ONVIF client protocol implementation for Node.js.
Tip
This page describes the 1.x version of the ONVIF library written in TypeScript. The 1.x version is currently in the release candidate stage. If you are looking for the README for the stable 0.x version, please see branch v0.x
The default npm installation still uses version 0.x. If you want to try this new version, install it with:
npm install onvif@rcA TypeScript-first ONVIF client for Node.js with typed WSDL interfaces, Promise-based APIs, and broad service coverage:
- Core media & PTZ — (Profiles S, T) device info, Media / Media2 profiles, stream URIs, imaging, presets, continuous and absolute moves
- Events — (Profile M) pull-point and WS-BaseNotification with topic
filters and
EventEmitterintegration - Recording & replay — (Profile G) NVR search, recordings, and replay URIs
- Physical access — (Profiles C, A) DoorControl, AccessControl, Credential, AccessRules, Schedule
- More services — (Profile T) Analytics, DeviceIO, Display, Action Engine, Thermal, Provisioning, AdvancedSecurity
- Discovery & auth — WS-Discovery on the LAN; WS-Security and Digest (MD5 / SHA-1 / SHA-256)
Works server-side on Node.js 18+, tested on GitHub Actions.
This is a new version of the ONVIF library. The previous version was written in JavaScript, while this version is written in TypeScript and includes interfaces describing ONVIF data structures.
At the moment, all the methods from v0.8 have been implemented in the new Onvif API, and a
v0.x compatibility layer is available
for existing projects.
Tip
The main 1.x API uses the Onvif class with service modules (onvif.device, onvif.media, onvif.ptz, …).
For v0.8 migration, import the separate compatibility modules (not part of require('onvif')):
require('onvif/compatibility') for callbacks or require('onvif/compatibility/promises') for async/await
(both export Cam and Discovery).
The documentation for the new library was generated with TypeDoc and is available here:
Thanks a lot for your interest!
I will be happy to answer any questions and hear your feedback.
- TypeScript interfaces for the latest ONVIF WSDL specification generated by onvif-generate-interfaces to provide code completion and type checking in the IDE for the requested and returned values
- Complete documentation
- Tests using the real ONVIF server from HappyTimeSoft
- Event support: pull-point, base ws-notification, filters, EventEmitter inheritance. See below
- Lazy loading of ONVIF service modules — see Performance / lazy loading
- Authentication with WS-Security and Digest (MD5, SHA-1, SHA-256), also Advanced Security (experimental)
- WS-Discovery support for finding devices on the local network
- Full:
Device,Events,Media,Media2,PTZ,Imaging,Analytics,AnalyticsDevice,Recording,Replay,Search,Receiver,DeviceIO,Display,Action Engine,Thermal,DoorControl,AccessControl,Credential,AccessRules,Schedule,Provisioning,AdvancedSecuritysupport.Not yet implemented (interfaces only, from ONVIF Network Interface Specifications): AuthenticationBehavior, Application Management (appmgmt), Uplink, FederatedSearch
- Improved error handling
- Compatible with the original API structure
- Optional v0.x layer (not in the main export):
require('onvif/compatibility')for callbacks,require('onvif/compatibility/promises')for Promises (both exportCamandDiscovery)
Before you can use most library methods, call connect() on your Onvif instance. This method performs
the initial handshake with the device and fills internal state so later SOAP requests are authenticated and
routed to the correct service endpoints.
connect() runs the following steps in order:
- Time synchronization —
getSystemDateAndTime()is called first. ONVIF WS-Security authentication includes a timestamp in the nonce digest, so the client must know the offset between its own clock and the device clock (timeShift). The library tries an unauthenticated request first, as the ONVIF spec allows, and retries with credentials when the device requires authentication (some Panasonic and Digital Barriers models behave this way). - Service discovery — the library calls
GetServices(the modern ONVIF approach introduced with Profile T) from a smallconnectionhelper, without loading the fulldevicemodule. If that fails on older devices, it falls back toGetCapabilities. Both methods populateonvif.uriwith the URLs for media, PTZ, events, replay, and other services that subsequent requests use. - Media configuration (only when the device advertises a Media service) —
GetProfilesandGetVideoSourcesrun in parallel, thengetActiveSources()matches each video source to a suitable media profile. This setsactiveSource,defaultProfile, anddefaultProfiles, including encoder settings and PTZ configuration. Devices without video (for example Profile C door stations) skip this step and still completeconnect()successfully.
