| id | title | slug |
|---|---|---|
README |
README |
/readme |
shiva is a library for Simple High dimensional Indexed Vector search Algorithms.
Basic guiding principle is to be:
- Simple (non-distributed, single threaded indexing, easy to use)
- Support high dimensional vectors, optimize memory for speed
- Support many distance metrics
- Scale out to different indices and algorithms
To use Shiva, add the following to your build.sbt
For release versions:
resolvers +=
"Sonatype OSS Releases" at "https://s01.oss.sonatype.org/content/repositories/releases"
libraryDependencies ++= Seq(
"io.github.saucam" %% "shiva-core" % "<version>"
)For snapshot versions:
resolvers +=
"Sonatype OSS Snapshots" at "https://s01.oss.sonatype.org/content/repositories/snapshots"
libraryDependencies ++= Seq(
"io.github.saucam" %% "shiva-core" % "<version>"
)The following gives a simple example on how to use the hnsw index in the library after adding the dependency:
val index = HnswIndexBuilder[Int, Double, IntDoubleIndexItem](
dimensions = 3,
maxItemCount = 1000000,
m = 32,
distanceCalculator = new EuclideanDistanceDouble
).build()
val item1 = IntDoubleIndexItem(1, Vector(4.05d, 1.06d, 7.8d))
val item2 = IntDoubleIndexItem(2, Vector(8.01d, 2.06d, 1.8d))
val item3 = IntDoubleIndexItem(3, Vector(9.34d, 3.06d, 3.1d))
index.add(item1)
index.add(item2)
index.add(item3)
val results = index.findKSimilarItems(item1.id, 10)
results.foreach(println())Currently supported distance metrics are:
- Inner Product
- Euclidean Distance
- Cosine Distance
- Manhattan Distance
- Minkowski Distance
See the contributor's guide
