package ru.mcs.metadatabook.service;
import com.fasterxml.jackson.databind.JsonNode;
import lombok.RequiredArgsConstructor;
import lombok.extern.slf4j.Slf4j;
import org.springframework.beans.factory.annotation.Value;
import org.springframework.stereotype.Service;
import org.springframework.transaction.annotation.Transactional;
import ru.mcs.metadatabook.client.FantLabClient;
import ru.mcs.metadatabook.client.GoogleBooksClient;
import ru.mcs.metadatabook.client.LibGenClient;
import ru.mcs.metadatabook.client.LlmFallbackClient;
import ru.mcs.metadatabook.client.OpenLibraryClient;
import ru.mcs.metadatabook.entity.MtData;
import ru.mcs.metadatabook.entity.UnresolvedLookup;
import ru.mcs.metadatabook.repository.UnresolvedLookupRepository;
import java.time.OffsetDateTime;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.LinkedHashMap;
import java.util.List;
import java.util.Map;
import java.util.stream.Collectors;
import java.util.stream.StreamSupport;
/**
* Обогащение метаданных из открытых источников.
*
* Два входных сценария:
* - enrichByIsbn: Open Library -> Google Books -> FantLab -> LibGen (если включён)
* -> LLM fallback (если включён) -> unresolved_lookup.
* - enrichByTitle: для изданий БЕЗ ISBN (старые советские книги и т.п.).
* Open Library (search.json) -> Google Books (intitle) -> FantLab -> LLM fallback.
* LibGen сюда не входит — его json.php API поддерживает поиск только по ISBN
* (addkeys=isbn), поиска по названию у него нет.
*
* Дедупликация записей идёт по параметрическому полю (ISBN либо Title) —
* см. persist(orgId, dedupeFieldName, dedupeValue, ...).
*/
@Service
@RequiredArgsConstructor
@Slf4j
public class PublicSourcesEnrichmentService {
private final OpenLibraryClient openLibraryClient;
private final GoogleBooksClient googleBooksClient;
private final FantLabClient fantLabClient;
private final LibGenClient libGenClient;
private final LlmFallbackClient llmFallbackClient;
private final BookQueryService bookQueryService;
private final BookWriteService bookWriteService;
private final UnresolvedLookupRepository unresolvedLookupRepository;
@Value("${llm.fallback.enabled:false}")
private boolean llmFallbackEnabled;
@Value("${llm.fallback.confidence:0.3}")
private double llmFallbackConfidence;
@Transactional
public MtData enrichByIsbn(Integer orgId, String isbn) {
List<String> attempted = new ArrayList<>();
JsonNode olBook = openLibraryClient.findByIsbn(isbn);
attempted.add("openlibrary");
if (olBook != null) {
return persist(orgId, "ISBN", isbn, mapOpenLibrary(olBook, isbn), "openlibrary", 0.7);
}
JsonNode gbVolume = googleBooksClient.findByIsbn(isbn);
attempted.add("googlebooks");
if (gbVolume != null) {
return persist(orgId, "ISBN", isbn, mapGoogleBooks(gbVolume, isbn), "googlebooks", 0.7);
}
JsonNode flEdition = fantLabClient.findByIsbn(isbn);
attempted.add("fantlab");
if (flEdition != null) {
return persist(orgId, "ISBN", isbn, mapFantLab(flEdition, isbn), "fantlab", 0.65);
}
JsonNode lgEdition = libGenClient.findByIsbn(isbn);
attempted.add("libgen");
if (lgEdition != null) {
return persist(orgId, "ISBN", isbn, mapLibGen(lgEdition, isbn), "libgen", 0.4);
}
if (llmFallbackEnabled) {
JsonNode llmResult = llmFallbackClient.findByIsbn(isbn);
attempted.add("llm_fallback");
if (llmResult != null) {
log.info("ISBN {} восстановлен через LLM fallback (низкая уверенность, рекомендуется ручная проверка)", isbn);
return persist(orgId, "ISBN", isbn, mapLlmFallback(llmResult, isbn), "llm_fallback", llmFallbackConfidence);
}
}
saveUnresolved(orgId, isbn, null, null, attempted);
log.info("ISBN {} not found in any source (attempted: {}), saved to unresolved_lookup", isbn, attempted);
return null;
}
/**
* Обогащение по названию — для изданий без ISBN (старые книги, редкие
* технические справочники и т.п.). Точность ниже, чем поиск по ISBN,
* так как совпадение названия не гарантирует то самое издание —
* особенно рискованно для книг с общими/родовыми названиями.
