Voice JS SDK v1: Changelog

Updated November 1, 2023

We have identified an issue on Chromium-based browsers running on MacOS 14 (Sonoma) where the audio deteriorates during a call. This issue happens due to the excessive calls to MediaDevices: enumerateDevices() API. With this release, the SDK calls this API only when necessary to avoid audio deterioration.

• Fixed a Chrome issue where audio is choppy when another application is also using the audio devices.

WebRTC API Overrides (Beta)

The SDK now allows you to override WebRTC APIs using the following options and events. If your environment supports WebRTC redirection, such as Citrix HDX's WebRTC redirection technologies, your application can use this new beta feature for improved audio quality in those environments.

• Device.Options.enumerateDevices - Overrides the native MediaDevices.enumerateDevices API.
• Device.Options.getUserMedia - Overrides the native MediaDevices.getUserMedia API.
• Device.Options.RTCPeerConnection - Overrides the native RTCPeerConnection class.
• connection.on('audio', handler(remoteAudio)) - Emitted after the HTMLAudioElement for the remote audio is created.

• Volume levels currently report as zero and do not reflect the actual volume levels of the call. This issue does not affect the quality of the call but, might affect application behavior if you rely on these values.
• Only inbound calls are currently supported. Outbound calls will be available in a future release.
• Using Device.audio.speakerDevices and Device.audio.setInputDevice will not properly switch speaker and microphone devices. You need to recreate the Device object or refresh the page when the default microphone or speaker device is changed.

Allow Connection specific ICE Servers

Developers can now opt to override rtcConfiguration set within Device.options per specific outgoing and incoming Connections.

To use this feature, a new parameter rtcConfiguration can be passed to Device.connect and Connection.accept. The function signatures are now described as below.

Passing the rtcConfiguration parameter to these functions will override any previously set rtcConfiguration within Device.options but not affect any other members set within Device.options.

• Added support for Tokyo and Sydney interconnect locations. These edges can be used by interconnect customers by specifying { edge: 'sydney-ix' } or { edge: tokyo-ix' } in Device.setup() or new Device() options.

1.13.0-beta2 has been promoted to 1.13.0 GA. Here's a summary of what's new in 1.13.0.

Voice diagnostics using runPreflight API

The SDK now supports a preflight test API which can help determine Voice calling readiness. The API creates a test call and will provide information to help troubleshoot call related issues. Please see the following for more details.

• API Docs
• Quick deploy App

Connection warning data

Connection.on('warning') now provides data associated with the warning. This data can provide more details about the warning such as thresholds and WebRTC samples collected that caused the warning. The example below is a warning for high jitter. Please see Voice Insights SDK Events Reference for a list of possible warnings.

Example output:

• Added high-packets-lost-fraction network warning . This new warning is raised when the average of the most recent seven seconds of packet-loss samples is greater than 3% . When the average packet-loss over the most recent seven seconds is less than or equal to 1% , then the warning is cleared.
• The behavior for raising the constant-audio-level warning has been updated. Now, the most recent ten seconds of volume values are recorded and then analyzed. If the standard deviation of these samples is less than 1% of the maximum audio value, then the warning is raised. When the standard deviation is greater than 1% and the warning has already been raised, then the warning is cleared.
• We now log an outgoing event to Insights when making an outbound call. This event also contains information whether the call is a preflight or not.
• Added a boolean field to the signaling payload for calls initiated by Device.runPreflight for debugging purposes.

• Fixed an issue introduced in Safari 13.1 that caused calls to continue playing after navigating away from the page.
• Fixed an issue where the disconnect sound plays after the caller cancelled the incoming call.

• Fixed an issue where an error is thrown if Device is imported and run in a NodeJS environment.

• The twilio.js SDK no longer supports the deprecated Edge Legacy browsers that rely on ORTC. See our deprecation notice for more details.

• We now log the selected ICE candidate pair to Insights. This will help with isolating issues should they arise.
• Fixed an issue where the transportClose event listener was not being cleaned up appropriately when a Connection is closed, causing MaxListenersExceededWarning on the console.

