywords
        for (Document doc : enriched) {
            String keywords = (String) doc.getMetadata()
                .get(KeywordMetadataEnricher.EXCERPT_KEYWORDS_METADATA_KEY);
            System.out.println("Keywords: " + keywords);
        }

        return enriched;
    }
}

Keyword Extraction with Builder

import org.springframework.ai.model.transformer.KeywordMetadataEnricher;

// Using builder for configuration
KeywordMetadataEnricher enricher = KeywordMetadataEnricher.builder(chatModel)
    .keywordCount(10)
    .build();

List<Document> enriched = enricher.apply(documents);

Custom Keyword Template

import org.springframework.ai.chat.prompt.PromptTemplate;

// Create custom template
PromptTemplate customTemplate = new PromptTemplate("""
    Text: {context_str}

    Extract the most important technical keywords and concepts.
    Return as a comma-separated list.
    Keywords:
    """);

// Use custom template
KeywordMetadataEnricher enricher = KeywordMetadataEnricher.builder(chatModel)
    .keywordsTemplate(customTemplate)
    .build();

List<Document> enriched = enricher.apply(documents);

Basic Summary Generation

import org.springframework.ai.model.transformer.SummaryMetadataEnricher;
import org.springframework.ai.model.transformer.SummaryMetadataEnricher.SummaryType;

// Generate only current document summaries
SummaryMetadataEnricher enricher = new SummaryMetadataEnricher(
    chatModel,
    List.of(SummaryType.CURRENT)
);

List<Document> enriched = enricher.apply(documents);

// Access summaries
for (Document doc : enriched) {
    String summary = (String) doc.getMetadata().get("section_summary");
    System.out.println("Summary: " + summary);
}

Summary with Adjacent Context

import org.springframework.ai.document.MetadataMode;

// Include summaries of previous, current, and next documents
SummaryMetadataEnricher enricher = new SummaryMetadataEnricher(
    chatModel,
    List.of(SummaryType.PREVIOUS, SummaryType.CURRENT, SummaryType.NEXT)
);

List<Document> enriched = enricher.apply(documents);

// Access all summary types
for (Document doc : enriched) {
    Map<String, Object> metadata = doc.getMetadata();

    String prevSummary = (String) metadata.get("prev_section_summary");
    String currentSummary = (String) metadata.get("section_summary");
    String nextSummary = (String) metadata.get("next_section_summary");

    System.out.println("Previous: " + prevSummary);
    System.out.println("Current: " + currentSummary);
    System.out.println("Next: " + nextSummary);
}

Custom Summary Template

String customTemplate = """
    Document content:
    {context_str}

    Provide a concise one-sentence summary highlighting the main topic.

    Summary:
    """;

SummaryMetadataEnricher enricher = new SummaryMetadataEnricher(
    chatModel,
    List.of(SummaryType.CURRENT),
    customTemplate,
    MetadataMode.ALL
);

Combined Enrichment Pipeline

@Service
public class DocumentProcessingPipeline {
    private final ChatModel chatModel;

    public List<Document> processDocuments(List<Document> documents) {
        // Step 1: Extract keywords
        KeywordMetadataEnricher keywordEnricher =
            KeywordMetadataEnricher.builder(chatModel)
                .keywordCount(5)
                .build();

        List<Document> withKeywords = keywordEnricher.apply(documents);

        // Step 2: Generate summaries
        SummaryMetadataEnricher summaryEnricher = new SummaryMetadataEnricher(
            chatModel,
            List.of(SummaryType.CURRENT)
        );

        List<Document> fullyEnriched = summaryEnricher.apply(withKeywords);

        // Now documents have both keywords and summaries
        return fullyEnriched;
    }
}

Using Enriched Metadata for Search

public class EnrichedDocumentSearch {
    public void demonstrateSearch(List<Document> enrichedDocs, String query) {
        // Filter documents by keywords
        List<Document> relevantDocs = enrichedDocs.stream()
            .filter(doc -> {
                String keywords = (String) doc.getMetadata()
                    .get(KeywordMetadataEnricher.EXCERPT_KEYWORDS_METADATA_KEY);
                return keywords != null && keywords.toLowerCase().contains(query.toLowerCase());
            })
            .toList();

        // Display summaries of relevant documents
        for (Document doc : relevantDocs) {
            String summary = (String) doc.getMetadata().get("section_summary");
            System.out.println("Relevant doc summary: " + summary);
        }
    }
}

Metadata Mode Configuration

import org.springframework.ai.document.MetadataMode;

// Include all metadata in summary generation
SummaryMetadataEnricher allMetadata = new SummaryMetadataEnricher(
    chatModel,
    List.of(SummaryType.CURRENT),
    SummaryMetadataEnricher.DEFAULT_SUMMARY_EXTRACT_TEMPLATE,
    MetadataMode.ALL  // Include existing metadata when summarizing
);

// Or exclude metadata from summarization
SummaryMetadataEnricher noMetadata = new SummaryMetadataEnricher(
    chatModel,
    List.of(SummaryType.CURRENT),
    SummaryMetadataEnricher.DEFAULT_SUMMARY_EXTRACT_TEMPLATE,
    MetadataMode.NONE  // Only use document content
);

Spring Configuration

@Configuration
public class EnrichmentConfig {

    @Bean
    public KeywordMetadataEnricher keywordEnricher(ChatModel chatModel) {
        return KeywordMetadataEnricher.builder(chatModel)
            .keywordCount(7)
            .build();
    }

    @Bean
    public SummaryMetadataEnricher summaryEnricher(ChatModel chatModel) {
        return new SummaryMetadataEnricher(
            chatModel,
            List.of(SummaryType.PREVIOUS, SummaryType.CURRENT, SummaryType.NEXT)
        );
    }
}

Install with Tessl CLI

npx tessl i tessl/maven-org-springframework-ai--spring-ai-model@1.1.1