neo4j/neo4j-javascript-driver

Neo4j Driver for JavaScript

This is the official Neo4j driver for JavaScript.

Starting with 6.0, the Neo4j Drivers is no longer on a monthly release cadence. Minor version releases will happen when there are sufficient new features or improvements to warrant them. This is to reduce the required work of users updating their driver.

As a policy, patch versions will not be released except on rare occasions. Bug fixes and updates will go into the latest minor version and users should upgrade to that. Driver upgrades within a major version will never contain breaking API changes.

See also: https://neo4j.com/developer/kb/neo4j-supported-versions/

Resources to get you started:

• API Documentation
• Neo4j Driver Manual
• Neo4j Cypher Cheatsheet

What's New in 6.x

• Changelog

Including the Driver

In Node.js application

The node version of the driver requires node 18 or later.

Stable channel:

npm install neo4j-driver

Pre-release channel:

npm install neo4j-driver@next

Please note that @next only points to pre-releases that are not suitable for production use. To get the latest stable release omit @next part altogether or use @latest instead.

var neo4j = require('neo4j-driver')

Driver instance should be closed when Node.js application exits:

driver.close() // returns a Promise

otherwise application shutdown might hang or it might exit with a non-zero exit code.

In web browser

We build a special browser version of the driver, which supports connecting to Neo4j over WebSockets. It can be included in an HTML page using one of the following tags:

<!-- Direct reference -->
<script src="lib/browser/neo4j-web.min.js"></script>

<!-- unpkg CDN non-minified -->
<script src="https://unpkg.com/neo4j-driver"></script>
<!-- unpkg CDN minified for production use, version X.Y.Z -->
<script src="https://unpkg.com/[email protected]/lib/browser/neo4j-web.min.js"></script>

<!-- jsDelivr CDN non-minified -->
<script src="https://cdn.jsdelivr.net/npm/neo4j-driver"></script>
<!-- jsDelivr CDN minified for production use, version X.Y.Z -->
<script src="https://cdn.jsdelivr.net/npm/[email protected]/lib/browser/neo4j-web.min.js"></script>

This will make a global neo4j object available, where you can create a driver instance with neo4j.driver:

var driver = neo4j.driver(
  'neo4j://localhost',
  neo4j.auth.basic('neo4j', 'password')
)

From 5.4.0, this version is also exported as ECMA Script Module. It can be imported from a module using the following statements:

// Direct reference
import neo4j from 'lib/browser/neo4j-web.esm.min.js'

// unpkg CDN non-minified , version X.Y.Z where X.Y.Z >= 5.4.0
import neo4j from 'https://unpkg.com/[email protected]/lib/browser/neo4j-web.esm.js'

// unpkg CDN minified for production use, version X.Y.Z where X.Y.Z >= 5.4.0
import neo4j from 'https://unpkg.com/[email protected]/lib/browser/neo4j-web.esm.min.js'

// jsDelivr CDN non-minified, version X.Y.Z where X.Y.Z >= 5.4.0
import neo4j from 'https://cdn.jsdelivr.net/npm/[email protected]/lib/browser/neo4j-web.esm.js'

// jsDelivr CDN minified for production use, version X.Y.Z where X.Y.Z >= 5.4.0
import neo4j from 'https://cdn.jsdelivr.net/npm/[email protected]/lib/browser/neo4j-web.esm.min.js'

It is not required to explicitly close the driver on a web page. Web browser should gracefully close all open WebSockets when the page is unloaded. However, driver instance should be explicitly closed when it's lifetime is not the same as the lifetime of the web page:

driver.close() // returns a Promise

Usage examples

Constructing a Driver

// Create a driver instance, for the user `neo4j` with password `password`.
// It should be enough to have a single driver per database per application.
var driver = neo4j.driver(
  'neo4j://localhost',
  neo4j.auth.basic('neo4j', 'password')
)

// Close the driver when application exits.
// This closes all used network connections.
await driver.close()

Acquiring a Session

Regular Session

// Create a session to run Cypher statements in.
// Note: Always make sure to close sessions when you are done using them!
var session = driver.session()

with a Default Access Mode of READ

var session = driver.session({ defaultAccessMode: neo4j.session.READ })

with Bookmarks

var session = driver.session({
  bookmarks: [bookmark1FromPreviousSession, bookmark2FromPreviousSession]
})

against a Database

var session = driver.session({
  database: 'foo',
  defaultAccessMode: neo4j.session.WRITE
})

Reactive Session

// Create a reactive session to run Cypher statements in.
// Note: Always make sure to close sessions when you are done using them!
var rxSession = driver.rxSession()

with a Default Access Mode of READ

var rxSession = driver.rxSession({ defaultAccessMode: neo4j.session.READ })

with Bookmarks

var rxSession = driver.rxSession({
  bookmarks: [bookmark1FromPreviousSession, bookmark2FromPreviousSession]
})

against a Database

var rxSession = driver.rxSession({
  database: 'foo',
  defaultAccessMode: neo4j.session.WRITE
})

