核心流程

本页面提供一个向量数据里 VikingDB 通过 Python SDK 创建数据集、写入数据、创建索引和检索查询的完整请求示例。

# 写给用户的样例
fields = [
    Field(
        field_name="doc_id",
        field_type=FieldType.String,
        is_primary_key=True
    ),
    Field(
        field_name="text_vector",
        field_type=FieldType.Vector,
        dim=10
    ),
    Field(
        field_name="like",
        field_type=FieldType.Int64,
        default_val=0
    ),
    Field(
        field_name="price",
        field_type=FieldType.Float32,
        default_val=0
    ),
    Field(
        field_name="author",
        field_type=FieldType.List_String,
        default_val=[]
    ),
    Field(
        field_name="aim",
        field_type=FieldType.Bool,
        default_val=True
    ),
]
res = vikingdb_service.create_collection("example", fields, "This is an example")
# 返回一个collection实例
print(res)

res = vikingdb_service.get_collection("example")
# 返回一个collection实例
print(res)

vikingdb_service.drop_collection("example")  # 无返回

res = vikingdb_service.list_collections()
# 返回一个列表
print(res)

vector_index = VectorIndexParams(distance=DistanceType.COSINE,index_type=IndexType.HNSW,
                                 quant=QuantType.Float)
res = vikingdb_service.create_index("example","example_index", vector_index, cpu_quota=2,
                                    description="This is an index", partition_by="like", scalar_index=None)
# 返回一个index实例
print(res)

res = vikingdb_service.get_index("example", "example_index")
# 返回一个index实例
print(res.description)

vikingdb_service.drop_index("example", "example_index")  # 无返回

res = vikingdb_service.list_indexes("example")
# 返回一个列表
print(res)

def gen_random_vector(dim):
    res = [0, ] * dim
    for i in range(dim):
        res[i] = random.random() - 0.5
    return res
collection = vikingdb_service.get_collection("example")
field1 = {"doc_id": "11", "text_vector": gen_random_vector(10), "like": 1, "price": 1.11,
          "author": ["gy"], "aim": True}
field2 = {"doc_id": "22", "text_vector": gen_random_vector(10), "like": 2, "price": 2.22,
          "author": ["gy", "xjq"], "aim": False}
field3 = {"doc_id": "33", "text_vector": gen_random_vector(10), "like": 1, "price": 3.33,
          "author": ["gy", "xjq"], "aim": False}
field4 = {"doc_id": "44", "text_vector": gen_random_vector(10), "like": 1, "price": 4.44,
          "author": ["gy", "xjq"], "aim": False}
data1 = Data(field1)
data2 = Data(field2)
data3 = Data(field3)
data4 = Data(field4)
datas = []
datas.append(data1)
datas.append(data2)
datas.append(data3)
datas.append(data4)
collection.upsert_data(datas)  # 无返回

collection = vikingdb_service.get_collection("example")
res = collection.fetch_data(["11", "22", "33", "44"])
# 返回一个列表
for item in res:
    print(item)
    print(item.fields)

collection = vikingdb_service.get_collection("example")
collection.delete_data("11")  # 无返回

index = vikingdb_service.get_index("example", "example_index")
res = index.fetch_data(["11", "33"], partition="1", output_fields=["doc_id", "like"])
# 返回一个列表
for item in res:
    print(item)
    print(item.fields)

index = vikingdb_service.get_index("example", "example_index")
res = index.search_by_id("11", limit=2, output_fields=["doc_id", "like", "text_vector"], partition="1")
# 返回一个列表
for item in res:
    print(item)
    print(item.fields)

index = vikingdb_service.get_index("example", "example_index")
def gen_random_vector(dim):
    res = [0, ] * dim
    for i in range(dim):
        res[i] = random.random() - 0.5
    return res
res = index.search_by_vector(gen_random_vector(10), limit=2, output_fields=["doc_id", "like", "text_vector"],
                             partition="1")
# 返回一个列表
for item in res:
    print(item)
    print(item.fields)

index = vikingdb_service.get_index("example", "example_index")
def gen_random_vector(dim):
    res = [0, ] * dim
    for i in range(dim):
        res[i] = random.random() - 0.5
    return res
res = index.search(order=VectorOrder(gen_random_vector(10)), limit=2,
                   output_fields=["doc_id", "like", "text_vector"],
                   partition="1", filter={"op": "range", "field": "price", "lt": 3.5})
# 返回一个列表
for item in res:
    print(item)
    print(item.fields)
res = index.search(order=ScalarOrder("price", Order.Desc), limit=6,
                   output_fields=["price"],
                   partition="1",
                   filter={"op": "range", "field": "price", "lt": 5})
# 返回一个列表
for item in res:
    print(item)
    print(item.fields)

# 含有text字段的测试
fields = [
    Field(
        field_name="doc_id",
        field_type=FieldType.String,
        is_primary_key=True
    ),
    Field(
        field_name="text",
        field_type=FieldType.Text,
        pipeline_name="text_split_bge_large_zh"
    ),
    Field(
        field_name="like",
        field_type=FieldType.Int64,
        default_val=0
    ),
    Field(
        field_name="price",
        field_type=FieldType.Float32,
        default_val=0
    ),
    Field(
        field_name="author",
        field_type=FieldType.List_String,
        default_val=[]
    ),
    Field(
        field_name="aim",
        field_type=FieldType.Bool,
        default_val=True
    ),
]
res = vikingdb_service.create_collection("example_text", fields, "This is an example include text")

vector_index = VectorIndexParams(distance=DistanceType.COSINE, index_type=IndexType.HNSW,
                                 quant=QuantType.Float)
res = vikingdb_service.create_index("example_text", "example_index_text", vector_index, cpu_quota=2,
                                    description="This is an index include text", partition_by="like",
                                    scalar_index=None)

collection = vikingdb_service.get_collection("example_text")
field1 = {"doc_id": "11", "text": {"text":"this is one"}, "like": 1, "price": 1.11,
          "author": ["gy"], "aim": True}
field2 = {"doc_id": "22", "text": {"text":"this is two"}, "like": 2, "price": 2.22,
          "author": ["gy", "xjq"], "aim": False}
field3 = {"doc_id": "33", "text": {"text":"this is three"}, "like": 1, "price": 3.33,
          "author": ["gy", "xjq"], "aim": False}
field4 = {"doc_id": "44", "text": {"text":"this is four"}, "like": 1, "price": 4.44,
          "author": ["gy", "xjq"], "aim": False}
data1 = Data(field1)
data2 = Data(field2)
data3 = Data(field3)
data4 = Data(field4)
datas = []
datas.append(data1)
datas.append(data2)
datas.append(data3)
datas.append(data4)
collection.upsert_data(datas)  # 无返回

index = vikingdb_service.get_index("example_text", "example_index_text")
res = index.search_by_text(Text(text="this is five"), filter={"op": "range", "field": "price", "lt": 4},
                           limit=3, output_fields=["doc_id", "text", "price", "like"], partition=1)
for item in res:
    print(item)
    print(item.fields)
list = [RawData("text","hello1"), RawData("text","hello2")]
res = vikingdb_service.embedding(EmbModel("bge_large_zh"), list)
print(res)
for item in res:
    print(item)