Whether to support phrase matching and excluding words from search queries.

Use the advancedSyntaxFeatures parameter to control which feature is supported.

Advanced search syntax features you want to support.

• exactPhrase. Phrases in quotes must match exactly. For example, sparkly blue "iPhone case" only returns records with the exact string "iPhone case".

• excludeWords. Query words prefixed with a - must not occur in a record. For example, search -engine matches records that contain "search" but not "engine".

This setting only has an effect if advancedSyntax is true.

Whether arrays with exclusively non-negative integers should be compressed for better performance. If true, the compressed arrays may be reordered.

Whether to allow typos on numbers in the search query.

Turn off this setting to reduce the number of irrelevant matches when searching in large sets of similar numbers.

Alternatives of query words that should be considered as exact matches by the Exact ranking criterion.

• ignorePlurals. Plurals and similar declensions added by the ignorePlurals setting are considered exact matches.

• singleWordSynonym. Single-word synonyms, such as "NY/NYC" are considered exact matches.

• multiWordsSynonym. Multi-word synonyms, such as "NY/New York" are considered exact matches.

Whether the best matching attribute should be determined by minimum proximity.

This setting only affects ranking if the Attribute ranking criterion comes before Proximity in the ranking setting. If true, the best matching attribute is selected based on the minimum proximity of multiple matches. Otherwise, the best matching attribute is determined by the order in the searchableAttributes setting.

Attribute that should be used to establish groups of results. Attribute names are case-sensitive.

All records with the same value for this attribute are considered a group. You can combine attributeForDistinct with the distinct search parameter to control how many items per group are included in the search results.

If you want to use the same attribute also for faceting, use the afterDistinct modifier of the attributesForFaceting setting. This applies faceting after deduplication, which will result in accurate facet counts.

Attributes used for faceting.

Facets are attributes that let you categorize search results. They can be used for filtering search results. By default, no attribute is used for faceting. Attribute names are case-sensitive.

Modifiers

• filterOnly("ATTRIBUTE"). Allows using this attribute as a filter, but doesn't evalue the facet values.

• searchable("ATTRIBUTE"). Allows searching for facet values.

• afterDistinct("ATTRIBUTE"). Evaluates the facet count after deduplication with distinct. This ensures accurate facet counts. You can apply this modifier to searchable facets: afterDistinct(searchable(ATTRIBUTE)).

Attributes to highlight.

By default, all searchable attributes are highlighted. Use * to highlight all attributes or use an empty array [] to turn off highlighting. Attribute names are case-sensitive.

With highlighting, strings that match the search query are surrounded by HTML tags defined by highlightPreTag and highlightPostTag. You can use this to visually highlight matching parts of a search query in your UI.

For more information, see Highlighting and snippeting.

Attributes to include in the API response.

To reduce the size of your response, you can retrieve only some of the attributes. Attribute names are case-sensitive.

• * retrieves all attributes, except attributes included in the customRanking and unretrievableAttributes settings.
• To retrieve all attributes except a specific one, prefix the attribute with a dash and combine it with the *: ["*", "-ATTRIBUTE"].
• The objectID attribute is always included.

Attributes for which to enable snippets. Attribute names are case-sensitive.

Snippets provide additional context to matched words. If you enable snippets, they include 10 words, including the matched word. The matched word will also be wrapped by HTML tags for highlighting. You can adjust the number of words with the following notation: ATTRIBUTE:NUMBER, where NUMBER is the number of words to be extracted.

Attributes, for which you want to support Japanese transliteration.

Transliteration supports searching in any of the Japanese writing systems. To support transliteration, you must set the indexing language to Japanese. Attribute names are case-sensitive.

Attributes for which to split camel case words. Attribute names are case-sensitive.

Characters and their normalized replacements. This overrides Algolia's default normalization.

Attributes to use as custom ranking. Attribute names are case-sensitive.

The custom ranking attributes decide which items are shown first if the other ranking criteria are equal.

Records with missing values for your selected custom ranking attributes are always sorted last. Boolean attributes are sorted based on their alphabetical order.

Modifiers

• asc("ATTRIBUTE"). Sort the index by the values of an attribute, in ascending order.

• desc("ATTRIBUTE"). Sort the index by the values of an attribute, in descending order.

If you use two or more custom ranking attributes, reduce the precision of your first attributes, or the other attributes will never be applied.

Searchable attributes to which Algolia should apply word segmentation (decompounding). Attribute names are case-sensitive.

Compound words are formed by combining two or more individual words, and are particularly prevalent in Germanic languages—for example, "firefighter". With decompounding, the individual components are indexed separately.

You can specify different lists for different languages. Decompounding is supported for these languages: Dutch (nl), German (de), Finnish (fi), Danish (da), Swedish (sv), and Norwegian (no).

Whether to split compound words into their building blocks.

For more information, see Word segmentation. Word segmentation is supported for these languages: German, Dutch, Finnish, Swedish, and Norwegian.

Searchable attributes for which you want to turn off the Exact ranking criterion. Attribute names are case-sensitive.

This can be useful for attributes with long values, where the likelyhood of an exact match is high, such as product descriptions. Turning off the Exact ranking criterion for these attributes favors exact matching on other attributes. This reduces the impact of individual attributes with a lot of content on ranking.

Searchable attributes for which you want to turn off prefix matching. Attribute names are case-sensitive.

Attributes for which you want to turn off typo tolerance. Attribute names are case-sensitive.

Returning only exact matches can help when:

• Searching in hyphenated attributes.
• Reducing the number of matches when you have too many. This can happen with attributes that are long blocks of text, such as product descriptions.

Consider alternatives such as disableTypoToleranceOnWords or adding synonyms if your attributes have intentional unusual spellings that might look like typos.

Words for which you want to turn off typo tolerance. This also turns off word splitting and concatenation for the specified words.

Determines how many records of a group are included in the search results.

Records with the same value for the attributeForDistinct attribute are considered a group. The distinct setting controls how many members of the group are returned. This is useful for deduplication and grouping.

The distinct setting is ignored if attributeForDistinct is not set.

Whether to enable Personalization.

Whether this search will use Dynamic Re-Ranking.

This setting only has an effect if you activated Dynamic Re-Ranking for this index in the Algolia dashboard.

Whether to enable rules.

Determines how the Exact ranking criterion is computed when the search query has only one word.

• attribute. The Exact ranking criterion is 1 if the query word and attribute value are the same. For example, a search for "road" will match the value "road", but not "road trip".

• none. The Exact ranking criterion is ignored on single-word searches.

• word. The Exact ranking criterion is 1 if the query word is found in the attribute value. The query word must have at least 3 characters and must not be a stop word. Only exact matches will be highlighted, partial and prefix matches won't.

HTML tag to insert after the highlighted parts in all highlighted results and snippets.

HTML tag to insert before the highlighted parts in all highlighted results and snippets.

Number of hits per page.

Treat singular, plurals, and other forms of declensions as equivalent. You should only use this feature for the languages used in your index.

Languages for language-specific processing steps, such as word detection and dictionary settings.

You should always specify an indexing language. If you don't specify an indexing language, the search engine uses all supported languages, or the languages you specified with the ignorePlurals or removeStopWords parameters. This can lead to unexpected search results. For more information, see Language-specific configuration.

Characters for which diacritics should be preserved.

By default, Algolia removes diacritics from letters. For example, é becomes e. If this causes issues in your search, you can specify characters that should keep their diacritics.

Maximum number of facet values to return when searching for facet values.

Maximum number of facet values to return for each facet.

Minimum proximity score for two matching words.

This adjusts the Proximity ranking criterion by equally scoring matches that are farther apart.

For example, if minProximity is 2, neighboring matches and matches with one word between them would have the same score.

Minimum number of characters a word in the search query must contain to accept matches with one typo.

Minimum number of characters a word in the search query must contain to accept matches with two typos.

Search mode the index will use to query for results.

This setting only applies to indices, for which Algolia enabled NeuralSearch for you.

Numeric attributes that can be used as numerical filters. Attribute names are case-sensitive.

By default, all numeric attributes are available as numerical filters. For faster indexing, reduce the number of numeric attributes.

If you want to turn off filtering for all numeric attributes, specifiy an attribute that doesn't exist in your index, such as NO_NUMERIC_FILTERING.

Modifier

• equalOnly("ATTRIBUTE"). Support only filtering based on equality comparisons = and !=.

Words that should be considered optional when found in the query.

By default, records must match all words in the search query to be included in the search results. Adding optional words can help to increase the number of search results by running an additional search query that doesn't include the optional words. For example, if the search query is "action video" and "video" is an optional word, the search engine runs two queries. One for "action video" and one for "action". Records that match all words are ranked higher.