*/
@Transactional
public MtData enrichByTitle(Integer orgId, String title) {
List<String> attempted = new ArrayList<>();
JsonNode olDoc = openLibraryClient.findByTitle(title);
attempted.add("openlibrary");
if (olDoc != null) {
return persist(orgId, "Title", title, mapOpenLibrarySearch(olDoc), "openlibrary", 0.5);
}
JsonNode gbVolume = googleBooksClient.searchByTitle(title, 1);
attempted.add("googlebooks");
if (gbVolume != null) {
return persist(orgId, "Title", title, mapGoogleBooks(gbVolume, null), "googlebooks", 0.5);
}
JsonNode flEdition = fantLabClient.findByTitle(title);
attempted.add("fantlab");
if (flEdition != null) {
return persist(orgId, "Title", title, mapFantLab(flEdition, null), "fantlab", 0.5);
}
if (llmFallbackEnabled) {
JsonNode llmResult = llmFallbackClient.findByTitle(title);
attempted.add("llm_fallback");
if (llmResult != null) {
log.info("Книга '{}' восстановлена через LLM fallback по названию (низкая уверенность, нужна ручная проверка)", title);
return persist(orgId, "Title", title, mapLlmFallback(llmResult, null), "llm_fallback", llmFallbackConfidence);
}
}
saveUnresolved(orgId, null, title, null, attempted);
log.info("Book '{}' not found by title in any source (attempted: {}), saved to unresolved_lookup", title, attempted);
return null;
}
private MtData persist(Integer orgId, String dedupeFieldName, String dedupeValue,
Map<String, String> values, String source, double confidence) {
var bookObject = bookQueryService.getBookObject(orgId);
var existing = bookQueryService.findByIndexedField(orgId, bookObject.getObjId(), dedupeFieldName, dedupeValue);
if (!existing.isEmpty()) {
return bookWriteService.enrichBook(orgId, existing.get(0).getGuid(), values, source, confidence);
}
return bookWriteService.createBook(orgId, values, source, confidence);
}
private Map<String, String> mapOpenLibrary(JsonNode book, String isbn) {
Map<String, String> values = new LinkedHashMap<>();
putIfPresent(values, "ISBN", isbn);
putIfPresent(values, "Title", textOrNull(book, "title"));
putIfPresent(values, "Subtitle", textOrNull(book, "subtitle"));
putIfPresent(values, "Pages", textOrNull(book, "number_of_pages"));
JsonNode authors = book.get("authors");
if (authors != null && authors.isArray()) {
String names = StreamSupport.stream(authors.spliterator(), false)
.map(a -> textOrNull(a, "name"))
.filter(java.util.Objects::nonNull)
.collect(Collectors.joining(", "));
putIfPresent(values, "AuthorName", names);
}
JsonNode publishers = book.get("publishers");
if (publishers != null && publishers.isArray() && !publishers.isEmpty()) {
putIfPresent(values, "Publisher", textOrNull(publishers.get(0), "name"));
}
putIfPresent(values, "PublishedYear", extractYear(textOrNull(book, "publish_date")));
JsonNode cover = book.get("cover");
if (cover != null) {
String coverUrl = textOrNull(cover, "large");
if (coverUrl == null) coverUrl = textOrNull(cover, "medium");
putIfPresent(values, "CoverUrl", coverUrl);
}
values.put("SourceStatus", "enriched_openlibrary");
return values;
}
/**
* Маппинг результата Open Library search.json (findByTitle) — формат ответа
* отличается от findByIsbn: author_name/publisher — плоские массивы строк,
* first_publish_year вместо publish_date, cover_i (числовой id) вместо
* объекта cover, isbn — массив (что удобно: иногда можно найти ISBN книги,
* который отсутствовал у вас изначально).
*/
private Map<String, String> mapOpenLibrarySearch(JsonNode doc) {
Map<String, String> values = new LinkedHashMap<>();
putIfPresent(values, "Title", textOrNull(doc, "title"));
JsonNode authorNames = doc.get("author_name");
if (authorNames != null && authorNames.isArray()) {
String names = StreamSupport.stream(authorNames.spliterator(), false)
.map(JsonNode::asText)
.collect(Collectors.joining(", "));
putIfPresent(values, "AuthorName", names);
}
JsonNode publishers = doc.get("publisher");
if (publishers != null && publishers.isArray() && !publishers.isEmpty()) {
putIfPresent(values, "Publisher", publishers.get(0).asText(null));
}
if (doc.has("first_publish_year") && !doc.get("first_publish_year").isNull()) {
putIfPresent(values, "PublishedYear", String.valueOf(doc.get("first_publish_year").asInt()));
}
JsonNode isbns = doc.get("isbn");
if (isbns != null && isbns.isArray() && !isbns.isEmpty()) {
putIfPresent(values, "ISBN", isbns.get(0).asText(null));
}
if (doc.has("cover_i") && !doc.get("cover_i").isNull()) {
putIfPresent(values, "CoverUrl", "https://covers.openlibrary.org/b/id/" + doc.get("cover_i").asText() + "-L.jpg");
}
values.put("SourceStatus", "enriched_openlibrary_by_title");
return values;
}
private Map<String, String> mapGoogleBooks(JsonNode volumeInfo, String isbn) {
Map<String, String> values = new LinkedHashMap<>();
putIfPresent(values, "ISBN", isbn);
putIfPresent(values, "Title", textOrNull(volumeInfo, "title"));
