inmemory

package
v0.60.0 Latest Latest
Warning

This package is not in the latest version of its module.

Go to latest
Published: Nov 4, 2026 License: Apache-2.0 Imports: 12 Imported by: 0

Documentation

Overview

Package inmemory implements exact vector search over process-local records. Upsert and Query accept precomputed coordinates in the configured space. Query evaluates metadata before scoring, scans all records under a read lock, and orders score ties by document ID. The Store owns no model or I/O.

Example
store, err := inmemory.NewStore(context.Background(), inmemory.StoreConfig{Space: testSpace, Dimensions: 4})
if err != nil {
	panic(err)
}
fmt.Println(store.Len())
Output:
0

Index

Examples

Constants

View Source
const Provider = "InMemory"

Variables

This section is empty.

Functions

func CosineSimilarity

func CosineSimilarity(left, right []float64) vectorstore.Score

CosineSimilarity is the default for StoreConfig.Similarity — cos(θ) mapped into [0, 1] via (1 + cos) / 2. Returns 0.5 (the "no information" midpoint) when either vector has zero magnitude rather than NaN.

func DotProductSimilarity

func DotProductSimilarity(left, right []float64) vectorstore.Score

DotProductSimilarity maps the unbounded inner product monotonically into the common score range when vector magnitude is meaningful. Products and their sum retain their exact values until the final score conversion.

func EuclideanSimilarity

func EuclideanSimilarity(left, right []float64) vectorstore.Score

EuclideanSimilarity maps distance d to 1/(1+d), preserving magnitude differences.

Types

type Similarity

type Similarity func(left, right []float64) vectorstore.Score

Similarity scores two equal-length vectors; higher means more similar. Implementations must be deterministic and symmetric: Similarity(a, b) == Similarity(b, a). Returning vectorstore.Score keeps custom strategies inside the same normalized contract as every provider.

Both slices are borrowed read-only for the duration of the call: an implementation must not modify or retain them. The store scores a record against the shared indexed vector, so an in-place write would corrupt the index and race concurrent searches. Honoring this, Similarity may be called concurrently, which is how Store.Query scans the index.

type Store

type Store struct {
	// contains filtered or unexported fields
}

Store is a concurrency-safe, in-process vector store. Upsert snapshots documents and coordinates and Query returns detached results in deterministic score order. Query scans the full index; all records are lost when the process exits.

func NewStore

func NewStore(_ context.Context, config StoreConfig) (*Store, error)

func (*Store) CheckDocuments added in v0.59.0

func (s *Store) CheckDocuments(ctx context.Context, space vectorstore.SpaceID, docs []*document.Document) error

CheckDocuments rejects an invalid space or document set before an index writer spends an embedding call. The memory store supports complete media.

func (*Store) CheckQuery added in v0.59.0

func (s *Store) CheckQuery(ctx context.Context, space vectorstore.SpaceID, opts vectorstore.SearchOptions) error

func (*Store) Clear

func (s *Store) Clear()

Clear removes all records while retaining the bound coordinate space.

func (*Store) DeleteIDs

func (s *Store) DeleteIDs(ctx context.Context, ids []string) error

func (*Store) DeleteWhere

func (s *Store) DeleteWhere(ctx context.Context, expr filter.Predicate) error

func (*Store) Len

func (s *Store) Len() int

func (*Store) Query added in v0.59.0

func (*Store) Upsert added in v0.59.0

func (s *Store) Upsert(ctx context.Context, batch *vectorstore.Batch) error

type StoreConfig

type StoreConfig struct {
	Space      vectorstore.SpaceID
	Dimensions int

	Similarity Similarity
}

StoreConfig binds one immutable coordinate space and dimension. A nil Similarity selects CosineSimilarity.

func (StoreConfig) Validate

func (s StoreConfig) Validate() error

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL