Keymaker is an analytics framework built on top of Neo4j. It allows you to write Cypher based pipleines on top of your Neo4j database without having to write any infrastructure code. For each Cypher pipleine you write, the framework exposes a graphql API endpoint which allows you to easily run the pipeline and/or consume the results from a frontend application. It also has built in collabration capabilities, performance testing features and encourages you to write more readable Cypher.
Keymaker has 3 components: the admin dashboard, admin API, and engine API. The admin dashboard and API facilitate user interaction with the framework (i.e. database connection management, engine creation, and writing Cypher). The engine API connects with your Neo4j database and exposes an endpoint which excecutes the engine.
Online help for Keymaker can be found here: https://help.neo4j.solutions/neo4j-solutions/keymaker/
Link to the Keymaker Blogpost can be found here: https://medium.com/@snehal.maid/keymaker-dc496d41b21f
The project is divided into 4 main folders:
- admin-api: This facilitate user interaction with the framework (i.e. database connection management, engine creation, and writing Cypher)
- engine-api: This connects with your Neo4j database and exposes an endpoint which excecutes the engine
- data: contains the initialization cypher script for running keymaker and also, the engine import json and the api graphql queries
- ui: contains the React UI
To run the code locally, you will need to do 3 things:
- Setup a Neo4j database
- Build and configure the admin-api
- Build and configure the engine-api
- Build and configure the ui
You will need to setup an Aura instance or a Neo4j database to use as the backend storage for Keymaker.
- If you are running Aura or Neo4j Community, you will use the default database neo4j.
- If you are running Neo4j Enterprise Edition, you can:
- Use the default database neo4j
- Create a new database. See this link for instructions on how to create a Neo4j database.
- For Neo4j Community or Enterprise Edition, you will also need to install APOC Core. For instructions on how to do this, please follow this link.
- Aura users will already have this installed.
Within the folder data/cypher
you will find a few Cypher scripts:
- constraints_v5.cypher
- cypher_init_data.cypher
If you are running a version older than 5, you may have to manually adjust the syntax to match the expected syntax of your Neo4j version.
Follow these steps to run the Cypher:
- Run either Neo4j Browser or Workspace and connect to your Neo4j database
- Copy/paste constraints_v5.cypher into Browser/Workspace and execute it
- If running in Browser, make sure this setting is checked: "Enable multi statement query editor"
- Copy/paste the cypher_init_data.cypher file into Browser/Workspace and execute it
Your database is now ready to be used.
Follow the instructions in the admin-api
README file to build and run the Admin API.
Follow the instructions in the engine-api
README file to build and run the Engine API.
Follow the instructions in the ui
README file to build and run the UI.
The create-user.sh command line utility has been provided to create users in Keymaker. You can find this script under admin-api/scripts/create-user.sh
If you run ./create-user.sh with no arguments, it will display a usage message:
Usage: ./create-user.sh <email> <password> <name> [picture] [graphql port] [https]
The parameters display in square brackets are optional, however, for this deployment you will need to specify the [graphql port], and since the parameters are positional, the [picture] also. The graphql port is specified in the docker-compose.yml file. The value will be the graphql port for admin-api server.
An example command looks like:
./create-user.sh admin@neo4j.com password "Keymaker Admin" "" 4000
The response looks like:
Creating admin@neo4j.com
{"data":{"createUserSignUp":{"email":"admin@neo4j.com","name":"Keymaker Admin","picture":""}}}
[source]
Upon success, this will create a node in the database with the :User.
Assuming you installed this locally, use your web browser to navigate to http://localhost:3000/login. You will be presented with a login screen.
Enter the email and password that you created in the Creating Users step and hit or click the Continue button. You should now be logged in.
Follow these steps to create a Neo4j sandbox instance for demo datasets and to prepare for running Keymaker:
-
Access Neo4j Sandbox:
- Navigate to Neo4j Sandbox. Sign up or log in as needed, completing any onboarding forms.
-
Select Dataset:
- Opt for the Recommendations Project for a pre-built movie reviews dataset. Click Create.
If you want to run keymaker against the Recommendations Project beyond the 3-day sandbox expiration, you can follow the below steps to get a database dump (Optional)
-
Request Backup:
- Within your project, find Backups > Request Backup. After processing, click Request Download Link to obtain the dump file.
-
Restore Database:
- Use the dump file to restore the database on your local Neo4j Desktop or Aura platform. For detailed instructions, refer to Neo4j's backup and restore guide.
This setup will support the use of Keymaker with a demo dataset, facilitating the creation of sample database connections and engines.