On success, connect() emits a connect event and returns the Onvif instance. Pass autoConnect: true
in the constructor to run this automatically after instantiation.
import { Onvif } from 'onvif';
const onvif = new Onvif({ hostname: '192.168.1.13', port: 8000, username: 'admin', password: 'admin' });
await onvif.connect();
const info = await onvif.device.getDeviceInformation();
console.log(info);const { Onvif } = require('onvif');
(async () => {
const onvif = new Onvif({ hostname: '192.168.1.13', port: 8000, username: 'admin', password: 'admin' });
await onvif.connect();
const info = await onvif.device.getDeviceInformation();
console.log(info);
})();import { Onvif } from 'onvif';
const onvif = new Onvif({ hostname: '192.168.1.13', port: 8000, username: 'admin', password: 'admin' });
await onvif.connect();
const info = await onvif.device.getDeviceInformation();
console.log(info);The 1.x package is organized so you only pay for the ONVIF services you actually use.
require('onvif') / import { Onvif } from 'onvif' loads a core set of modules (client, connection helpers,
events, discovery, utils) — on the order of ~100 KiB of compiled JS. Large service implementations such as
device, media, media2, ptz, recording, or advancedsecurity are not pulled in at import time.
Rough sizes of a few compiled service files (illustrative; exact numbers change with releases):
| Module | Approx. size |
|---|---|
Core (onvif + connection + events + utils + …) |
~95 KiB |
media.js |
~78 KiB |
media2.js |
~70 KiB |
device.js |
~39 KiB |
| Remaining service modules combined | ~200+ KiB |
Eagerly loading every service would put the initial JS footprint well over 400 KiB. With lazy loading, a typical camera client that only uses Device + Media + PTZ loads those modules on first use instead of at process start.
- Service namespaces (
onvif.device,onvif.media,onvif.ptz,onvif.thermal, …) are lazy proxies. The corresponding module is loaded the first time you call a method on it (for exampleawait onvif.ptz.getNodes()). connect()uses dedicated helpers inconnection.tsfor the handshake SOAP (GetServices/GetCapabilities, MediaGetProfiles/GetVideoSources). It does not load the fulldevice/media/media2class modules. Profiles and video sources are stored on theOnvifinstance (onvif.profiles,onvif.videoSources); whenMediais later loaded, it reuses that cache.Eventsis constructed eagerly (needed foronvif.on('event', …)). Everything else stays deferred.- The main package entry exports service classes as TypeScript types only, so CommonJS
require('onvif')does not force-loadRecording,Thermal, and similar modules just because they appear in the type surface.
import { Onvif } from 'onvif';
const onvif = new Onvif({ hostname: '192.168.1.13', username: 'admin', password: 'admin' });
await onvif.connect(); // handshake only — no full Media/Device class modules yet
const info = await onvif.device.getDeviceInformation(); // loads device.js on first use
const uri = await onvif.media.getStreamUri({ protocol: 'RTSP' }); // loads media.js on first use
// onvif.thermal is never loaded unless you call itIf you only need Discovery or Events, you can avoid Media entirely: Profile C / door-control style devices complete
connect() without a Media service, and unused namespaces stay unloaded for the lifetime of the process.
If you like the library, please fill out this form so we know which devices it supports. There are a lot of devices, and not all of them correctly support the ONVIF specification. We're trying our best to make it work with as many devices as possible, so your feedback is important to us.
@RogerHardiman tested this lib on a test bed with 5 x Axis, 2 x Bosch, 1 x Canon, 2 x Hanwha, 4 x HikVision, 1 x Panasonic, 2 x Sony and 2 x unknown vendor cameras. There is a mix of PTZ and Fixed cameras and a mix of Pre-Profile, Profile S, Profile G and Profile T devices.
But we want to learn about as wide a range of devices as possible. So yes, please leave your feedback, it is important
Just run console.log(await onvif.device.getDeviceInformation());, you will get something like this:
{
"manufacturer": "tp-link",
"model": "Tapo C220",
"firmwareVersion": "1.4.4 Build 260515 Rel.24570n",
"serialNumber": "7461572b",
"hardwareId": 1
}and put it here with your comments please: https://docs.google.com/forms/d/e/1FAIpQLSfXsVZv802YFDISGCZaLaJaC_isw2wKQpJ11UurvgO5veYzUw/viewform
Version 1.x introduces a new typed Onvif API. Compatibility modules are not re-exported from require('onvif') —
import them explicitly: onvif/compatibility (callbacks) or onvif/compatibility/promises (async/await).
const { Cam, Discovery } = require('onvif/compatibility');
const cam = new Cam(
{ hostname: '192.168.1.13', port: 8000, username: 'admin', password: 'admin' },
(error) => {
if (error) throw error;
cam.getDeviceInformation((err, info) => {
if (err) throw err;
console.log(info);
});
},
);See compatibility.cjs.
const { Cam, Discovery } = require('onvif/compatibility/promises');
const cam = new Cam({ hostname: '192.168.1.13', port: 8000, username: 'admin', password: 'admin' });
(async () => {
await cam.connect();
console.log(await cam.getDeviceInformation());
})();See compatibilityPromises.cjs.
