PostgreSQL is a powerful database that we’re using. Not only the common data, but the Vector data.
PostgreSQL supports vector data types by a extension called pgvector.
In this article, we’re going to install PostgreSQL and pgvector.
Prerequisites
Assume you have a Linux system, and currently we are using CentOS 8.
And PostgreSQL 15 is the last version that supports CentOS 8. By the way, pgvector needs PostgreSQL 11 or later.
Install PostgreSQL
Go to the official site to download the rpm package.
Or you can just copy the following command to install it.
# Install the repository RPM:
yum install -y https://download.postgresql.org/pub/repos/yum/reporpms/EL-7-x86_64/pgdg-redhat-repo-latest.noarch.rpm# Install PostgreSQL:
yum install -y postgresql15-server# Optionally initialize the database and enable automatic start:
/usr/pgsql-15/bin/postgresql-15-setup initdb
systemctl enable postgresql-15
systemctl start postgresql-15
Check the PostgreSQL if it’s installed successfully.
psql --version
Some useful commands:
systemctl status postgresql-15
systemctl stop postgresql-15
systemctl restart postgresql-15
Setup PostgreSQL
After the installation, it will create a user named postgres automatically. Switch to the postgres user.
su - postgres
Now we enter the PostgreSQL shell, and change the password of the postgres user.
psql
ALTER USER postgres WITH PASSWORD 'postgres';
After completing these tasks, enter the exit command to quit.
Beware that the default password of Linux postgres user should be changed.
# 修改 linux 的 postgres 用户密码
# 注意，这个密码很容易被猜到，被攻击sudo passwd -d postgressudo -u postgres passwd
Install pgvector
By default, install pgvetor by compiling from source code. However, it’s such a tough task with a lot of errors.
Eventually, we found a way to install pgvector by using the RPM package. It’s SUPER easy and works well.
yum install pgvector_15
Create a user
Enter the PostgreSQL shell again: psql.
# 创建用户
CREATE USER dbroot WITH PASSWORD '123root';# 创建数据库
CREATE DATABASE vectdb OWNER dbroot;# 赋权
GRANT ALL PRIVILEGES ON DATABASE vectdb TO dbroot;# 添加 superuser 权限，用于安装 pgvector 插件
alter user dbroot with superuser;
Then just create a vector table to test if it works.
# 本地登录
> psql -U dbroot -h localhost -p 5432 -d vectdb# 创建插件
CREATE EXTENSION vector;# 创建一个带有向量类型的表
CREATE TABLE items (id bigserial PRIMARY KEY, embedding vector(3));# 写入一条数据
INSERT INTO items (embedding) VALUES ('[1,2,3]'), ('[4,5,6]');# 向量检索
SELECT * FROM items ORDER BY embedding <-> '[3,1,2]' LIMIT 5;
Open PostgreSQL to the outside accessing
> cd /var/lib/pgsql/15/data/> vim postgresql.conf
#将监听地址修改为*
#默认listen_addresses配置是注释掉的，所以可以直接在配置文件开头加入该行
listen_addresses='*'> vim pg_hba.conf
#在文件尾部加入
#注意，开启所有外网都能访问时，服务容易被攻击
host all all 0.0.0.0/0 md5#重启PostgreSQL服务
sudo systemctl restart postgresql-15
Some useful commands
#查看所有数据库
> \l#切换当前数据库
> \c vectdb#查看当前数据库下所有表
> \d#退出数据库
> \q# 查询配置文件所在位置
> show config_file;# 查询数据储存目录
> show data_directory;
Thanks to these articles that helped me a lot!
- https://juejin.cn/post/7316121883635466259
- https://www.cnblogs.com/xfishup/p/17904882.html
- https://blog.csdn.net/m0_62627802/article/details/135027327
)
