Configuration

This reference collects the supported indexer, writer, client, and offline options without presenting planned settings as current behavior.

BuildIndexOptions

• fieldBoosts
• synonyms
• fuzzy and fuzzyMaxEdits
• hierarchicalFacets
• rangeFacetBuckets
• allowedUrlOrigins and canonicalBaseUrl

buildIndex(sources, defaultLanguage, options) is synchronous. Source IDs must be stable and unique. Vector building is separate because embedding can be asynchronous.

WriteIndexOptions

• maxShardGzipBytes and shardByPrefix
• termShardFormat, fuzzyShardFormat, and docStoreFormat
• docStoreShardSize
• vectors

All three format options default to "json"; prefix sharding defaults on. vectors is the BuiltVectors value returned by buildVectorShards.

SearchClientOptions

• required indexUrl
• worker and workerUrl
• allowCrossOriginShards and strict
• embedQuery and validateVectorProvider

Query and offline options

Search configuration is per call through SearchOptions; see Client API. registerOfflineCaching(swUrl, indexUrl, options) accepts mode: "cache-first" | "stale-while-revalidate", optional languages, and allowCrossOriginShards.

For a small site, defaults are appropriate. Larger corpora can lower maxShardGzipBytes, set docStoreShardSize, or opt hot shard families into binary after measuring. Internationalization, synonyms, fuzzy dictionaries, and vectors should be configured only for languages and features the deployment actually uses.