The promisified Cam wraps the callback implementation: no auto-connect (call await cam.connect()),
methods return Promises, getters and EventEmitter APIs are forwarded, and _cam exposes the underlying instance.
Both compatibility entry points also export Discovery (callback or Promise probe), matching v0.x usage.
Discovered cams include xaddrs (all ProbeMatch XAddrs as URL[]).
The Cam surface from v0.8 is largely covered. Remaining behavioral differences:
gotoPresetaccepts both{ presetToken }(ONVIF / 1.x) and the v0 alias{ preset }(sent as PresetToken)rawResponsemay omit thestatusCodesecond argument that v0 emittedsetNTP(options)mutates the passedoptionsobject (fillsNTPManual) — same as v0.x Callbacks match v0.x(err, data, xml?): the third argument is the raw SOAP response XML from the underlying request (Onvif.lastResponseXml).getPresets/cam.presetsfollow token → preset (duplicate names kept; 0.8.1+ intent). Note: publishedonvif@0.8.2still returns name → preset from thegetPresetscallback while storing token → preset oncam.presets— compatibility aligns both with the token-keyed shape.
located in the Examples Folder on the Github
Tip
Not all of them were reworked for version 1.x.
- compatibility.cjs - v0.x callback API (
require('onvif/compatibility')): connect and printgetDeviceInformation - compatibilityPromises.cjs - v0.x Promise API (
require('onvif/compatibility/promises')): same, with async/await - events.with.filter.ts - ONVIF Events. With filters, pull-point, push-sub subscriptions
- example.js - Move camera to a pre-defined position then server the RTSP URL up via a HTTP Server. Click on the RTSP address in a browser to open the video (if you have the VLC plugin installed)
- example2.js - takes an IP address range, scans the range for ONVIF devices (brute force scan) and displays information about each device found including make and model and RTSP URLs For Profile S Cameras and Encoders it displays the default RTSP address For Profile G Recorders it displays the RTSP address of the first recording
- example3.js - reads the command line cursor keys and sends PTZ commands to the Camera
- example4.js - uses Discovery to find cameras on the local network
- example5.js - connect to a camera via SOCKS proxy. Note SSH includes a SOCKS proxy so you can use this example to connect to remote cameras via SSH
- example6.js - ONVIF Events. Example can be switched btween using Pull Point Subscriptions and using Base Subscribe with a built in mini HTTP Server
- example7.js - legacy v0.x Promise example (for 1.x use compatibilityPromises.cjs instead)
- example8.js - example setting OSD On Screen Display. (also uses Promises API)
To subscribe to all events using pull-point subscription you can just use .on() method, since the Onvif class
inherits from the EventEmitter class.
const onvif = new Onvif();
function eventHandler(msg) {
console.log(msg);
onvif.off('event');
}
onvif.on('event', eventHandler);If you need to subscribe to events, you can use the Subscription class. This class is for the specific subscriptions,
for example, when we need to subscribe to events from the camera with the filters. or add some more subscriptions
than the common one. It uses the pull-point subscription. It inherits from EventEmitter.
And emits two events: data and error. To use it you need to call subscribe() method. And to stop the device
subscription and remove all listeners you need to call unsubscribe() method.
The first and the only one argument for data is the NotificationMessage object. And an error raised only when
the connection to the device is lost.
await cam.connect();
const sub = new Subscription(cam, {
filter: {
topicExpression: [
{
expression: 'tns1:RuleEngine/CellMotionDetector/Motion',
dialect: 'http://www.onvif.org/ver10/tev/topicExpression/ConcreteSet',
},
],
},
});
sub.on('data', async (data) => {
console.log(new Date().toLocaleTimeString(), 'motion', data.topic._, data.message.message.data);
await sub.unsubscribe();
});
await sub.subscribe();For a full interactive example, see events.with.filter.ts.
This class is used internally by the Onvif class for the common event listener.
With push (WS-BaseNotification), the camera sends event notifications to an HTTP endpoint you host, instead of you polling the device.
