Package sk.iway.iwcm.rag.search

Class MarkdownSearchService

java.lang.Object

sk.iway.iwcm.rag.search.MarkdownSearchService

@Service
public class MarkdownSearchService
extends Object

Searches configured Markdown documentation using the shared semantic, hybrid and RAG pipelines.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static final record	MarkdownSearchService.SearchResponse	Carries ranked documentation sources and an optional generated answer.
static final record	MarkdownSearchService.SearchResult	Represents one documentation source returned to a search client.

Constructor Summary

Constructors

Constructor	Description
MarkdownSearchService(MarkdownSourceService sourceService, SemanticSearchService semanticSearchService, EmbeddingChunkRepository chunkRepository, RagService ragService)

Method Summary

Modifier and Type	Method	Description
MarkdownSearchService.SearchResponse	search(String query, String language, String directory, Integer domainId, jakarta.servlet.http.HttpServletRequest request)	Searches accessible globally configured roots in the selected language in the shared Markdown domain zero.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

MarkdownSearchService

public MarkdownSearchService(MarkdownSourceService sourceService,
 SemanticSearchService semanticSearchService,
 EmbeddingChunkRepository chunkRepository,
 RagService ragService)

Method Details

search

public MarkdownSearchService.SearchResponse search(String query,
 String language,
 String directory,
 Integer domainId,
 jakarta.servlet.http.HttpServletRequest request)

Searches accessible globally configured roots in the selected language in the shared Markdown domain zero. Search availability is checked before validation, source access and cache lookup. Access checks run before embeddings and again for individual sources before answer generation. Embeddings and answers use the requesting domain's assistants, credentials and usage accounting. Answers are generated automatically when ragAnswerAllowed is enabled and authorized context is available. When ragMarkdownSearchRequireLogin is enabled, any authenticated user can search; administrator access is not required. The shared search spam protection limits requests before any AI provider call. Retrieval and answers are cached for 30 minutes using a lowercase, accent-free question key. Cached chunks are authorized again and the current answer setting is checked on every request. Missing cached chunk IDs trigger fresh retrieval after reindexing or deletion.

Parameters:

query - question or search terms

language - required documentation language

directory - optional viewer directory including its language and leading/trailing slashes; descendants are always included

domainId - current request domain used for assistants and usage, not Markdown storage

request - current HTTP request used for access checks and AI providers

Returns:

unique source results with optional plain-text and safe HTML answers