• Fixed an issue where calls on Safari 13.1.2 will intermittently fail and raise a 31003 error when establishing media connection. This usually happens when receiving or initiating the call.

• Fixed an issue where a device.on('disconnect') is emitted before raising a device.on('cancel') event. This usually happens when the caller cancels the incoming call before the SDK accepts it.
• Fixed an issue where sao-paolo is expected as an edge name instead of sao-paulo .

SHAKEN/STIR Verification Status for incoming calls

Twilio Client's Connection class now has Connection.callerInfo.isVerified, that can be used to display a trust indicator to the recipient when an incoming call, say from the public telephone network, has been verified under the SHAKEN/STIR framework.

A verified call that has been given highest attestation under SHAKEN/STIR means that the carrier that originated the call both (1) knows the identity of the caller, and (2) knows the caller has the right to use the phone number as the caller ID.

When your application receives a request webhook, that has the new StirStatus parameter all you have to do is <Dial><Client> and Twilio will implicitly pass the StirStatus to the Javascript Client.

CallerInfo

The Connection.callerInfo field returns caller verification information about the caller. If no caller verification information is available this will return null.

A CallerInfo provides caller verification information.

Attributes

• isVerified - Whether or not the caller's phone number has been attested by the originating carrier and verified by Twilio using SHAKEN/STIR. True if the caller has been verified at highest attestation 'A', false if the caller has been attested at any lower level or verification has failed.

Example:

Read here to learn more about making and receiving SHAKEN/STIR calls to/from the public telephone network.

Twilio Edge Locations

This release includes support for the expansion of Twilio's Global Infrastructure via Edge Locations which allows connectivity control into and out of Twilio's platform. The Voice Client JS SDK uses these Edges to connect to Twilio's infrastructure via the new parameter Twilio.Device.Options.edge. This new parameter supersedes the now deprecated Twilio.Device.Options.region. See Twilio.Device.Options.edge API documentation for migration instructions.

Example

**Support for automatic Edge location Fallback (Beta) ###

Deployments designed to connect to multiple Twilio Edge locations can take advantage of the new fallback mechanism. To enable the edge fallback, specify an array of edge names via Twilio.Device.Options.edge. When enabled and a connection failure is encountered, the SDK will reattempt the connection to the next region in the list. For more details about how the fallback works, refer to Edge Fallback documentation.

Example

Application Name and Version Logging Support

This release also introduces two new Device options: appName and appVersion. The values will be logged to Insights. These can be used to correlate other insights events with the application generating them. This is useful for debugging purposes in cases where multiple versions are deployed e.g. When performing A/B testing.

Support for Microsoft Edge Legacy is now deprecated. Running device.setup() on this browser will result with the console warning below. See this advisory for more information about important upcoming breaking changes with Microsoft Edge Legacy.

• Fixed an issue where rtcSample.rtt raised by Connection.on('sample', rtcSample => ...) was reported in seconds instead of milliseconds on Firefox and Chrome 81+. If your application converts rtcSample.rtt to milliseconds on these browsers, please remove the conversion.
• Fixed an issue where, in rare cases, call insights were generated after a call was disconnected.

• Fixed an issue where an Angular project will not build when the SDK is used as a npm module.

Also with this release, we addressed an issue where certain device event handlers, when an exception is thrown, causes some connection event handlers to stop working. This causes potential side effects such as incoming ringtone not being able to stop after receiving a call.

Example
In the following example, connection.on('accept') will not trigger if device.on('connect') throws an error. With this fix, connection.on('accept') handler should now receive the event.

Events affected
The following are the events affected and should be fixed with this release.

More information about NodeJS Events
As mentioned in our public documentation, the Device and Connection objects are EventEmitters. This release doesn't change the default behavior of EventEmitters, where if one of the handlers on the same EventEmitter object throws an exception, the rest of the event handlers will not receive the event. Consider the following example.

In the above example, testevent has three handlers and are on the same EventEmitter object myEmitter. If one of the handlers, in this case handler number 2, throws an error, the rest of the event handlers will not receive the event. In this case, handler 3 will not receive testevent. This is a normal behavior on EventEmitters and this SDK release doesn't change this behavior. This release only fixes the issue where if the events are coming from two different EventEmitter objects - Connection and Device.

