Skip to content

Configuring a node

Neurobagel is designed to be easily deployed with a single command without deep configuration. In many cases however, you will want to customize your deployment to fit your needs.

If you already have a running Neurobagel node, after making any configuration changes (including changing the data you want to be available in the graph database), follow the instructions to restart your services for the changes to take effect.

Deployment

Available services

The Neurobagel Docker Compose recipe includes several services and coordinates them to work together:

(In parentheses are the names of services within the Docker Compose stack)

  • Neurobagel node API/n-API (api): The API that communicates with a single graph store and determines how detailed the response to a query should be from that graph.
  • Graph store (graph): A third-party RDF store that stores Neurobagel-harmonized data to be queried. At the moment our recipe uses the free tier of GraphDB for this.
  • Neurobagel federation/f-API (federation): A special API that can federate over one or more Neurobagel nodes to provide a single point of access to multiple distributed databases. By default it will federate over all public nodes and any local nodes you specify.
  • Neurobagel query tool (query_tool): A web app that provides a graphical interface for users to query a federation API and view the results from one or more nodes. Because the query tool is a static app and is run locally in the user's browser, this service simply hosts the app.

Available profiles

Neurobagel offers different deployment profiles that allow you to spin up specific combinations of services (listed below), depending on your use case.

  1. full_stack: Best profile to get started with Neurobagel. It includes all services you need to run a local Neurobagel node and have the ability to query public nodes, along with a graphical query tool.

    • api
    • graph
    • federation
    • query_tool

    Info

    By default this profile will also federate over all publicly accessible Neurobagel nodes, although this behaviour can be disabled in the f-API using the environment variable [`NB_FEDERATE_REMOTE_PUBLIC_NODES`](#environment-variables).

  2. local_node: Best profile if you want to run a standalone Neurobagel node but rely on a separate deployment for providing federation and a graphical query tool (such as Neurobagel's own hosted public instances).

    • api
    • graph

    Info

    This is the default profile if you don't specify one.

  3. local_federation: Best profile if you already have multiple standalone (local or non-publicly-accessible) Neurobagel node deployments running and you now want to provide federation over them.

    • federation
    • query_tool

    Info

    If you only want to federate over a single local node and all public Neurobagel nodes, we recommend using the full_stack profile to set up your node and federation in one step. If you choose to use the local_federation profile, you will have to manually configure your local_nb_nodes.json file.

You can then launch these profiles by using the --profile flag with docker compose, e.g.:

docker compose --profile full_stack up -d

Take a look at the getting started guide for more information setting up for a first launch.

Environment variables

Below are all the possible Neurobagel environment variables that can be set in .env.

Environment variable Set manually in .env? Description Default value if not set Used in these installation modes
NB_GRAPH_ADMIN_PASSWORD Yes Secure password to set for the admin user. - Docker
NB_GRAPH_USERNAME Yes Username to set for the graph database user. - Docker, Python
NB_GRAPH_PASSWORD Yes Secure password to set for the graph database user. - Docker, Python
NB_GRAPH_DB Yes Name to give your graph database (e.g., for a GraphDB database, use the format repositories/{database_name}) repositories/my_db Docker, Python
LOCAL_GRAPH_DATA Yes Path on your filesystem to the JSONLD files you want to upload to the graph database ./data Docker
NB_API_ALLOWED_ORIGINS Yes, if using a frontend query tool ‡ Origins allowed to make cross-origin resource sharing requests. Multiple origins must be separated with spaces in a single string enclosed in quotes. See ‡ for more info "" Docker, Python
NB_API_QUERY_URL Yes URL (and port number, if needed) of the Neurobagel API that the query tool will send its requests to. The query tool sends requests from a user's machine, so ensure the API URL is provided as a user would access it from their own machine. See also the query tool README. - Docker
NB_RETURN_AGG Yes Whether to return only aggregate, dataset-level query results (excluding subject/session-level attributes). One of [true, false] true Docker, Python
NB_NAPI_TAG No Docker image tag for the Neurobagel node API latest Docker
NB_NAPI_PORT_HOST No Port number on the host machine to map the Neurobagel node API container port to 8000 Docker
NB_NAPI_PORT No Port number on which to run the Neurobagel node API in the API container 8000 Docker
NB_FAPI_TAG No Docker image tag for the Neurobagel federation API latest Docker
NB_FAPI_PORT_HOST No Port number on the host machine to map the Neurobagel federation API container port to 8080 Docker
NB_FAPI_PORT No Port number on which to run the Neurobagel federation API in the API container 8000 Docker
NB_GRAPH_IMG No Graph server Docker image ontotext/graphdb:10.3.1 Docker
NB_GRAPH_ADDRESS No IP address for the graph database (or container name, if graph is hosted locally) 206.12.99.17 (graph) ** Docker, Python
NB_GRAPH_PORT_HOST No Port number on the host machine to map the graph server container port to 7200 Docker
NB_GRAPH_PORT No Port number used by the graph server container 7200 * Docker, Python
NB_QUERY_TAG No Docker image tag for the query tool latest Docker
NB_QUERY_PORT_HOST No Port number used by the query_tool on the host machine 3000 Docker
NB_FEDERATE_REMOTE_PUBLIC_NODES Yes If "True", include public nodes in federation. If "False", only locally specified nodes in local_nb_nodes.json are queried. true Docker, Python
NB_QUERY_APP_BASE_PATH No The URL path for the query tool, determines the specific URL at which the app should be rendered for users to access it / Docker
NB_ENABLE_AUTH Yes (Experimental, for dev deployments only) Whether to enable authentication for cohort queries. One of [true, false] false Docker, Python
NB_QUERY_CLIENT_ID Yes (Experimental, for dev deployments only) OAuth client ID for the query tool. Required if NB_ENABLE_AUTH is set to true. - Docker, Python
Ensure that shell variables do not clash with .env file

