全部文档
当前文档

暂无内容

如果没有找到您期望的内容,请尝试其他搜索词

文档中心

管理向量(Python)

最近更新时间:2026-11-18 16:33:48

管理向量(Python)

写入向量

以下代码用于向索引写入向量数据:

from ks3.vectors import VectorsClient
from ks3.xmlParsers.vectorData import InputVector, VectorData

client = VectorsClient('<YOUR_ACCESS_KEY>', '<YOUR_SECRET_KEY>',
                       host='ks3vectors-cn-beijing.ksyuncs.com', is_secure=True)

vectors = [
    InputVector(key='vec1', data=VectorData(float32=[0.1] * 128), metadata={'tag1': 'a'}),
    InputVector(key='vec2', data=VectorData(float32=[0.2] * 128), metadata={'tag2': 'b'}),
]
ret = client.put_vectors(vectors,
                         vector_bucket_name='<YOUR_VECTOR_BUCKET_NAME>',
                         index_name='<YOUR_INDEX_NAME>')
print('put_vectors, request_id:', ret.response_metadata.request_id)

获取向量

以下代码用于按 key 批量读取向量:

from ks3.vectors import VectorsClient

client = VectorsClient('<YOUR_ACCESS_KEY>', '<YOUR_SECRET_KEY>',
                       host='ks3vectors-cn-beijing.ksyuncs.com', is_secure=True)

ret = client.get_vectors(keys=['vec1', 'vec2'],
                         vector_bucket_name='<YOUR_VECTOR_BUCKET_NAME>',
                         index_name='<YOUR_INDEX_NAME>',
                         return_data=True, return_metadata=True)
for v in ret:
    print(v.key, v.data.float32[:3] if v.data else None, v.metadata)
print('request_id:', ret.response_metadata.request_id)

列举向量

以下代码用于分页列举索引中的向量:

from ks3.vectors import VectorsClient

client = VectorsClient('<YOUR_ACCESS_KEY>', '<YOUR_SECRET_KEY>',
                       host='ks3vectors-cn-beijing.ksyuncs.com', is_secure=True)

ret = client.list_vectors(vector_bucket_name='<YOUR_VECTOR_BUCKET_NAME>',
                          index_name='<YOUR_INDEX_NAME>',
                          max_results=100,
                          return_data=True, return_metadata=True)
for v in ret:
    print(v.key, v.data.float32[:3] if v.data else None, v.metadata)
print('next_token:', ret.next_token, ', request_id:', ret.response_metadata.request_id)

分段列举向量

以下代码用于将索引数据划分为 N 段并行拉取:

from ks3.vectors import VectorsClient

client = VectorsClient('<YOUR_ACCESS_KEY>', '<YOUR_SECRET_KEY>',
                       host='ks3vectors-cn-beijing.ksyuncs.com', is_secure=True)

segment_count = 4
for segment_index in range(segment_count):
    ret = client.list_vectors(vector_bucket_name='<YOUR_VECTOR_BUCKET_NAME>',
                              index_name='<YOUR_INDEX_NAME>',
                              max_results=100,
                              segment_count=segment_count,
                              segment_index=segment_index)
    print('segment %d/%d, count: %d' % (segment_index, segment_count, len(ret)))

相似性检索向量

以下代码用于按查询向量检索最近邻:

from ks3.vectors import VectorsClient
from ks3.xmlParsers.vectorData import VectorData

client = VectorsClient('<YOUR_ACCESS_KEY>', '<YOUR_SECRET_KEY>',
                       host='ks3vectors-cn-beijing.ksyuncs.com', is_secure=True)

ret = client.query_vectors(
    query_vector=VectorData(float32=[0.15] * 128),
    top_k=10,
    vector_bucket_name='<YOUR_VECTOR_BUCKET_NAME>',
    index_name='<YOUR_INDEX_NAME>',
    filter={'eq': {'tag': 'a'}},
    return_data=True, return_metadata=True, return_distance=True)
for v in ret:
    print(v.key, 'distance=', v.distance, v.metadata)
print('request_id:', ret.response_metadata.request_id)

删除向量

以下代码用于按 key 列表删除向量:

from ks3.vectors import VectorsClient

client = VectorsClient('<YOUR_ACCESS_KEY>', '<YOUR_SECRET_KEY>',
                       host='ks3vectors-cn-beijing.ksyuncs.com', is_secure=True)

ret = client.delete_vectors(keys=['vec1', 'vec2'],
                            vector_bucket_name='<YOUR_VECTOR_BUCKET_NAME>',
                            index_name='<YOUR_INDEX_NAME>')
print('delete_vectors, request_id:', ret.response_metadata.request_id)

文档导读
纯净模式常规模式

纯净模式

点击可全屏预览文档内容
文档反馈