HNSW Index

High-performance approximate nearest neighbor index

Public fields

dimension

Vector dimension

metric

Distance metric

n_trees

Number of trees (for Annoy)

search_k

Search parameter

Methods

Public methods

• HNSWIndex$new()

• HNSWIndex$add_items()

• HNSWIndex$build()

• HNSWIndex$search()

• HNSWIndex$get_vector()

• HNSWIndex$get_ids()

• HNSWIndex$size()

• HNSWIndex$remove_items()

• HNSWIndex$clear()

• HNSWIndex$save()

• HNSWIndex$load()

• HNSWIndex$clone()

---

Method new()

Create a new HNSWIndex

Usage

HNSWIndex$new(dimension, metric = "angular", n_trees = 50, search_k = -1)

Arguments

dimension

Vector dimension

metric

Distance metric: "angular", "euclidean", "manhattan", "dot"

n_trees

Number of trees for index (higher = more accuracy)

search_k

Search parameter (higher = more accuracy, -1 = auto)

---

Method add_items()

Add items to the index

Usage

HNSWIndex$add_items(ids, vectors)

Arguments

ids

Character vector of IDs

vectors

Matrix of vectors (rows = items)

Returns

Self

---

Method build()

Build the index (required before searching)

Usage

HNSWIndex$build()

Returns

Self

---

Method search()

Search for nearest neighbors

Usage

HNSWIndex$search(query, k = 10, include_distances = TRUE)

Arguments

query

Query vector

k

Number of neighbors

include_distances

Return distances

Returns

Data frame with id, distance columns

---

Method get_vector()

Get vector by ID

Usage

HNSWIndex$get_vector(id)

Arguments

id

Item ID

Returns

Vector or NULL

---

Method get_ids()

Get all IDs

Usage

HNSWIndex$get_ids()

Returns

Character vector

---

Method size()

Get item count

Usage

HNSWIndex$size()

Returns

Integer

---

Method remove_items()

Remove items from index

Usage

HNSWIndex$remove_items(ids)

Arguments

ids

IDs to remove

Returns

Self

---

Method clear()

Clear the index

Usage

HNSWIndex$clear()

Returns

Self

---

Method save()

Save index to file

Usage

HNSWIndex$save(path)

Arguments

path

File path

---

Method load()

Load index from file

Usage

HNSWIndex$load(path)

Arguments

path

File path

Returns

Self

---

Method clone()

The objects of this class are cloneable with this method.

Usage

HNSWIndex$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

if (FALSE) { # \dontrun{
# Create index
index <- HNSWIndex$new(dimension = 128, metric = "angular")

# Add vectors
index$add_items(ids = c("a", "b", "c"),
                vectors = matrix(rnorm(384), nrow = 3))

# Search
results <- index$search(query = rnorm(128), k = 5)
} # }