This docker container is intended to help setup and run an environment to learn and practice vector databases with a focus on pgvector. The container is not optimized for size but instead includes most of the tooling needed to get started withou having to build the env from scratch. This means the following is installed at a minimum.

• rocky linux 8
• postgres 16
• pgvector
• python 3.9
• bunch of needed python packages
• python scripts to generate embeddings using openai and minilm. ( Located in / )
• sample schema
• sample test data

Note: If using openai, you will need to generate an api key. Visit https://platform.openai.com/api-keys for more details

I have added the sample data file to this repo it is 400MB. I used the lfs plugin for git.

The entry script will perform the following:

• Initiate the postgres db if its the first time running
• Set the postgres password to "postgres"
• Create a role named "vectordb" with a login password of "vectordb"
• Create a database called "vectordb" owned by "vectordb"
• Create the necessary tables inside the vectordb database
• Populate the tables using the vectordb_data.sql

• All entries in pg_hba.conf have all been set to trust
• The data directory is /pgdata/16/data
• The data directory should be preserved even if you remove the container. Obviouldy removing the volume will deletethe data

For some reason git clone will not copy the complete compressed data file into your local repo. You must clone the repo then click on vectordb_data.sql.gz via the browser by right clicking and downloading it directly to your local repo before building your docker image. It is a 400MB+ file.

You can also curl the raw version after cloning:

curl -LO https://github.com/jtorral/vectorDB/raw/refs/heads/main/vectordb_data.sql.gz

Try this and see if you get a 400MB+ file. If so it works

curl -LO https://github.com/jtorral/vectorDB/raw/refs/heads/main/vectordb_data.sql.gz

The image takes a few minutes to build since there is a lot mojo in it.

docker build -t vectordb .

In the below example I am mapping port 6432 to the internal 5432 in the container. You can use whatever port you like

docker run -p 6432:5432 --env=PGPASSWORD=postgres -v vectorDB:/pgdata --name vectorDB -d vectordb

/getFromMinilm.py
/getFromOpenai.py

docker stop vectorDB; docker rm vectorDB;

Keep in mind removing container volume will permanatly delete data you may have saved and will require you to run the container again which will recreate the volume

docker volume rm vectorDB

psql -h localhost -p 6432 -U vectordb

psql (16.4)
Type "help" for help.

vectordb=> \dt
              List of relations
 Schema |      Name       | Type  |  Owner   
--------+-----------------+-------+----------
 public | articles        | table | vectordb
 public | articles_minilm | table | vectordb
 public | articles_openai | table | vectordb
(3 rows)

docker exec -it vectorDB /bin/bash

[root@bb99b4fe8b73 /]# psql -U vectordb

psql (16.4)
Type "help" for help.

vectordb=> \dt
              List of relations
 Schema |      Name       | Type  |  Owner   
--------+-----------------+-------+----------
 public | articles        | table | vectordb
 public | articles_minilm | table | vectordb
 public | articles_openai | table | vectordb
(3 rows)

vectordb=> 

In one command:

docker exec -it vectorDB psql -U vectordb
psql (16.5)
Type "help" for help.

vectordb=>

Send me an email or message with issues or suggestions.