Transaction functions

// Transaction functions provide a convenient API with minimal boilerplate and
// retries on network fluctuations and transient errors. Maximum retry time is
// configured on the driver level and is 30 seconds by default:
// Applies both to standard and reactive sessions.
neo4j.driver('neo4j://localhost', neo4j.auth.basic('neo4j', 'password'), {
  maxTransactionRetryTime: 30000
})

Reading with Async Session

// It is possible to execute read transactions that will benefit from automatic
// retries on both single instance ('bolt' URI scheme) and Causal Cluster
// ('neo4j' URI scheme) and will get automatic load balancing in cluster deployments
var readTxResultPromise = session.executeRead(txc => {
  // used transaction will be committed automatically, no need for explicit commit/rollback

  var result = txc.run('MATCH (person:Person) RETURN person.name AS name')
  // at this point it is possible to either return the result or process it and return the
  // result of processing it is also possible to run more statements in the same transaction
  return result
})

// returned Promise can be later consumed like this:
readTxResultPromise
  .then(result => {
    console.log(result.records)
  })
  .catch(error => {
    console.log(error)
  })
  .then(() => session.close())

Reading with Reactive Session

rxSession
  .executeRead(txc =>
    txc
      .run('MATCH (person:Person) RETURN person.name AS name')
      .records()
      .pipe(map(record => record.get('name')))
  )
  .subscribe({
    next: data => console.log(data),
    complete: () => console.log('completed'),
    error: err => console.log(error)
  })

Writing with Async Session

// It is possible to execute write transactions that will benefit from automatic retries
// on both single instance ('bolt' URI scheme) and Causal Cluster ('neo4j' URI scheme)
var writeTxResultPromise = session.executeWrite(async txc => {
  // used transaction will be committed automatically, no need for explicit commit/rollback

  var result = await txc.run(
    "MERGE (alice:Person {name : 'Alice'}) RETURN alice.name AS name"
  )
  // at this point it is possible to either return the result or process it and return the
  // result of processing it is also possible to run more statements in the same transaction
  return result.records.map(record => record.get('name'))
})

// returned Promise can be later consumed like this:
writeTxResultPromise
  .then(namesArray => {
    console.log(namesArray)
  })
  .catch(error => {
    console.log(error)
  })
  .then(() => session.close())

Writing with Reactive Session

rxSession
  .executeWrite(txc =>
    txc
      .run("MERGE (alice:Person {name: 'James'}) RETURN alice.name AS name")
      .records()
      .pipe(map(record => record.get('name')))
  )
  .subscribe({
    next: data => console.log(data),
    complete: () => console.log('completed'),
    error: error => console.log(error)
  })

Consuming Records

Consuming Records with Streaming API

// Run a Cypher statement, reading the result in a streaming manner as records arrive:
session
  .run('MERGE (alice:Person {name : $nameParam}) RETURN alice.name AS name', {
    nameParam: 'Alice'
  })
  .subscribe({
    onKeys: keys => {
      console.log(keys)
    },
    onNext: record => {
      console.log(record.get('name'))
    },
    onCompleted: () => {
      session.close() // returns a Promise
    },
    onError: error => {
      console.log(error)
    }
  })

Subscriber API allows following combinations of onKeys, onNext, onCompleted and onError callback invocations:

• zero or one onKeys,
• zero or more onNext followed by onCompleted when operation was successful. onError will not be invoked in this case
• zero or more onNext followed by onError when operation failed. Callback onError might be invoked after couple onNext invocations because records are streamed lazily by the database. onCompleted will not be invoked in this case.

Consuming Records with Promise API

// the Promise way, where the complete result is collected before we act on it:
session
  .run('MERGE (james:Person {name : $nameParam}) RETURN james.name AS name', {
    nameParam: 'James'
  })
  .then(result => {
    result.records.forEach(record => {
      console.log(record.get('name'))
    })
  })
  .catch(error => {
    console.log(error)
  })
  .then(() => session.close())

Consuming Records with Reactive API

rxSession
  .run('MERGE (james:Person {name: $nameParam}) RETURN james.name AS name', {
    nameParam: 'Bob'
  })
  .records()
  .pipe(
    map(record => record.get('name')),
    concatWith(rxSession.close())
  )
  .subscribe({
    next: data => console.log(data),
    complete: () => console.log('completed'),
    error: err => console.log(err)
  })

Explicit Transactions

With Async Session

// run statement in a transaction
const txc = session.beginTransaction()
try {
  const result1 = await txc.run(
    'MERGE (bob:Person {name: $nameParam}) RETURN bob.name AS name',
    {
      nameParam: 'Bob'
    }
  )
  result1.records.forEach(r => console.log(r.get('name')))
  console.log('First query completed')

  const result2 = await txc.run(
    'MERGE (adam:Person {name: $nameParam}) RETURN adam.name AS name',
    {
      nameParam: 'Adam'
    }
  )
  result2.records.forEach(r => console.log(r.get('name')))
  console.log('Second query completed')

  await txc.commit()
  console.log('committed')
} catch (error) {
  console.log(error)
  await txc.rollback()
  console.log('rolled back')
} finally {
  await session.close()
}

