qmilvus provides the simplest way to use milvus.
- auto build schema
- auto build index
- auto insert collection entity
- auto search
So fantastic as if milvus is transparent. Access milvus service can be so easy!
- define you schema like this:
package qmilvus
import (
"context"
"github.com/milvus-io/milvus-sdk-go/v2/entity"
milvus "github.com/yangkequn/q-milvus-driver-for-go"
)
type FooEntity struct {
Id int64 `schema:"in,out" primarykey:"true"`
Name string `schema:""`
Detail string `schema:""`
Vector []float32 `dim:"384" schema:"in" index:"true"`
Score float32 ``
}
//Index: define your index field name and type here. required
func (v FooEntity) Index() (indexFieldName string, index entity.Index) {
index, _ = entity.NewIndexIvfFlat(entity.IP, 256)
return "Vector", index
}
//BuildSearchVector: Calculate vector here; if precalculated, Just return it
func (v FooEntity) BuildSearchVector(ctx context.Context) (Vector []float32) {
//return Foo.CalculateVector(ctx, v.Detail)
return v.Vector
}
var FooCollection *milvus.Collection = milvus.NewCollection("milvus.vm:19530", FooEntity{}, "partitionName")
- using FooCollection, you can do the the left things easily:
- insert entities
err:=FooCollection.Insert(c context.Context, bar []*Bar)
here struct Bar will cast （borrow from c++ cast ） into struct FooEntity. structs fields with same name and type will be copied, Other Bar fields will be neglected.
The casted FooEntity will insert into milvus Collection according to go Tag.
Id int64
schema:"in,out" primarykey:"true", means Id Field will insert to collection, and will returned in searchField.
- search
ids,scores,err:=FooCollection.Search(ctx context.Context, query []float32)
- Remove entity
err:=FooCollection.RemoveByKey(c context.Context, bar []*Bar)