milvus-api 0.2.1.2

A Milvus API used for quickly setting up a Milvus vector database

milvus_api

milvus_api is a Python library that primarily provides functionalities to manage databases, including creating collections, inserting data, searching and updating index.

Installation

This project depends on several Python libraries. Use the following command to install the necessary libraries:

pip install pymilvus==2.4.0 milvus-model==0.2.0 pydantic==2.7.0 tqdm==4.66.2 FlagEmbedding==1.2.9 torch==2.2.2 numpy==1.24.4

Usage Examples

Creating a Database Collection

Here's how to create a new database collection using milvus_api:

from milvus_api.db_model.milvus_db import MilvusDB
from pymilvus import FiledSchema, DataType

host = 'localhost'
port = '19530'
db = MilvusDB(host, port)

fields = [  
    FieldSchema(name="id", dtype=DataType.INT64, auto_id=True, is_primary=True),
    FieldSchema(name="embeddings", dtype=DataType.FLOAT_VECTOR, dim=1024)
]

db.create_collection("my_collection", 1000, fields, "FLAT", "COSINE")

Data Encoding Using the Model

Here's how to perform query encoding using the embedding model in milvus_api:

from milvus_api.embed_model.bgem3 import PyMilvusBGEM3

query = ["Are you afraid of the earthquake"]
model_name = 'bgem3_model'

# Initialize the model
bgem3 = PyMilvusBGEM3(model_name)
results = bgem3.encode(query)

Release Note

• 0.2.1:(Feat): support IVF_PQ indexing algorithm for switch_index.
• 0.2.0:(Feat): add the switch_index operation to MilvusDB object.

from milvus_api.db_model.milvus_db import MilvusDB

db = MilvusDB(host, port)

# Assume you already have a collection with index.
db.switch_index("your_collection_name", "drop_index_name", "new_index_type", "field_name", metric_type")

• 0.1.4:(BugFix): Fix the logic error when do search operation.
• 0.1.3:(BugFix): Fix the error caused by the non-specific field argument during creating collection.
• 0.1.2:(BugFix): Cancel partial information of info function.
• 0.1.1: Modify this project structure to ensure that users can import 'milvus_api' normally.

Contributions

Thanks to all developers who contributed to this project. If you would like to contribute code, please follow the existing code style and submit a pull request.

Copyright and License

This project is licensed under the Apache License 2.0.

Other Information

This README.md file is now complete, including an introduction to the project, installation instructions, usage examples, contribution guidelines, and copyright and licensing information. You can save this content as the README.md file in the root directory of your project. If you need further modifications or wish to add additional information, feel free to let me know!