Python Sdk for Milvus
Project description
Milvus Python SDK -- pymilvus
Using Milvus python sdk for Milvus Download
Pymilvus only supports python >= 3.5
, is fully tested under 3.5, 3.6, 3.7.
Pymilvus can be downloaded via pip
or pip3
for python3
$ pip install pymilvus
Different versions of Milvus and lowest/highest pymilvus version supported accordingly
Milvus version | Recommended pymilvus version |
---|---|
0.3.0 | 0.1.13 |
0.3.1 | 0.1.25 |
0.4.0 | 0.2.2 |
0.5.0 | 0.2.3 |
0.5.1 | 0.2.3 |
0.5.2 | 0.2.3 |
0.5.3 | 0.2.5 |
You can download a specific version by:
$ pip install pymilvus==0.2.5
If you want to upgrade pymilvus
to newest version
$ pip install --upgrade pymilvus
Import
from milvus import Milvus, IndexType, MetricType, Status
Getting started
Initial a Milvus
instance and connect
to the sever
>>> milvus = Milvus()
>>> milvus.connect(host='localhost', port='19530')
Status(code=0, message='Successfully connected! localhost:19530')
Once successfully connected, you can get the version of server
>>> milvus.server_version()
(Status(code=0, message='Success'), '0.5.0') # this is example version, the real version may vary
Add a new table
First set param
>>> dim = 32 # Dimension of the vector
>>> param = {'table_name':'test01', 'dimension':dim, 'index_file_size':1024, 'metric_type':MetricType.L2}
Then create table
>>> milvus.create_table(param)
Status(code=0, message='Create table successfully!')
Describe the table we just created
>>> milvus.describe_table('test01')
(Status(code=0, message='Describe table successfully!'), TableSchema(table_name='test01', dimension=32, index_file_size=1024, metric_type=<MetricType: L2>))
Add vectors into table test01
First create 20 vectors of 256-dimension.
- Note that
random
andpprint
we used here is for creating fake vectors data and pretty print, you may not need them in your project
>>> import random
>>> from pprint import pprint
# Initialize 20 vectors of 256-dimension
>>> vectors = [[random.random() for _ in range(dim)] for _ in range(20)]
Then add vectors into table test01
>>> status, ids = milvus.insert(table_name='test01', records=vectors)
>>> print(status)
Status(code=0, message='Add vectors successfully!')
>>> pprint(ids) # List of ids returned
[1571123848227800000,
1571123848227800001,
...........
1571123848227800018,
1571123848227800019]
You can also specify vectors id
>>> vector_ids = [i for i in range(20)]
>>> status, ids = milvus.insert(table_name='test01', records=vectors, ids=vector_ids)
>>> pprint(ids)
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19]
Get vectors num
>>> milvus.count_table('test01')
(Status(code=0, message='Success!'), 20)
Load vectors into memory
>>> milvus.preload_table('test01')
Status(code=0, message='')
Create index
>>> index_param = {'index_type': IndexType.FLAT, 'nlist': 128}
>>> milvus.create_index('test01', index_param)
Status(code=0, message='Build index successfully!')
Then show index information
>>> milvus.describe_index('test01')
(Status(code=0, message='Successfully'), IndexParam(_table_name='test01', _index_type=<IndexType: FLAT>, _nlist=128))
Search vectors
# create 5 vectors of 256-dimension
>>> q_records = [[random.random() for _ in range(dim)] for _ in range(5)]
Then get results
>>> status, results = milvus.search(table_name='test01', query_records=q_records, top_k=1, nprobe=16)
>>> print(status)
Status(code=0, message='Search vectors successfully!')
>>> pprint(results) # Searched top_k vectors
[
[(id:15, distance:2.855304718017578),
(id:16, distance:3.040700674057007)],
[(id:11, distance:3.673950433731079),
(id:15, distance:4.183730602264404)],
........
[(id:6, distance:4.065953254699707),
(id:1, distance:4.149323463439941)]
]
Drop index
>>> milvus.drop_index('test01')
Status(code=0, message='')
Delete the table we just created
>>> milvus.drop_table(table_name='test01')
Status(code=0, message='Delete table successfully!')
Disconnect with the server
>>> milvus.disconnect()
Status(code=0, message='Disconnect successfully')
Example python
There are some small examples in examples/
, you can find more guide there.
You can find api doc in API Doc
If you encounter any problems or bugs, please open new issues
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for pymilvus_test-0.2.36-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | bcc94a538b631bce19a6fee67ba0f75e4596ffdaf641f63d0832d3537d8d3994 |
|
MD5 | 884977f673ad4e0d07f3bfc7d01a03a7 |
|
BLAKE2b-256 | a84201dac376bb369a1c6582c3d7be48228186e5ac2de9efe5ac6708b5a6edee |