Skip to main content

Interface: ISearchEngine

Defined in: src/types/interfaces/search.interface.ts:273

Search engine interface for Elasticsearch.

Remarks​

Provides full-text search, faceted search, and autocomplete for eprints.

Implementation notes:

  • Uses Elasticsearch 8+
  • Indices prefixed with chive-eprints-
  • Ingest pipeline for PDF text extraction
  • Custom analyzers for academic text

Methods​

autocomplete()​

autocomplete(prefix, limit?): Promise<readonly string[]>

Defined in: src/types/interfaces/search.interface.ts:381

Provides autocomplete suggestions.

Parameters​

prefix​

string

Partial query string

limit?​

number

Maximum number of suggestions

Returns​

Promise<readonly string[]>

Suggested completions

Remarks​

Uses Elasticsearch completion suggester for fast autocomplete. Suggestions are based on indexed titles and keywords.

Example​

const suggestions = await searchEngine.autocomplete('neura', 5);
// ['neural', 'neural networks', 'neuromorphic', ...]

deleteDocument()​

deleteDocument(uri): Promise<void>

Defined in: src/types/interfaces/search.interface.ts:401

Removes a document from the index.

Parameters​

uri​

AtUri

Document URI

Returns​

Promise<void>

Promise resolving when deleted

Remarks​

Use when an eprint is deleted from the user's PDS.

Example​

await searchEngine.deleteDocument(
  toAtUri('at://did:plc:abc/pub.chive.eprint.submission/xyz')!
);

facetedSearch()​

facetedSearch(query): Promise<FacetedSearchResults>

Defined in: src/types/interfaces/search.interface.ts:360

Performs faceted search with aggregations.

Parameters​

query​

FacetedSearchQuery

Faceted search query

Returns​

Promise<FacetedSearchResults>

Results with facet counts

Remarks​

Extends regular search with aggregation results for each facet dimension. Use for browse interfaces with faceted navigation.

Example​

const results = await searchEngine.facetedSearch({
  q: 'neural networks',
  facets: ['subjects', 'year'],
  limit: 20
});

console.log('Subjects:', results.facets.subjects);
// [{ value: "Computer Science", count: 150 }, ...]

findSimilarByText()​

findSimilarByText(documentUri, options?): Promise<readonly MLTResult[]>

Defined in: src/types/interfaces/search.interface.ts:429

Finds documents similar to a given document using text content.

Parameters​

documentUri​

AtUri

AT-URI of the source document

options?​

MLTOptions

Options controlling the More Like This query

Returns​

Promise<readonly MLTResult[]>

Matching documents sorted by similarity score

Remarks​

Uses Elasticsearch More Like This query on title, abstract, and keywords to find content-similar documents. Serves as a fallback when embedding-based similarity (SPECTER2) is unavailable.

Example​

const similar = await searchEngine.findSimilarByText(
  toAtUri('at://did:plc:abc/pub.chive.eprint.submission/xyz')!,
  { limit: 10, minimumShouldMatch: '30%' }
);

for (const result of similar) {
  console.log(`${result.uri} (score: ${result.score})`);
}

indexEprint()​

indexEprint(eprint): Promise<void>

Defined in: src/types/interfaces/search.interface.ts:304

Indexes an eprint document.

Parameters​

eprint​

IndexableEprintDocument

Eprint to index

Returns​

Promise<void>

Promise resolving when indexed

Remarks​

Indexes or updates the eprint in Elasticsearch. If full-text extraction is enabled, triggers PDF processing via ingest pipeline.

Example​

await searchEngine.indexEprint({
  uri: toAtUri('at://did:plc:abc/pub.chive.eprint.submission/xyz')!,
  author: toDID('did:plc:abc')!,
  authorName: 'Dr. Jane Smith',
  title: 'Neural Networks in Biology',
  abstract: 'This paper explores...',
  keywords: ['neural networks', 'biology'],
  fieldNodes: [
    { id: 'cs-uuid', label: 'Computer Science' },
    { id: 'bio-uuid', label: 'Biology' }
  ],
  createdAt: new Date(),
  indexedAt: new Date()
});

search(query): Promise<SearchResults>

Defined in: src/types/interfaces/search.interface.ts:334

Searches eprints.

Parameters​

query​

SearchQuery

Search query

Returns​

Promise<SearchResults>

Search results

Remarks​

Performs full-text search across title, abstract, and full-text (if available). Returns results sorted by relevance score.

Example​

const results = await searchEngine.search({
  q: 'neural networks biology',
  filters: {
    subjects: ['Computer Science'],
    dateFrom: new Date('2024-01-01')
  },
  limit: 20
});

results.hits.forEach(hit => {
  console.log(`${hit.uri} (score: ${hit.score})`);
});