If the shell you run docker compose from already has any shell variable of the same name set, the shell variable will take precedence over the configuration of .env! In this case, make sure to unset the local variable first.

For more information, see Docker's environment variable precedence.

Tip

Double check that any environment variables you have customized in .env are resolved with your expected values using the command docker compose config.

Change security relevant variables

At minimum, we recommend reviewing and changing the values of the following variables in .env for security purposes:

NB_GRAPH_ADMIN_PASSWORD
NB_GRAPH_USERNAME
NB_GRAPH_PASSWORD
NB_GRAPH_DB NB_RETURN_AGG
NB_API_QUERY_URL

Make sure to use a secure password for the NB_GRAPH_ADMIN_PASSWORD and NB_GRAPH_PASSWORD variables. Here is how you can create a random password in the terminal:

openssl rand -hex 16
Passwords are handled as Docker secrets

Although the NB_GRAPH_ADMIN_PASSWORD and NB_GRAPH_PASSWORD variables are set in the .env file, they are passed to the containers as Docker secrets. This ensures that your passwords are not exposed in the container logs or in the docker-compose.yml file.

Make sure to not share your .env file with others, especially if it contains sensitive information like passwords.

local_nb_nodes.json

This file is only used by deployment profiles that include the federation API. local_nb_nodes.json contains the URLs and (arbitrary) names of the node APIs of local nodes you wish to federate over. Each node must be denoted by a dictionary {} with two key-value pairs:

  • "NodeName": name of the node
  • "ApiURL": URL of the node API exposed for that node

Multiple nodes must be wrapped in a list [].

Example: Let's assume there are two local nodes already running on different servers of your institutional network, and you want to set up federation across both nodes:

  • a node named "My Institute" running on your local computer (localhost), on port 8000 and
  • a node named "Node Recruitment" running on a different computer with the local IP 192.168.0.1, listening on the default http port 80.

In your local_nb_nodes.json file you would configure this as follows:

local_nb_nodes.json
[
  {
    "NodeName": "My Institute",
    "ApiURL": "https://neurobagel.myinstitute.edu",
  },
  {
    "NodeName": "Node Recruitment",
    "ApiURL": "http://192.168.0.1"
  }
]

Do not use localhost/127.0.0.1 in local_nb_nodes.json

Even if the local node API(s) you are federating over are running on the same host machine as your federation API, you cannot use localhost for the ApiURL and instead have to provide a network-accessible URL or IP address. For an example, see the configuration for the node called "My Institute" above.

Nodes that do not need to be manually configured

We maintain a list of public Neurobagel nodes here. By default every new f-API will lookup this list on startup and include it in the list of nodes to federate over. This also means that you do not have to manually configure public nodes, i.e. you do not have to explicitly add them to your local_nb_nodes.json file.

To add one or more local nodes to the list of nodes known to your f-API, simply add more dictionaries to this file.

Manually setting up a Neurobagel graph backend

The Neurobagel Docker Compose recipe will automatically set up and configure all services for you after deployment. The automated setup steps are explained in more detail below.

For advanced users / debugging purposes only

The following steps are only documented as a reference and for advanced users. You should not need to do this in most cases.

Configuring the graph store

These are manual steps for configuring the GraphDB backend after launching the Neurobagel stack.

1: Set the password of the default admin superuser and enable password-based access to databases

Details

When you first launch the graph server, a default admin user with superuser privilege will automatically be created for you. This admin user is meant to create other database users and modify their permissions. (For more information, see the official GraphDB documentation.)

Doing this manually with curl

First, change the password for the admin user that has been automatically created by GraphDB:

curl -X PATCH --header 'Content-Type: application/json' http://localhost:7200/rest/security/users/admin -d '
{"password": "NewAdminPassword"}'
(make sure to replace "NewAdminPassword" with your own, secure password).

