pymilvus 0.2.7

Python Sdk for Milvus

Milvus Python SDK -- pymilvus

If you want to contribute to this repo, please read our contribution guidelines.

You can find api doc in Reference Doc

Using Milvus python sdk for Milvus Download

New features

• Add new metric type HAMMING, JACCARD, TANIMOTO for binary vectors. examples about binary vectors in examples/example_binary.py

---

Pymilvus only supports python >= 3.5, is fully tested under 3.5, 3.6, 3.7, 3.8.

Pymilvus can be downloaded via pip or pip3 for python3

$ pip install pymilvus

Different versions of Milvus and recommended 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
0.6.0	0.2.6

You can download a specific version by:

$ pip install pymilvus==0.2.7

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

Create table

>>> 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'), '0.6.0')  # 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>))

---

Insert vectors

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

# 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

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 32-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=8)
>>> 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)]
]

---

Partition operations

Create table named demo01

>>> param = {'table_name':'demo01', 'dimension':dim, 'index_file_size':1024, 'metric_type':MetricType.L2}
>>> milvus.create_table(param)

Create a new partition named partition01 under table demo01, and specify tag tag01

>>> milvus.create_partition('demo01', 'partition01', 'tag01')
Status(code=0, message='OK')

Specify partition vectors insert into

>>> status = milvus.insert('demo01', vectors, partition_tag="tag01")
>>> status
(Status(code=0, message='Add vectors successfully!')

Show partitions

milvus.show_partitions(table_name='demo01')

Search vectors in a designated partition

>>> milvus.search(table_name='test01', query_records=q_records, top_k=1, nprobe=8, partition_tags=['tag01'])

When you not specify partition_tags, milvus will search in whole table.

Drop operations

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.

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