Docker template for basic data science packages to interface with Neo4j

This container is based on the Juptyer data science container along side creating a Neo4j database on your local machine. This container does NOT include support for GPU-based deep learning packages (that will be a separate repo). There are a lot of Python packages in this container that you might not use and so this container might be a bit bloated. It is more of a generic container to get you started.

First, build the container from the CLI:

docker-compose build

Then start the container:

docker-compose up

This will start up both a Jupyter Lab notebook at localhost:8888 (note that there will be a token associated with this which you will need to grab as that container is starting) as well as the Neo4j browser at localhost:7474. This are both configurable to whatever port you want.

When you are done with the container, stop it vial CTRL-c and then:

docker-compose down

• There are two different Python packages that can be used to connect to Neo4j from within Python. It is probably easiest to just pick one and go with it.
  • neo4j(https://neo4j.com/docs/api/python-driver/current/): The official, Neo4j-supported Python driver
  • py2neo(https://py2neo.org/2021.0/): A community-developed driver with lots of solid documentation and examples out there
• The container is set up to run Neo4j with the user name neo4j and the password as 1234. You will want to change the password in particular.
• There is a notebook in notebooks/ that tests to make sure that Jupyter can properly connect to Neo4j. Run this to verify.
• The docker-compose.yml is set to create and read data into Neo4j (including creating the database itself) in the directory $HOME/graph_data/my_data. You should change this to whatever directory you want to store the database. You will also use these directorys should you want to read .csv files into the database.
• The docker-compose.yml file includes some optional environment variables for setting memory values. Since this container includes the Graph Data Science (GDS) library, it is a good idea to add some heap memory. The exact values will depend on your machine. You will want to experiment with tuning this based on these instructions
• The Neo4j portion of the container includes two libraries that are very helpful for doing data science with graphs. These are APOC and GDS. The latest versions should be pulled when this container executes. Links to their use are provided below.

• Neo4j
  • Awesome Procedures on Cypher (APOC)
  • Cypher Manual
  • Cypher Reference Card
  • Graph Data Science (GDS) Library
  • Python Driver API Docs