Next, enable GraphDB security to only allow authenticated users access: 

curl -X POST --header 'Content-Type: application/json' -d true http://localhost:7200/rest/security

and confirm that this was successful: 

➜ curl -X GET http://localhost:7200/rest/security                                                  
true

2: Create a new graph database user based on credentials defined in your .env file

Details

We do not recommend using admin for normal read and write operations, instead we can create a regular database user.

The .env file created as part of the docker compose setup instructions declares the NB_GRAPH_USERNAME and NB_GRAPH_PASSWORD for the database user. The Neurobagel API will send requests to the graph using these credentials.

Doing this manually with curl

When you launch the RDF store for the first time, we have to create a new database user:

curl -X POST --header 'Content-Type: application/json' -u "admin:NewAdminPassword" -d '
{
"username": "DBUSER",
"password": "DBPASSWORD"
}' http://localhost:7200/rest/security/users/DBUSER

Make sure to use the exact NB_GRAPH_USERNAME and NB_GRAPH_PASSWORD you defined in the .env file when creating the new database user. Otherwise the Neurobagel API will not have the correct permission to query the graph.

3: Create a new graph database with the name defined in your .env

Details

When you first launch the graph store, there are no graph databases. You have to create a new one to store your metadata.

By default the Neurobagel API will query a graph database named my_db. If you have defined a custom NB_GRAPH_DB name in the .env file, you will first need to create a database with a matching name.

Doing this manually with curl

In GraphDB, graph databases are called resources. To create a new one, you will also have to prepare a data-config.ttl file that contains the settings for the resource you will create (for more information, see the GraphDB docs).

You can edit this example file and save it as data-config.ttl locally. Ensure the value for rep:repositoryID in data-config.ttl matches the value in NB_GRAPH_DB in your .env file. For example, if NB_GRAPH_DB=repositories/my_db, then rep:repositoryID "my_db" ;.

Then, create a new graph database with the following command (replace "my_db" as needed). If your data-config.ttl is not in the current directory, replace "@data-config.ttl" in the command with "@PATH/TO/data-config.ttl".

curl -X PUT -u "admin:NewAdminPassword" http://localhost:7200/repositories/my_db --data-binary "@data-config.ttl" -H "Content-Type: application/x-turtle"

4: Grant the newly created user from step 2 permissions to access the database

Doing this manually with curl 
curl -X PUT --header 'Content-Type: application/json' -d '
{"grantedAuthorities": ["WRITE_REPO_my_db","READ_REPO_my_db"]}' http://localhost:7200/rest/security/users/DBUSER -u "admin:NewAdminPassword"
  • "WRITE_REPO_my_db": Grants write permission.
  • "READ_REPO_my_db": Grants read permission.

Make sure you replace my_db with the name of the graph db you have just created.

Non-automated options for interacting with the GraphDB backend
  1. Directly send HTTP requests to the HTTP REST endpoints of the GraphDB backend e.g. using curl. GraphDB uses the RDF4J API specification.
  2. Use the GraphDB web interface (called the Workbench), which offers a more accessible way to manage the GraphDB instance. Once your local GraphDB backend is running you can connect to the Workbench at http://localhost:7200. The Workbench is well documented on the GraphDB website.

Uploading data to the graph store

Data are automatically uploaded to the graph from the path specified with the LOCAL_GRAPH_DATA in the .env configuration file when the Neurobagel stack is (re-)started.

If you instead prefer to upload data manually, you can use the add_data_to_graph.sh script:

./add_data_to_graph.sh PATH/TO/YOUR/GRAPH-DATA \
  localhost:7200 repositories/my_db DBUSER DBPASSWORD \

Warning

To update any existing datasets in your graph database, you can clear the database and reupload all datasets using add_data_to_graph.sh following the command above and including the --clear-data flag. Ensure that you also re-upload the Neurobagel vocabulary file nb_vocab.ttl following the section below.

Adding vocabulary files to the graph database

Why we need vocabulary files in the graph

In the context of an RDF store, in addition to information about specific observations of given standardized concepts such as "subject", "age", and "diagnosis" (represented in the subject-level JSONLDs generated by Neurobagel tools), hierarchical relationships between concepts themselves can also be represented. Including these relationships in a graph is important to be able to answer questions such as how many different diagnoses are represented in a graph database, to query for higher-order concepts for a given variable, and more.

The participant variables modeled by Neurobagel are named using Neurobagel's own vocabulary (for more information, see this page on controlled terms). This vocabulary, which defines internal relationships between vocabulary terms, is serialized in the file nb_vocab.ttl available from the neurobagel/recipes repository. If you have cloned this repository, you will already have downloaded the vocabulary file.

The nb_vocab.ttl file should be added to every created Neurobagel graph database. Below is an example of how you would do this using the same script we used to upload the dataset JSONLD files, add_data_to_graph.sh.

(assumes you are in the scripts subdirectory inside the recipes repository):

./add_data_to_graph.sh ../vocab \
  localhost:7200 repositories/my_db DBUSER DBPASSWORD