xichenn/embedindb

A lightweight vector database

Embedin

Installation

pip install embedin

Quick Start

Using memery DB

from embedin.client import Client

client = Client(collection_name="test_collection")
client.add_data(texts=["This is a test"], meta_data=[{"source": "abc4"}])
result = client.query("These are tests", top_k=1)

print("result: ", result)

Using sqlite with local storage

from embedin.client import Client

url = 'sqlite:///test.db'
client = Client(collection_name="test_collection", url=url)
client.add_data(texts=["This is a test"], meta_data=[{"source": "abc4"}])
result = client.query("These are tests", top_k=1)

Using PostgreSQL

import os

from embedin.client import Client

url = os.environ.get('EMBEDIN_POSGRES_URL', "postgresql+psycopg2://embedin:embedin@localhost/embedin_db")
client = Client(collection_name="test_collection", url=url)
client.add_data(texts=["This is a test"], meta_data=[{"source": "abc4"}])
result = client.query("These are tests", top_k=1)

Using MySQL

import os

from embedin.client import Client

url = os.environ.get('EMBEDIN_MYSQL_URL', "mysql+pymysql://embedin:embedin@localhost/embedin_db")
client = Client(collection_name="test_collection", url=url)
client.add_data(texts=["This is a test"], meta_data=[{"source": "abc4"}])
result = client.query("These are tests", top_k=1)

Using MS-SQL

import os

from embedin.client import Client

url = os.environ.get('EMBEDIN_MSSQL_URL', "mssql+pymssql://sa:StrongPassword123@localhost/tempdb")
client = Client(collection_name="test_collection", url=url)
client.add_data(texts=["This is a test"], meta_data=[{"source": "abc4"}])
result = client.query("These are tests", top_k=1)

For development

Clone the repo

git clone https://github.com/EmbedInAI/embedin.git

Start PostgreSQL DB (Only necessary for using PostgreSQL as the storage)

cd docker
docker-compose up embedin-postgres

Start MySQL DB (Only necessary for using MySQL as the storage)

cd docker
docker-compose up embedin-mysql

Start MS-SQL DB (Only necessary for using MS-SQL as the storage)

cd docker
docker-compose up embedin-mssql

Start Oracle DB (Only necessary for using Oracle as the storage) - not supported yet

docker login container-registry.oracle.com

cd docker
docker-compose up embedin-oracle
```yaml
  embedin-oracle:
    container_name: embedin-oracle
    image: container-registry.oracle.com/database/enterprise:21.3.0.0
    environment:
      - ORACLE_SID=ORCLCDB
      - ORACLE_PWD=password
      - ORACLE_PDB=ORCLPDB1
      - ORACLE_CHARACTERSET=AL32UTF8
    volumes:
      - ./oracle/data:/opt/oracle/oradata
    ports:
      - "1521:1521"

Run unit test with coverage report

coverage run -m unittest discover -s tests -p '*.py'
coverage report

To publish to PyPI

pip install twine

python setup.py sdist && python setup.py bdist_wheel && twine upload dist/*