Following a successful Keymaker installation, proceed with establishing database connections and configuring query pipelines for the recommendations project.
-
Access Database Connections:
- Navigate to Database Connections and select New Connection (+).
-
Input Connection Details:
- Name:
Movies Sandbox
- Connection URL: Replace
<enter-your-connection-url>
with your specific connection URL. - Private: Choose to enable or disable based on your preference by replacing
<you-can-enable-or-disable>
. - Username: Substitute
<enter-username>
with your username. - Password: Substitute
<enter-password>
with your password.
- Name:
-
Submission and Verification:
- Click Submit, then refresh the page to view the Database Connection Card for the Movies Sandbox.
Note: For users of the Recommendations project on a Neo4j Sandbox, obtain your connection details (username, password, and bolt URL) from the sandbox's Connection Details section. If you see the Green check, it means that the database connection is successful. Always make sure to have apoc installed on the database connection that you are trying to create.
To create and configure query engines, follow the general steps below for each engine type. Replace specific placeholders and values as indicated for each engine.
-
Navigate to Engine Creation:
- Go to http://localhost:3000 and under Engines, select New Engine (+).
-
Enter Engine Information:
- Engine Name, Description, Engine ID: Use the specific values listed under each engine type below.
- Database Connection: Select the database connection card that we created in the earlier step.
- Database Name: Default, or specify the database name if using an enterprise/commercial Neo4j license.
- Cypher Workbench Data Model (Optional):
None
- Return Label:
Movie
- Return Properties:
title, url, plot, movieId
-
Submit and Refresh:
- Click Submit, then refresh the page to view the Engine Card. Select the engine card or use the provided link to navigate directly.
-
Import Queries:
- Click Import > from json.
- Navigate to the specific path in the GitHub repo listed for each engine, copy the content, paste it, and click Submit.
Movies - Cold Start Query Engine
- Engine Name:
Movies - Cold Start
- Description:
Provides recommendations without user context or preference.
- Engine ID:
movies-cold-start
- Import Path:
data/engineImports/movies-cold-start.json
(Creates 5 phases)
Movies - Collaborative Query Engine
- Engine Name:
Movies - Collaborative
- Description:
Recommends movies based on user preference.
- Engine ID:
movies-collaborative-based-recommendation
- Import Path:
data/engineImports/movies-collaborative-based-recommendation.json
(Creates 3 phases)
Movies - Content Based Engine
- Engine Name:
Movies - Content Based
- Description:
Generates recommendations based on shared movie attributes.
- Engine ID:
movies-content-based-recommendation
- Import Path:
data/engineImports/movies-content-based-recommendation.json
(Creates 5 phases)
Final Step:
- Repeat the process for each engine type as needed. By the end, you should have one database connection for the movies sandbox and three configured query engines.
-
Access Apollo Sandbox: Go to
http://localhost:4001/graphql
(or your engine-API server) and open the Apollo Sandbox by clicking on Query your Server. -
Enable Access: If access issues arise, ensure
INTROSPECTION
andPLAYGROUND_ENABLED
environment variables are set totrue
. -
Configure Request:
- In the Operation tab, prepare your GraphQL query.
- For API key authentication, go to Headers, click + New Header, enter
api-key
as the key andLZmIbId7He4dh0hH7ZUVKz5A
as the value.
After setting the API key, you're ready to run your query pipelines.
query {
recommendations(
engineID: "movies-cold-start"
first: 20
skip: 0
) {
item
score
details
}
}
query {
recommendations(
engineID: "movies-collaborative-based-recommendation",
params: {startMovieTitle: "Shawshank Redemption, The"}
first: 20
skip: 0
) {
item
score
details
}
}
query {
recommendations(
engineID: "movies-content-based-recommendation",
params: {startMovieTitle: "Tales from the Crypt Presents: Demon Knight"}
first: 20
skip: 0
) {
item
score
details
}
}
These sections describe further optional configuration.
When using the default local authentication scheme, there is a simple encryption process for encrypting a user's password. The encryption key used is set both at the UI and API layers. It is not required to modify this key, but it is recommended before creating any users. Once you have created users, do not modify it again.
admin-api/.env: ENCRYPTION_KEY=keymakerEncryptionKey
engine-api/.env: ENCRYPTION_KEY=keymakerEncryptionKey
Pick a new value, and make sure the new value is set in both files. You will need to restart the admin-api and engine-api projects if they are running.
Additionally, any user that was created before the modifying the encryption key will no longer work. You will have to re-run the created users step to fix this issue.