putIfPresent(values, "Subtitle", textOrNull(volumeInfo, "subtitle"));
putIfPresent(values, "Publisher", textOrNull(volumeInfo, "publisher"));
putIfPresent(values, "PublishedYear", extractYear(textOrNull(volumeInfo, "publishedDate")));
putIfPresent(values, "Pages", textOrNull(volumeInfo, "pageCount"));
putIfPresent(values, "Description", textOrNull(volumeInfo, "description"));
putIfPresent(values, "Language", textOrNull(volumeInfo, "language"));
JsonNode authors = volumeInfo.get("authors");
if (authors != null && authors.isArray()) {
String names = StreamSupport.stream(authors.spliterator(), false)
.map(JsonNode::asText)
.collect(Collectors.joining(", "));
putIfPresent(values, "AuthorName", names);
}
JsonNode categories = volumeInfo.get("categories");
if (categories != null && categories.isArray()) {
String genres = StreamSupport.stream(categories.spliterator(), false)
.map(JsonNode::asText)
.collect(Collectors.joining(", "));
putIfPresent(values, "Genre", genres);
}
JsonNode imageLinks = volumeInfo.get("imageLinks");
if (imageLinks != null) {
putIfPresent(values, "CoverUrl", textOrNull(imageLinks, "thumbnail"));
}
values.put("SourceStatus", isbn != null ? "enriched_googlebooks" : "enriched_googlebooks_by_title");
return values;
}
private Map<String, String> mapFantLab(JsonNode edition, String isbn) {
Map<String, String> values = new LinkedHashMap<>();
putIfPresent(values, "ISBN", isbn);
putIfPresent(values, "Title", FantLabClient.stripBbCode(textOrNull(edition, "name")));
putIfPresent(values, "AuthorName", FantLabClient.stripBbCode(textOrNull(edition, "autors")));
putIfPresent(values, "Publisher", FantLabClient.stripBbCode(textOrNull(edition, "publisher")));
putIfPresent(values, "Series", FantLabClient.stripBbCode(textOrNull(edition, "series")));
putIfPresent(values, "PublishedYear", textOrNull(edition, "year"));
putIfPresent(values, "Genre", "Фантастика");
values.put("SourceStatus", isbn != null ? "enriched_fantlab" : "enriched_fantlab_by_title");
return values;
}
private Map<String, String> mapLibGen(JsonNode edition, String isbn) {
Map<String, String> values = new LinkedHashMap<>();
putIfPresent(values, "ISBN", isbn);
putIfPresent(values, "Title", textOrNull(edition, "title"));
putIfPresent(values, "AuthorName", textOrNull(edition, "author"));
putIfPresent(values, "Publisher", textOrNull(edition, "publisher"));
putIfPresent(values, "Pages", textOrNull(edition, "pages"));
putIfPresent(values, "PublishedYear", extractYear(textOrNull(edition, "year")));
putIfPresent(values, "CoverUrl", textOrNull(edition, "cover_url"));
values.put("SourceStatus", "enriched_libgen");
return values;
}
/**
* Маппинг результата LLM fallback. source_url сохраняется в поле SourceUrl
* для ручной проверки — этот источник наименее надёжен из всех.
*/
private Map<String, String> mapLlmFallback(JsonNode result, String isbn) {
Map<String, String> values = new LinkedHashMap<>();
putIfPresent(values, "ISBN", isbn != null ? isbn : textOrNull(result, "isbn"));
putIfPresent(values, "Title", textOrNull(result, "title"));
putIfPresent(values, "AuthorName", textOrNull(result, "author"));
putIfPresent(values, "Publisher", textOrNull(result, "publisher"));
putIfPresent(values, "PublishedYear", textOrNull(result, "year"));
String sourceUrl = textOrNull(result, "source_url");
putIfPresent(values, "SourceUrl", sourceUrl);
values.put("SourceStatus", "enriched_llm_fallback_needs_review");
return values;
}
private void saveUnresolved(Integer orgId, String isbn, String title, String authorName, List<String> attempted) {
UnresolvedLookup existing = isbn != null
? unresolvedLookupRepository.findByOrgIdAndIsbn(orgId, isbn).orElse(null)
: null;
if (existing != null) {
List<String> sources = new ArrayList<>(java.util.Arrays.asList(existing.getAttemptedSources()));
for (String s : attempted) {
if (!sources.contains(s)) sources.add(s);
}
existing.setAttemptedSources(sources.toArray(new String[0]));
existing.setUpdatedAt(OffsetDateTime.now());
unresolvedLookupRepository.save(existing);
return;
}
UnresolvedLookup lookup = UnresolvedLookup.builder()
.orgId(orgId)
.isbn(isbn)
.title(title)
.authorName(authorName)
.attemptedSources(attempted.toArray(new String[0]))
.status("pending")
.rawContext(new HashMap<>())
.build();
unresolvedLookupRepository.save(lookup);
}
private void putIfPresent(Map<String, String> map, String key, String value) {
if (value != null && !value.isBlank()) map.put(key, value);
}
private String extractYear(String dateStr) {
if (dateStr == null) return null;
java.util.regex.Matcher m = java.util.regex.Pattern.compile("(\\d{4})").matcher(dateStr);
return m.find() ? m.group(1) : null;
}
private String textOrNull(JsonNode node, String field) {
if (node == null) return null;
JsonNode value = node.get(field);
return (value == null || value.isNull()) ? null : value.asText();
}
}