To use it: start an HTTP server reachable from the camera, call subscribe with that URL as the consumer reference, keep the subscription alive with renew before it expires, and call unsubscribe when you are done. Method signatures are in the Events class documentation. A working flow is shown in events.with.filter.ts — the HTTP server at lines 50–65, and subscribe / unsubscribe at lines 143–164.
Interfaces are generated according to the latest version of the ONVIF specification.
All methods accept options defined by the ONVIF specification and return data from the corresponding <method_name>Response.
For example, the getCapabilities method accepts a single argument of type GetCapabilities and returns a result of type Capabilities.
Below is the internal structure of the GetCapabilitiesResponse type:
export interface GetCapabilitiesResponse {
/** Capability information. */
capabilities?: Capabilities;
}
class Device {
// ...
async getCapabilities(options?: GetCapabilities): Promise<Capabilities> {
// ...
}
// ...
}In general, the library tries to avoid returning objects that contain only a single property.
In some cases, where native JavaScript types are more convenient, interfaces are extended with additional fields.
For example:
SetSystemDateAndTimeSetSystemDateAndTimeExtended
The extended version adds a more convenient field:
export interface SetSystemDateAndTimeExtended extends SetSystemDateAndTime {
/**
* Javascript Date object to use instead of UTCDateTime
*/
dateTime?: Date;
// ...
}The ONVIF specifications include numerous extension points, which presents a challenge:
- on one hand, we want simple and convenient interfaces
- on the other hand, we need a unified mechanism for handling undocumented vendor-specific data
This data is usually provided through:
<xs:any namespace="##any" processContents="lax" minOccurs="0" maxOccurs="unbounded"/>This mechanism is important for:
- backward compatibility
- forward compatibility
- XML ↔ JavaScript transformation in ONVIF get/set methods
Suppose we have an ElementItem structure defined in:
Example schema:
<xs:element name="ElementItem" minOccurs="0" maxOccurs="unbounded">
<xs:annotation>
<xs:documentation>Complex value structure.</xs:documentation>
</xs:annotation>
<xs:complexType>
<xs:sequence>
<xs:any namespace="##any" processContents="lax">
<xs:annotation>
<xs:documentation>
XML tree containing the element value as defined
in the corresponding description.
</xs:documentation>
</xs:annotation>
</xs:any>
</xs:sequence>
<xs:attribute name="Name" type="xs:string" use="required">
<xs:annotation>
<xs:documentation>Item name.</xs:documentation>
</xs:annotation>
</xs:attribute>
</xs:complexType>
</xs:element>This becomes the following autogenerated TypeScript interface:
export interface ElementItem {
/** Item name. */
name: string;
/** XML tree containing the element value as defined in the corresponding description. */
[key: string]: unknown;
}For example, in MetadataConfiguration:
<Parameters>
<ElementItem>
<Name>elementItem1</Name>
<Param1>
<Data>42</Data>
</Param1>
<Param2>param2</Param2>
</ElementItem>
</Parameters>After parsing, the object looks like this:
{
elementItem : [{
name : 'elementItem1',
param1 : {
data : 42
},
param2 : 'param2',
__any__ : {
'Name' : ['elementItem1'],
'Param1' : [{
'Data' : ['42']
}],
'Param2' : 'param2'
}
}]
}This object contains:
- the required
namefield - parsed
xs:anyfields (param1,param2) - the raw
__any__field
The __any__ field contains the original unprocessed object returned by xml2js
A reasonable question is:
Why keep the raw XML structure?
The answer is simple.
When configuring ONVIF devices, extensions, or vendor-specific parameters, we often do not know how to serialize a clean JavaScript object back into the correct XML structure. At the same time, we still want to work with the data in a convenient way.
So when modifying device configuration (for example using setMetadataConfiguration), follow two simple rules:
elementItem[0].name = 'hello'elementItem[0].__any__.Param2 = 'hi'This structure can then be automatically converted back into the appropriate SOAP XML.
All tests are written using Jest.
Run them with:
npm testThe tests use happytime-onvif-server as a test device,
including integration suites for the v0.x compatibility layer (onvif/compatibility and onvif/compatibility/promises).
Golden suite __tests__/compatibility.golden.test.ts runs the same scenarios against npm onvif@0.8.2
(onvif-v0) and the master compatibility Cam, comparing callback args (err, data, xml), key result fields,
post-connect properties, and rawRequest / rawResponse events.
Thanks to HappyTimeSoft for providing the opportunity to test the full ONVIF specification.
Products are available here:
