Search content

17 min

Optimizely Graph provides full-text search using the match operator. Sort results with traditional relevance-based algorithms or with semantic search powered by AI.

match operator

The match operator searches full-text fields across content stored in Optimizely Graph. It supports multiple ranking modes and can be combined with filters, boosting, and facets.

Ranking modes

Relevance

Use relevance ranking for keyword-based searches. It uses traditional algorithms such as TF-IDF and BM25. Relevance ranking is fast, predictable, and works well with exact matches.

Semantic

Use semantic ranking for natural-language queries. It uses AI models to understand meaning and context. Semantic ranking understands synonyms, intent, and related concepts, which makes it suitable for user questions and conceptual searches.

When to use each ranking type

Use case	Recommended ranking	Reason
Product catalog search	Relevance	Exact matches for product names and SKUs.
Technical documentation	Relevance	Matches precise terminology.
Category browsing	Relevance	Matches categories and tags.
Knowledge base search	Semantic	Understands user intent and question.
News or article search	Semantic	Finds related topics and concepts.
FAQ search	Semantic	Handles natural language questions.

Basic GraphQL queries

Run a basic text search

Search across all content types by matching a query term.

query SearchContent($query: String!) {
  Content(
    where: { _fulltext: { match: $query } }
    orderBy: { _ranking: RELEVANCE }
  ) {
    items {
      Name
      RelativePath
      _score
    }
    total
  }
}

Variables

{ "query": "sustainability" }

Sort results by relevance

Relevance ranking uses traditional keyword-based algorithms.

query SearchByRelevance($query: String!) {
  ArticlePage(
    where: { _fulltext: { match: $query } }
    orderBy: { _ranking: RELEVANCE }
    limit: 10
  ) {
    items {
      Name
      TeaserText
      RelativePath
      _score
    }
    total
  }
}

Sort results by semantic ranking

Semantic ranking uses AI models to understand meaning and context.

query SearchBySemantic($query: String!) {
  ArticlePage(
    where: { _fulltext: { match: $query } }
    orderBy: { _ranking: SEMANTIC }
    limit: 10
  ) {
    items {
      Name
      TeaserText
      RelativePath
      _score
    }
    total
  }
}

Search across multiple fields

Target specific fields while maintaining relevance scoring.

query MultiFieldSearch($query: String!) {
  ArticlePage(
    where: {
      _or: [
        { Name: { match: $query } },
        { TeaserText: { match: $query } },
        { MainBody: { match: $query } }
      ]
    }
    orderBy: { _ranking: RELEVANCE }
  ) {
    items {
      Name
      TeaserText
      RelativePath
      _score
    }
    total
  }
}

Filter by content type

Combine search with content type filtering.

query SearchArticles($query: String!) {
  Content(
    where: {
      _and: [
        { _fulltext: { match: $query } },
        { _typeName: { eq: "ArticlePage" } }
      ]
    }
    orderBy: { _ranking: SEMANTIC }
    limit: 5
  ) {
    items {
      Name
      RelativePath
      _score
    }
    total
  }
}

Boost specific fields

Increase the relevance of certain fields.

query AdvancedSearch($query: String!) {
  ArticlePage(
    where: {
      _or:[
        { Name: { match: $query, boost: 5 } },
        { TeaserText: { match: $query, boost: 3 } },
        { _fulltext: { match: $query } }
      ]
    }
    orderBy: { _ranking: RELEVANCE }
  ) {
    items {
      Name
      TeaserText
      _score
    }
    total
  }
}

Apply date range filters

Restrict results to recently published content.

query SearchWithFilters($query: String!, $fromDate: DateTime!) {
  ArticlePage(
    where: {
      _and: [
        { _fulltext: { match: $query } },
        { StartPublish: { gte: $fromDate } }
      ]
    }
    orderBy: { _ranking: SEMANTIC }
    limit: 20
  ) {
    items {
      Name
      TeaserText
      RelativePath
      _score
      StartPublish
    }
    total
  }
}

Use faceted search

Retrieve facet counts for filtering.

query FacetedSearch($query: String!) {
  search: Content(
    where: { _fulltext: { match: $query } }
    orderBy: { _ranking: RELEVANCE }
    limit: 10
  ) {
    items {
      Name
      RelativePath
      _score
    }
    total
  }

  typeFacets: Content(
    where: { _fulltext: { match: $query } }
    facet: { _typeName(limit: 10) }
  ) {
    facet {
      _typeName {
        name
        count
      }
    }
  }
}

Paginate search results

Retrieve search results in pages.

query PaginatedSearch($query: String!, $skip: Int = 0, $limit: Int = 10) {
  Content(
    where: { _fulltext: { match: $query } }
    orderBy: { _ranking: SEMANTIC }
    skip: $skip
    limit: $limit
  ) {
    items {
      Name
      RelativePath
      _score
    }
    total
  }
}

References

Search scores

The _score field indicates the relevance of a search result.

Higher scores mean higher relevance.

Scores vary depending on the ranking type.

Filter results based on a minimum _score.

const { data } = useQuery(SEARCH_QUERY, {
  variables: { query: searchTerm }
});

const relevantResults = data.Content.items.filter(item => item._score > 0.5);

C# implementation

public async Task<SearchResult> SearchContentAsync(string query, SearchType searchType = SearchType.Semantic)
{
    var orderBy = searchType == SearchType.Semantic ? "SEMANTIC" : "RELEVANCE";
    
    var graphQLQuery = $@"
        query SearchContent($query: String!) {{
            Content(
                where: {{ _fulltext: {{ match: $query }} }}
                orderBy: {{ _ranking: {orderBy} }}
                limit: 20
            ) {{
                items {{
                    Name
                    RelativePath
                    _score
                }}
                total
            }}
        }}";

    var request = new GraphQLRequest 
    { 
        Query = graphQLQuery,
        Variables = new { query }
    };

    var response = await _graphQLClient.SendQueryAsync<SearchResponse>(request);
    return response.Data;
}

public enum SearchType
{
    Relevance,
    Semantic
}