Typescript declarations are now included with our NPM package. In the following example, Device, Connection, and their functions should have the correct typings.

• Fixed an issue where Device.on('incoming') event is not raised when the incoming sound is stopped right after playing it. This is a timing issue which can happen if multiple incoming connections comes in almost at the same time.
• Fixed an issue causing Android chrome to throw the error This browser does not support audio output selection . We now check if this is supported on the browser before attempting to update the output device.

Added the ability to access the SDK logger instance using the loglevel npm module. Please refer to the loglevel documentation for a list of logger APIs.

For example, to set the log level:

Also in 1.10.0

• https://sdk.twilio.com is now being used for serving the sound files. You may need to update your firewall rules and allow list AWS CloudFront IPs . Please refer to AWS IP address ranges documentation for more details.
• Updated npm dependencies to support node version 12.
• We now log RTCDtlsTransport state changes to Insights. This will help with isolating issues should they arise.

Added an experimental feature for Chrome browser to enable Aggressive ICE Candidate Nomination. This feature can be enabled by setting forceAggressiveIceNomination to true. If your deployment is on devices with one network interface and your RTT to Twilio's Servers is typically greater than 96 milliseconds, this feature may help reduce call connect time. As this is an experimental feature, we don't recommend enabling this until after testing it thoroughly in your deployment.

Example:

• New improvements to media reconnection . ICE restart is now also requested when ICE gathering fails (transitions to complete and no ICE candidates were gathered), or ICE gathering exceeds 15 seconds and no ICE candidates were gathered.
• Locally gathered ICE candidates are now logged for debugging purposes.

• You can now connect to our interconnect region in Singapore by setting the region option to sg1-ix . See Twilio Client Regions for the list of supported regions. Note that with this release, to support new regions without requiring an SDK update, we have removed the check for the region name passed to Device.setup . If an unsupported region is supplied, Device.on('error') will be called.
• We now log PeerConnection state changes to Insights. This will help with isolating issues should they arise.

• We now report audioInputLevel and audioOutputLevel within the last second in the connection sample object Connection.on('sample', handler(sample)) .

• Update querystring to clear cached audio files without CORS headers.
• Fixed an issue where constant-audio-input-level warning is not being emitted.

• Fixed an issue where audio files sometimes shows CORS errors on the console.

• Fixed an issue on Safari where sample.mos , emitted from Connection.on('sample', handler(sample)) , is always null.

• The Device sounds are now cached. They are only downloaded when Device.setup() is invoked.

Max Average Bandwidth API

By default, the Opus codec is set up with a transmission rate of around 32 kbps (40-50kbps on the wire). With this release, you are able to set a custom max average bitrate to better control how much bandwidth your VoIP application should use. See RFC-7587 section 7.1 for information about Max Average Bitrate.

The main purpose of this API is to set a lower max average bitrate to minimise bandwidth usage. This is particularly useful in deployments where bandwidth is at a premium. Where bandwidth is not of concern, you do not need to use this API. Max Average Bitrate can be set to as low as 6,000bps and as high as 51,000 bps. Values outside this range are ignored and the default Opus operation mode is used. See API Docs for more information.

As would be expected, lowering the max average bitrate impacts audio quality. We don't recommend setting max average bitrate to a value below 8,000 bps. On the other hand, setting values over 32,000 bps will have negligible audio quality improvements.

This is currently not supported in Firefox due to this bug.

Example, to set a new max average bitrate to 16,000 bps:

• Fixed an issue causing multiple devices that are created in the same tab to get disconnected when one of the devices disconnects a connection.

• Fixed an issue causing audio levels to be reported as zero when running as an extension, or when the browser tab is inactive or minimized.
• Fixed an issue causing Connection.status() to return pending instead of closed after calling Connection.reject() .

Media Reconnection States and Events

This feature, when enableIceRestart is enabled, allows for detecting when media connection fails which will trigger automatic media reconnection, and for detecting when media connection is restored.

