Documentation
¶
Overview ¶
Package search provides search implementations for gognee's knowledge graph.
Index ¶
Constants ¶
This section is empty.
Variables ¶
var ErrNoSeeds = errors.New("graph search requires seed node IDs in SearchOptions.SeedNodeIDs")
ErrNoSeeds is returned when graph search is attempted without seed nodes.
Functions ¶
func ApplyDefaults ¶
func ApplyDefaults(opts *SearchOptions)
ApplyDefaults sets default values for unspecified search options.
Types ¶
type DecayingSearcher ¶
type DecayingSearcher struct {
// contains filtered or unexported fields
}
DecayingSearcher is a decorator that applies time-based decay to search results. It wraps any Searcher implementation and modifies scores based on node age.
func NewDecayingSearcher ¶
func NewDecayingSearcher( underlying Searcher, graphStore store.GraphStore, memoryStore store.MemoryStore, enabled bool, halfLifeDays int, basis string, accessFrequencyEnabled bool, referenceAccessCount int, ) *DecayingSearcher
NewDecayingSearcher creates a new decaying searcher wrapper.
Parameters:
- underlying: The base searcher to wrap
- graphStore: Graph store for retrieving node timestamps
- memoryStore: Memory store for retrieving access counts (M2: Plan 021)
- enabled: Whether decay is enabled
- halfLifeDays: Number of days for half-life decay
- basis: "access" (use last_accessed_at) or "creation" (use created_at)
- accessFrequencyEnabled: Whether to apply access frequency heat multiplier (M2: Plan 021)
- referenceAccessCount: Reference count for full heat protection (M2: Plan 021)
func (*DecayingSearcher) Search ¶
func (d *DecayingSearcher) Search(ctx context.Context, query string, opts SearchOptions) ([]SearchResult, error)
Search performs search with decay applied to scores.
func (*DecayingSearcher) SetLogger ¶ added in v1.6.0
func (d *DecayingSearcher) SetLogger(logger *slog.Logger)
SetLogger sets the structured logger for this DecayingSearcher (M8: Plan 023). When nil, logging is disabled (zero overhead).
type GraphSearcher ¶
type GraphSearcher struct {
// contains filtered or unexported fields
}
GraphSearcher performs graph traversal search from seed nodes.
func NewGraphSearcher ¶
func NewGraphSearcher(graphStore store.GraphStore) *GraphSearcher
NewGraphSearcher creates a new graph searcher.
func (*GraphSearcher) Search ¶
func (g *GraphSearcher) Search(ctx context.Context, query string, opts SearchOptions) ([]SearchResult, error)
Search performs graph traversal from seed nodes. The query parameter is ignored (graph search uses opts.SeedNodeIDs).
type HybridSearcher ¶
type HybridSearcher struct {
// contains filtered or unexported fields
}
HybridSearcher combines vector similarity and graph traversal search.
func NewHybridSearcher ¶
func NewHybridSearcher( embClient embeddings.EmbeddingClient, vectorStore store.VectorStore, graphStore store.GraphStore, ) *HybridSearcher
NewHybridSearcher creates a new hybrid searcher.
func (*HybridSearcher) Search ¶
func (h *HybridSearcher) Search(ctx context.Context, query string, opts SearchOptions) ([]SearchResult, error)
Search performs hybrid search combining vector similarity and graph expansion. Score formula: combined_score = vector_score + graph_score where vector_score = 0 if not found by vector, graph_score = 0 if not found by graph.
type SearchOptions ¶
type SearchOptions struct {
Type SearchType // Type of search to perform
TopK int // Maximum number of results to return (default: 10)
GraphDepth int // Maximum graph traversal depth (default: 1)
// SeedNodeIDs specifies starting nodes for graph search.
// Required for SearchTypeGraph; ignored for SearchTypeVector.
// For SearchTypeHybrid, seeds augment vector results.
SeedNodeIDs []string
// IncludeMemoryIDs enables memory provenance enrichment (v1.0.0+).
// Default: true. Set to false to skip provenance lookup for performance.
IncludeMemoryIDs *bool
// TraceEnabled enables detailed timing instrumentation for performance analysis.
// Default: false (off by default to minimize overhead).
TraceEnabled bool
}
SearchOptions configures search behavior.
type SearchResult ¶
type SearchResult struct {
NodeID string // Unique identifier of the node
Node *store.Node // Full node data (nil if node was deleted)
Score float64 // Combined relevance score (higher is better)
Source string // Origin: "vector", "graph", or "hybrid"
// GraphDepth indicates the minimum graph distance from the search origin.
// 0 for direct vector hits, >0 for nodes discovered via graph expansion.
GraphDepth int
// MemoryIDs lists memory IDs that reference this node (v1.0.0+).
// Sorted by memory updated_at DESC (most recent first).
// Empty for legacy nodes (created via Add/Cognify without provenance).
MemoryIDs []string
}
SearchResult represents a single search result with scoring metadata.
type SearchType ¶
type SearchType string
SearchType specifies the type of search to perform.
const ( // SearchTypeVector performs vector similarity search only. SearchTypeVector SearchType = "vector" // SearchTypeGraph performs graph traversal search only (requires seed nodes). SearchTypeGraph SearchType = "graph" // SearchTypeHybrid combines vector similarity and graph traversal. SearchTypeHybrid SearchType = "hybrid" )
type Searcher ¶
type Searcher interface {
// Search performs a search based on the query and options.
// For vector/hybrid search, query is the text to embed and search.
// For graph search, query is ignored (uses opts.SeedNodeIDs instead).
Search(ctx context.Context, query string, opts SearchOptions) ([]SearchResult, error)
}
Searcher defines the interface for knowledge graph search.
type VectorSearcher ¶
type VectorSearcher struct {
// contains filtered or unexported fields
}
VectorSearcher performs vector similarity search.
func NewVectorSearcher ¶
func NewVectorSearcher( embClient embeddings.EmbeddingClient, vectorStore store.VectorStore, graphStore store.GraphStore, ) *VectorSearcher
NewVectorSearcher creates a new vector searcher.
func (*VectorSearcher) Search ¶
func (v *VectorSearcher) Search(ctx context.Context, query string, opts SearchOptions) ([]SearchResult, error)
Search performs vector similarity search. It embeds the query, searches the vector store, and enriches results with full node data.
Source Files