Milvus Python SDK -- pymilvus

Using Milvus python sdk for Milvus Download

Pymilvus only supports python >= 3.4, is fully tested under 3.4, 3.5, 3.6, 3.7.

Pymilvus can be downloaded via pip. If no use, try pip3

$ pip install pymilvus

Different versions of Milvus and lowest/highest pymilvus version supported accordingly

Milvus version	Lowest pymilvus version supported	Highest pymivus version supported
0.3.0	-	0.1.13
0.3.1	0.1.14	0.1.25
0.4.0	0.2.0	-

You can download a specific version by:

$ pip install pymilvus==0.2.0

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='SERVER-HOST', port='SERVER-PORT')
Status(code=0, message='Successfully connected!')

Once successfully connected, you can get the version of server

>>> milvus.server_version()
(Status(code=0, message='Success'), 0.4.0)  # this is example version, the real version may vary

Add a new table

First set param

>>> param = {'table_name':'test01', 'dimension':256, '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=256, index_file_size=1024, metric_type=<MetricType: L2>))

Add vectors into table test01

First create 20 vectors of 256-dimension.

Note that random and pprint 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

>>> dim = 256  # Dimension of the vector

# 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.add_vectors(table_name='test01', records=vectors)
>>> print(status)
Status(code=0, message='Success')
>>> pprint(ids) # List of ids returned
23455321135511233
12245748929023489
...

You can also specify vectors id

>>> vector_ids = [i for i in range(20)]
>>> status, ids = milvus.add_vectors(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.get_table_row_count('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.IVFLAT, 'nlist': 16384}
>>> milvus.create_index('test01', index_param)
Status(code=0, message='Build index successfully!')

Then show index information

>>> client.describe_index('test01')
(Status(code=0, message='Successfully'), IndexParam(_table_name='test01', _index_type=<IndexType: IVFLAT>, _nlist=16384))

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_vectors(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
[
[QueryResult(id=0, distance=34.85963439941406)],
[QueryResult(id=0, distance=36.73900604248047)],
[QueryResult(id=0, distance=34.35655975341797)],
[QueryResult(id=18, distance=36.19701385498047)],
[QueryResult(id=5, distance=39.11549758911133)]
]

Drop index

>>> milvus.drop_index('test01')
Status(code=0, message='')

Delete vectors by date range

>>> milvus.delete_vectors_by_range('test01', '2019-06-01', '2020-01-01')
Status(code=0, message='')

Delete the table we just created

>>> milvus.delete_table(table_name='test01')
Status(code=0, message='Success')

Disconnect with the server

>>> milvus.disconnect()
Status(code=0, message='Success')

Example python

There are some small examples in examples/, you can find more guide there.

Build docs

$ sphinx-build -b html doc/en/ doc/en/build

If you encounter any problems or bugs, please open new issues

Metadata

Release files for pymilvus 0.2.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for pymilvus 0.2.0
File	Size	Uploaded
pymilvus-0.2.0.tar.gz	22.3 kB	Sep 12, 2019	Details

Built distribution (wheel)

Table of built distributions (wheels) for pymilvus 0.2.0
File	Interpreter	ABI	Platform	Reset
pymilvus-0.2.0-py3-none-any.whl 28.7 kB Sep 12, 2019	Python 3	none	any	Details

Total release size: 51.0 kB

Release files / pymilvus-0.2.0.tar.gz

Download URL	pymilvus-0.2.0.tar.gz
Size	22.3 kB
Tags	Source
SHA-256 checksum How to use checksums	`dbd094cc4a87db44739adac7b66366248d166d8f9e881d73fa51919a10fc65bf`
BLAKE2b-256 checksum How to use checksums	`357fe7c934e6d9e89e7f8e1552c21ff1cdde741a1a1047acf585f7b74f31b499`
Upload date	Sep 12, 2019
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/1.13.0 pkginfo/1.5.0.1 requests/2.22.0 setuptools/41.2.0 requests-toolbelt/0.9.1 tqdm/4.34.0 CPython/3.6.9`

Release files / pymilvus-0.2.0-py3-none-any.whl

Download URL	pymilvus-0.2.0-py3-none-any.whl
Size	28.7 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`2362d011e33322b1029dea870b8441a94a8860d4277bb1a7ad5de9fa8b3ba1bf`
BLAKE2b-256 checksum How to use checksums	`dcc1288e3bc7481e969cf71c8a53fca6c6485439aea2c31517385fc0e84f9b4d`
Upload date	Sep 12, 2019
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/1.13.0 pkginfo/1.5.0.1 requests/2.22.0 setuptools/41.2.0 requests-toolbelt/0.9.1 tqdm/4.34.0 CPython/3.6.9`