New Events

We've added two new events that will fire when enableIceRestart is enabled:

• Connection.on('reconnecting', handler(error)) - raised when media connection fails and automatic reconnection has been started by issuing ICE Restarts. During this period, Connection.status() will be set to reconnecting.
  • error - Error object { code: 53405, message: 'Media connection failed.' }
  • Media reconnection triggers
    • ICE Connection state transitions to disconnect and bytes sent and received in the last 3 seconds is zero.
    • ICE Connection state or PeerConnection state transitions to failed . Only Chrome browser will attempt an ICE restart with this trigger. Other browsers will immediately disconnect the call and raise an error 31003 . This is due to browsers not fully supporting connection states during an ICE restart.
• Connection.on('reconnected', handler()) - raised when media connection has been restored which is detected when media starts flowing. Once reconnected, Connection.status() will be set to open .

Retries

ICE restarts will be retried in the event that previous ICE restarts are unsuccessful. Retry attempts will happen when ICE Connection state or PeerConnection state transitions to failed. If more than 30 seconds has elapsed during this transition, the call will disconnect and raise an error 31003.

• Added de1-ix to valid list of regions.
• Added Device.version to return sdk version
• When applicable, errors emitted through Device.on('error') and Connection.on('error') now contain a twilioError field, providing more information about the error. This twilioError represents the new TwilioError format that will become the default Error format in 2.0.

• Fixed an issue causing local environment information to get bundled into the build artifact in local npm builds.
• Fixed an issue where ringing will not stop when network is disconnected.

• Added de1-ix to valid list of regions.

• Fixed an issue causing local environment information to get bundled into the build artifact in local npm builds.

Automatic Media Reconnection

This feature was first introduced in 1.7.4 and was enabled by default. With this release, we have introduced the enableIceRestart reconnect flag to enable or disable Automatic Media Reconnection. The default is disabled. This will allow you to transition your code to utilise this feature. Example usage:

Twilio.Device.setup(TOKEN, { enableIceRestart: true });

• The console now shows a more descriptive error if WebRTC support isn't detected when not loaded over https.

• Fixed an issue where active connection was getting disconnected when the token expired.
• Fixed an issue where an answer during ICE reconnection was applied without a valid offer, resulting in the console error: Failed to set remote answer sdp: Called in wrong state: kStable .

• Checking whether plan-b or unified-plan is default on the browser now happens on Device.setup() or on device initialization with a token, instead of on page load. This offloads creation of an RTCPeerConnection until we know Device will actually be used on the page.

• Fixed an issue where ICE restarts would continue to retry when a call got disconnected while ringing.
• Device.destroy() now properly disconnects all connections.
• Fixed an issue where an answer was applied multiple times after creating an offer, causing console errors.
• Fixed an issue where low-bytes warning was raised if total bytes sent and received was zero or not supported, potentially causing some issues in Edge preventing audio from being heard after an ICE restart.
• Fixed an issue where ICE restart would not stop when connection drops on Firefox, causing console errors.

Updated July 16, 2019

The introduction of Automatic Media Reconnection in 1.7.4 is enabled by default. This functionality may affect program flow if you rely on Device.on('error', …) with error code 31003 to update your UI or reconnect logic. This error is not thrown at the time of media interruption any longer. It is now sent after ICE restart is attempted and fails which may take 10s of seconds.

Automatic Media Reconnection

A call may be inadvertently disconnected when media is temporarily lost. With this release, we will attempt to reconnect the media before dropping the call with a process known as ICE restart.

If you are relying on Device.on('error', …) with error code 31003, to update your UI or to initiate a reconnect, you will need to update your code to use Device.on('offline', …) instead. The 31003 error code may not be reported for some time as ICE restarts are continually attempted.

Updated July 16, 2019

We have identified this as a potential breaking change and we will include an opt-in feature switch in the 1.7.6 release. We apologize for any inconvenience this may have caused you.

• ICE Connections will now attempt to reconnect when they transition to the failed state, rather than immediately disconnecting.
• We now report bytesSent and bytesReceived within the last second in the webrtc sample object ( RTCSample ).
• We now begin monitoring for warnings 5 seconds after the start of a call (originally at 20 seconds).