For a search query with 4 or more words and all its words are optional, the number of matched words required for a record to be included in the search results increases for every 1,000 records:

• If optionalWords has less than 10 words, the required number of matched words increases by 1: results 1 to 1,000 require 1 matched word, results 1,001 to 2000 need 2 matched words.
• If optionalWords has 10 or more words, the number of required matched words increases by the number of optional words dividied by 5 (rounded down). For example, with 18 optional words: results 1 to 1,000 require 1 matched word, results 1,001 to 2000 need 4 matched words.

For more information, see Optional words.

Maximum number of search results that can be obtained through pagination.

Higher pagination limits might slow down your search. For pagination limits above 1,000, the sorting of results beyond the 1,000th hit can't be guaranteed.

Languages for language-specific query processing steps such as plurals, stop-word removal, and word-detection dictionaries.

This setting sets a default list of languages used by the removeStopWords and ignorePlurals settings. This setting also sets a dictionary for word detection in the logogram-based CJK languages. To support this, you must place the CJK language first.

You should always specify a query language. If you don't specify an indexing language, the search engine uses all supported languages, or the languages you specified with the ignorePlurals or removeStopWords parameters. This can lead to unexpected search results. For more information, see Language-specific configuration.

Determines if and how query words are interpreted as prefixes.

By default, only the last query word is treated as prefix (prefixLast). To turn off prefix search, use prefixNone. Avoid prefixAll, which treats all query words as prefixes. This might lead to counterintuitive results and makes your search slower.

For more information, see Prefix searching.

Determines the order in which Algolia returns your results.

By default, each entry corresponds to a ranking criteria. The tie-breaking algorithm sequentially applies each criterion in the order they're specified. If you configure a replica index for sorting by an attribute, you put the sorting attribute at the top of the list.

Modifiers

• asc("ATTRIBUTE"). Sort the index by the values of an attribute, in ascending order.
• desc("ATTRIBUTE"). Sort the index by the values of an attribute, in descending order.

Before you modify the default setting, you should test your changes in the dashboard, and by A/B testing.

Relevancy threshold below which less relevant results aren't included in the results.

You can only set relevancyStrictness on virtual replica indices. Use this setting to strike a balance between the relevance and number of returned results.

Removes stop words from the search query.

Stop words are common words like articles, conjunctions, prepositions, or pronouns that have little or no meaning on their own. In English, "the", "a", or "and" are stop words.

You should only use this feature for the languages used in your index.

Strategy for removing words from the query when it doesn't return any results. This helps to avoid returning empty search results.

• none. No words are removed when a query doesn't return results.

• lastWords. Treat the last (then second to last, then third to last) word as optional, until there are results or at most 5 words have been removed.

• firstWords. Treat the first (then second, then third) word as optional, until there are results or at most 5 words have been removed.

• allOptional. Treat all words as optional.

For more information, see Remove words to improve results.

Extra data that can be used in the search UI.

You can use this to control aspects of your search UI, such as, the order of facet names and values without changing your frontend code.

Whether to replace a highlighted word with the matched synonym.

By default, the original words are highlighted even if a synonym matches. For example, with home as a synonym for house and a search for home, records matching either "home" or "house" are included in the search results, and either "home" or "house" are highlighted.

With replaceSynonymsInHighlight set to true, a search for home still matches the same records, but all occurences of "house" are replaced by "home" in the highlighted response.

Creates replica indices.

Replicas are copies of a primary index with the same records but different settings, synonyms, or rules. If you want to offer a different ranking or sorting of your search results, you'll use replica indices. All index operations on a primary index are automatically forwarded to its replicas. To add a replica index, you must provide the complete set of replicas to this parameter. If you omit a replica from this list, the replica turns into a regular, standalone index that will no longer by synced with the primary index.

Modifier

• virtual("REPLICA"). Create a virtual replica, Virtual replicas don't increase the number of records and are optimized for Relevant sorting.

Restrict Dynamic Re-Ranking to records that match these filters.

Properties to include in the API response of search and browse requests.

By default, all response properties are included. To reduce the response size, you can select, which attributes should be included.

You can't exclude these properties: message, warning, cursor, serverUsed, indexUsed, abTestVariantID, parsedQuery, or any property triggered by the getRankingInfo parameter.

Don't exclude properties that you might need in your search UI.

Whether to restrict highlighting and snippeting to items that at least partially matched the search query. By default, all items are highlighted and snippeted.

Attributes used for searching. Attribute names are case-sensitive.

By default, all attributes are searchable and the Attribute ranking criterion is turned off. With a non-empty list, Algolia only returns results with matches in the selected attributes. In addition, the Attribute ranking criterion is turned on: matches in attributes that are higher in the list of searchableAttributes rank first. To make matches in two attributes rank equally, include them in a comma-separated string, such as "title,alternate_title". Attributes with the same priority are always unordered.

For more information, see Searchable attributes.

Modifier

• unordered("ATTRIBUTE"). Ignore the position of a match within the attribute.

Without modifier, matches at the beginning of an attribute rank higer than matches at the end.

Settings for the semantic search part of NeuralSearch. Only used when mode is neuralSearch.

Controls which separators are indexed.

Separators are all non-letter characters except spaces and currency characters, such as $€£¥. By default, separator characters aren't indexed. With separatorsToIndex, Algolia treats separator characters as separate words. For example, a search for C# would report two matches.

String used as an ellipsis indicator when a snippet is truncated.

Order in which to retrieve facet values.

• count. Facet values are retrieved by decreasing count. The count is the number of matching records containing this facet value.

• alpha. Retrieve facet values alphabetically.

This setting doesn't influence how facet values are displayed in your UI (see renderingContent). For more information, see facet value display.

Whether typo tolerance is enabled and how it is applied.

If typo tolerance is true, min, or strict, word splitting and concetenation is also active.

Attributes that can't be retrieved at query time.

This can be useful if you want to use an attribute for ranking or to restrict access, but don't want to include it in the search results. Attribute names are case-sensitive.

Index name (case-sensitive).

Operation to perform on the index.

Coordinates for the center of a circle, expressed as a comma-separated string of latitude and longitude.

Only records included within circle around this central location are included in the results. The radius of the circle is determined by the aroundRadius and minimumAroundRadius settings. This parameter is ignored if you also specify insidePolygon or insideBoundingBox.

Maximum radius for a search around a central location.

This parameter works in combination with the aroundLatLng and aroundLatLngViaIP parameters. By default, the search radius is determined automatically from the density of hits around the central location. The search radius is small if there are many hits close to the central coordinates.

Filter the search by facet values, so that only records with the same facet values are retrieved.

Prefer using the filters parameter, which supports all filter types and combinations with boolean operators.

• [filter1, filter2] is interpreted as filter1 AND filter2.
• [[filter1, filter2], filter3] is interpreted as filter1 OR filter2 AND filter3.
• facet:-value is interpreted as NOT facet:value.

While it's best to avoid attributes that start with a -, you can still filter them by escaping with a backslash: facet:\-value.

Filter expression to only include items that match the filter criteria in the response.

You can use these filter expressions:

• Numeric filters. <facet> <op> <number>, where <op> is one of <, <=, =, !=, >, >=.
• Ranges. <facet>:<lower> TO <upper> where <lower> and <upper> are the lower and upper limits of the range (inclusive).
• Facet filters. <facet>:<value> where <facet> is a facet attribute (case-sensitive) and <value> a facet value.
• Tag filters. _tags:<value> or just <value> (case-sensitive).
• Boolean filters. <facet>: true | false.

You can combine filters with AND, OR, and NOT operators with the following restrictions:

• You can only combine filters of the same type with OR. Not supported: facet:value OR num > 3.
• You can't use NOT with combinations of filters. Not supported: NOT(facet:value OR facet:value)
• You can't combine conjunctions (AND) with OR. Not supported: facet:value OR (facet:value AND facet:value)

Use quotes around your filters, if the facet attribute name or facet value has spaces, keywords (OR, AND, NOT), or quotes. If a facet attribute is an array, the filter matches if it matches at least one element of the array.

For more information, see Filters.

Coordinates for a rectangular area in which to search.

Each bounding box is defined by the two opposite points of its diagonal, and expressed as latitude and longitude pair: [p1 lat, p1 long, p2 lat, p2 long]. Provide multiple bounding boxes as nested arrays. For more information, see rectangular area.

Coordinates of a polygon in which to search.

Polygons are defined by 3 to 10,000 points. Each point is represented by its latitude and longitude. Provide multiple polygons as nested arrays. For more information, see filtering inside polygons. This parameter is ignored if you also specify insideBoundingBox.

Filter by numeric facets.

Prefer using the filters parameter, which supports all filter types and combinations with boolean operators.

You can use numeric comparison operators: <, <=, =, !=, >, >=. Comparsions are precise up to 3 decimals. You can also provide ranges: facet:<lower> TO <upper>. The range includes the lower and upper boundaries. The same combination rules apply as for facetFilters.