With Reactive Session

rxSession
  .beginTransaction()
  .pipe(
    mergeMap(txc =>
      concatWith(
        txc
          .run(
            'MERGE (bob:Person {name: $nameParam}) RETURN bob.name AS name',
            {
              nameParam: 'Bob'
            }
          )
          .records()
          .pipe(map(r => r.get('name'))),
        of('First query completed'),
        txc
          .run(
            'MERGE (adam:Person {name: $nameParam}) RETURN adam.name AS name',
            {
              nameParam: 'Adam'
            }
          )
          .records()
          .pipe(map(r => r.get('name'))),
        of('Second query completed'),
        txc.commit(),
        of('committed')
      ).pipe(catchError(err => txc.rollback().pipe(throwError(() => err))))
    )
  )
  .subscribe({
    next: data => console.log(data),
    complete: () => console.log('completed'),
    error: error => console.log(error)
  })

Numbers and the Integer type

The Neo4j type system uses 64-bit signed integer values. The range of values is between -(264- 1) and (263- 1).

However, JavaScript can only safely represent integers between Number.MIN_SAFE_INTEGER -(253- 1) and Number.MAX_SAFE_INTEGER (253- 1).

In order to support the full Neo4j type system, the driver will not automatically convert to javascript integers. Any time the driver receives an integer value from Neo4j, it will be represented with an internal integer type by the driver.

Any javascript number value passed as a parameter will be recognized as Float type.

Writing integers

Numbers written directly e.g. session.run("CREATE (n:Node {age: $age})", {age: 22}) will be of type Float in Neo4j.

To write the age as an integer the neo4j.int method should be used:

var neo4j = require('neo4j-driver')

session.run('CREATE (n {age: $myIntParam})', { myIntParam: neo4j.int(22) })

To write an integer value that are not within the range of Number.MIN_SAFE_INTEGER -(253- 1) and Number.MAX_SAFE_INTEGER (253- 1), use a string argument to neo4j.int:

session.run('CREATE (n {age: $myIntParam})', {
  myIntParam: neo4j.int('9223372036854775807')
})

Reading integers

In Neo4j, the type Integer can be larger what can be represented safely as an integer with JavaScript Number.

It is only safe to convert to a JavaScript Number if you know that the number will be in the range Number.MIN_SAFE_INTEGER -(253- 1) and Number.MAX_SAFE_INTEGER (253- 1).

In order to facilitate working with integers the driver include neo4j.isInt, neo4j.integer.inSafeRange, neo4j.integer.toNumber, and neo4j.integer.toString.

var smallInteger = neo4j.int(123)
if (neo4j.integer.inSafeRange(smallInteger)) {
  var aNumber = smallInteger.toNumber()
}

If you will be handling integers that is not within the JavaScript safe range of integers, you should convert the value to a string:

var largeInteger = neo4j.int('9223372036854775807')
if (!neo4j.integer.inSafeRange(largeInteger)) {
  var integerAsString = largeInteger.toString()
}

Enabling native numbers

Starting from 1.6 version of the driver it is possible to configure it to only return native numbers instead of custom Integer objects. The configuration option affects all integers returned by the driver. Enabling this option can result in a loss of precision and incorrect numeric values being returned if the database contains integer numbers outside of the range [Number.MIN_SAFE_INTEGER, Number.MAX_SAFE_INTEGER]. To enable potentially lossy integer values use the driver's configuration object:

var driver = neo4j.driver(
  'neo4j://localhost',
  neo4j.auth.basic('neo4j', 'password'),
  { disableLosslessIntegers: true }
)

Writing and reading Vectors

Neo4j supports storing vector embeddings in a dedicated vector type. Sending large lists with the driver will result in significant overhead as each value will be transmitted with type information, so the 6.0.0 release of the driver introduced the Neo4j Vector type.

The Vector type supports signed integers of 8, 16, 32 and 64 bits, and floats of 32 and 64 bits. The Vector type is a wrapper for JavaScript TypedArrays of those types.

To create a neo4j Vector in your code, do the following:

var neo4j = require('neo4j-driver')

var typedArray = Float32Array.from([1, 2, 3]) //this is how to convert a regular array of numbers into a TypedArray, useful if you handle vectors as regular arrays in your code

var neo4jVector = neo4j.vector(typedArray) //this creates a neo4j Vector of type Float32, containing the values [1, 2, 3]

driver.executeQuery('CREATE (n {embeddings: $myVectorParam})', { myVectorParam: neo4jVector })

To access the data in a retrieved Vector you can do the following:

var retrievedTypedArray = neo4jVector.asTypedArray() //This will return a TypedArray of the same type as the Vector

var retrievedArray = Array.from(retrievedTypedArray) //This will convert the TypedArray to a regular array of Numbers. (Not safe for Int64 arrays)