README
¶
🧬🔍🗄️ Vecgo
Vecgo is a Go library designed for efficient vector indexing and searching, supporting various index types and emphasizing approximate nearest neighbor search. It provides a versatile and user-friendly interface for managing and querying vast collections of high-dimensional vectors.
⚠ This is experimental and subject to breaking changes.
Features
- Support for multiple index types, including flat and HNSW (Hierarchical Navigable Small World) algorithms.
- Customizable options for memory usage and search performance.
- Efficient handling of high-dimensional vectors.
- Embeddable vector store for seamless integration with Go applications.
Usage
Here's a basic example demonstrating how to perform nearest neighbor search with Vecgo:
package main
import (
"fmt"
"log"
"github.com/hupe1980/vecgo"
)
func main() {
vg := vecgo.NewHNSW[string]()
// vg := vecgo.NewFlat[string]()
_, err := vg.Insert(vecgo.VectorWithData[string]{
Vector: []float32{1.0, 2.0, 2.5},
Data: "Hello World!",
})
if err != nil {
log.Fatal(err)
}
k := 5
result, err := vg.KNNSearch([]float32{1.0, 2.0, 2.5}, k)
if err != nil {
log.Fatal(err)
}
fmt.Println(result[0].Data)
}
For more information on each method, please refer to the GoDoc documentation.
Contributing
Contributions to Vecgo are welcome! Feel free to open issues for bug reports, feature requests, or submit pull requests with improvements.
License
Vecgo is licensed under the MIT License. See the LICENSE file for details.
Documentation
¶
Overview ¶
Package vecgo provides functionalities for an embedded vector store database.
Index ¶
- Variables
- type BruteSearchOptions
- type FilterFunc
- type KNNSearchOptions
- type SearchResult
- type Vecgo
- func (vg *Vecgo[T]) BruteSearch(query []float32, k int, optFns ...func(o *BruteSearchOptions)) ([]SearchResult[T], error)
- func (vg *Vecgo[T]) Get(id uint32) (T, error)
- func (vg *Vecgo[T]) Insert(item VectorWithData[T]) (uint32, error)
- func (vg *Vecgo[T]) KNNSearch(query []float32, k int, optFns ...func(o *KNNSearchOptions)) ([]SearchResult[T], error)
- func (vg *Vecgo[T]) PrintStats()
- func (vg *Vecgo[T]) SaveToFile(filename string) error
- func (vg *Vecgo[T]) SaveToWriter(w io.Writer) error
- type VectorWithData
Constants ¶
This section is empty.
Variables ¶
var ( // ErrNotFound is returned when an item is not found. ErrNotFound = errors.New("not found") // ErrInvalidEFValue is returned when the explore factor (ef) is less than the value of k. ErrInvalidEFValue = errors.New("explore factor (ef) must be at least the value of k") )
Functions ¶
This section is empty.
Types ¶
type BruteSearchOptions ¶ added in v0.0.5
type BruteSearchOptions struct {
// FilterFunc is a function used to filter search results.
FilterFunc FilterFunc
}
BruteSearchOptions contains options for brute-force search.
type FilterFunc ¶ added in v0.0.5
FilterFunc is a function type used for filtering search results.
type KNNSearchOptions ¶
type KNNSearchOptions struct {
// EF (Explore Factor) specifies the size of the dynamic list for the nearest neighbors during the search.
// Higher EF leads to more accurate but slower search.
// EF cannot be set lower than the number of queried nearest neighbors (k).
// The value of EF can be anything between k and the size of the dataset.
EF int
// FilterFunc is a function used to filter search results.
FilterFunc FilterFunc
}
KNNSearchOptions contains options for KNN search.
type SearchResult ¶
type SearchResult[T any] struct { index.SearchResult // Data is the associated data of the search result. Data T }
SearchResult represents a search result.
type Vecgo ¶
type Vecgo[T any] struct { // contains filtered or unexported fields }
Vecgo is a vector store database.
func NewFromFilename ¶
NewFromFilename creates a new Vecgo instance from a file.
func NewFromReader ¶
NewFromReader creates a new Vecgo instance from an io.Reader.
func (*Vecgo[T]) BruteSearch ¶
func (vg *Vecgo[T]) BruteSearch(query []float32, k int, optFns ...func(o *BruteSearchOptions)) ([]SearchResult[T], error)
BruteSearch performs a brute-force search.
func (*Vecgo[T]) Insert ¶
func (vg *Vecgo[T]) Insert(item VectorWithData[T]) (uint32, error)
Insert inserts a vector along with associated data into the database.
func (*Vecgo[T]) KNNSearch ¶
func (vg *Vecgo[T]) KNNSearch(query []float32, k int, optFns ...func(o *KNNSearchOptions)) ([]SearchResult[T], error)
KNNSearch performs a K-nearest neighbor search.
func (*Vecgo[T]) PrintStats ¶
func (vg *Vecgo[T]) PrintStats()
PrintStats prints statistics about the database.
func (*Vecgo[T]) SaveToFile ¶
SaveToFile saves the Vecgo database to a file.
type VectorWithData ¶
VectorWithData represents a vector along with associated data.
Source Files
Directories
¶
| Path | Synopsis |
|---|---|
|
_examples
|
|
|
Package index provides interfaces and types for vector search indexes.
|
Package index provides interfaces and types for vector search indexes. |
|
flat
Package flat provides an implementation of a flat index for vector storage and search.
|
Package flat provides an implementation of a flat index for vector storage and search. |
|
hnsw
Package hnsw implements the Hierarchical Navigable Small World (HNSW) graph for approximate nearest neighbor search.
|
Package hnsw implements the Hierarchical Navigable Small World (HNSW) graph for approximate nearest neighbor search. |
|
internal
|
|