# Search DSL

Curiosity Workspace search can be used via UI components and APIs. This page documents the common request shape and mental model for building search experiences programmatically.

Most search requests specify:

• Query text: what the user typed
• Type scope: which node types are eligible results
• Filters/facets: property facets and related facets
• Target set (optional): restrict search to a set of nodes (often computed by a graph query)
• Sort mode: relevance, recency, or a domain-specific ordering
• Hybrid settings (optional): how keyword and vector retrieval combine

var request = SearchRequest.For("MacBook Air");
request.BeforeTypesFacet = new([] { "SupportCase", "Device", "Part" });

var query = await Graph.CreateSearchAsync(request);
return query.Emit();

This pattern is key to “search within context”.

var request = SearchRequest.For("screen flicker");
request.BeforeTypesFacet = new([] { "SupportCase" });

// Restrict search to nodes returned by a graph traversal (example: all items for a manufacturer)
request.TargetUIDs = Q().StartAt("Manufacturer", "Apple").Out().AsUIDEnumerable().ToArray();

var query = await Graph.CreateSearchAsync(request);
return query.Emit();

If embedding indexes are enabled, similarity search can be used as a standalone retrieval method or as part of hybrid ranking.

return Q().StartAtSimilarText("Apple screen issue", nodeTypes: new[] { "SupportCase" })
          .EmitWithScores();

• Explicit type scope: always constrain to the relevant types for the user experience.
• Use facets: facets are the primary UX mechanism to reduce result sets.
• Prefer context constraints: compute a target set using graph traversal and pass it into search.
• Evaluate changes: changes to boosts, cutoffs, and hybrid logic should be tested with a query set.

• Search concepts: Core Concepts → Search Model
• Tuning: Search → Ranking Tuning