• Fixed a bug where changing the input device then later calling Connection.mute() will not work.
• Fixed a bug causing some signaling errors to not trigger an error event from Connection.

The following known issues will be addressed in an upcoming 1.7.5 release:

• ICE restarts may continue after successfully reconnecting, or a call has ended. This is benign, but may cause errors to be visible in the browser console.
• ICE restarts may begin after 5 seconds in Edge 42, causing audio loss for the local client. This does not affect Edge 44.

• We now report audio codec and whether DSCP is enabled to Insights Metrics.
• Added new getter on Connection, Connection.codec , which will be populated with the audio codec used in the call as soon as the SDK receives that information from WebRTC. We currently do not get the audio code from FireFox.
• We now emit a webrtc sample object ( RTCSample ) every second through a new event, Connection.on('sample') .

• Fixed a bug causing the input stream to not be released after a Connection is created without calling Device.audio.setInputDevice and then later calling Device.audio.setInputDevice during the call when using a browser that supports the unified-plan SDP semantic.
• Fixed an issue where audio ring tone plays on a different output device after reconnecting an external output device.

• Fixed an issue where some audio resources weren't being released after a call. Thank you to 西口瑛一(Eiichi Nishiguchi) 土橋雅一(Masakazu Tsuchihashi) and Leverages Co., Ltd. for helping us isolate this issue.

Unified Plan support

• Added support for Unified Plan.
• With this version, twilio-client.js uses the browser's default SDP format. Unified Plan SDP format is the default format used by Google Chrome 72+, Safari 12.2+, and Firefox since forever.
• Note If you are using twilio-client.js versions 1.7.0 or older, changing audio input devices during a call will break on Safari 12.2 onwards.
• See this advisory for SDP format migration impact.

Updated the algorithm used to report the "Audio input level" and "Audio output level" Insight metrics. The levels are obtained directly from an AudioContext and are no longer read out from webrtc's legacy stats.

Opus and Codec Preference API

Opus and PCMU are the two codecs now offered and accepted by Twilio Client JS.

For 1.7 and all further 1.x releases, G.711 (PCMU) will continue to be the preferred codec offered. This is to avoid any potential breaking changes.

The following code illustrates how to make Opus the default codec:

The option codecPreferences can be passed in when instantiating a Device instance or when calling Device.setup to set the codecs preference. The default is set to [‘pcmu', ‘opus']

• Fixed a DSCP bug, which was causing the dscp flag to be ignored in Chrome due to a breaking change introduced in Chrome M72.

• Added device.audio.setAudioConstraints() and device.audio.unsetAudioConstraints(). These methods allow setting a MediaTrackConstraints object to be applied to every time device.audio.setInputDevice() is called, and any time an active input device is lost and the SDK gets new user media to fall back to another input device. If an input device is already set via device.audio.setInputDevice(), these methods will immediately call setInputDevice() internally and return the resulting Promise, otherwise they will return a resolved Promise. The currently set audio constraints can be seen on the new read-only field, device.audio.audioConstraints, which defaults to null.

  Example:

• Applied a workaround to a Chrome regression introduced in M72 affecting speakerDevices.test() and ringtoneDevices.test().

Browser Support

• We are now using RTCP values for RTT when available. Initially, this will not affect Chrome because Chrome has not yet implemented support . Additionally, having access to RTT will allow Firefox to calculate and report MOS, however FireFox is currently affected by a regression causing jitter to be reported as 0, which may make MOS scores appear slightly higher until it's patched by Mozilla.

• Fixed a regression introduced in 1.6.0 causing falsey TwiML params (undefined, null, 0, false) to be stripped.
• Twilio.Device.audio.disconnect() will now toggle whether the disconnect sound should play while already on an active call.

• Now sending ice gathering state change events to Insights as ice-gathering-state.
• Increased default websocket backoff maximum from 3000ms to 20000ms, and added the backoffMaxMs option to Device.setup() options that takes a time in milliseconds to override this default. The minimum allowable value is 3000ms. Example: Device.setup(token, { backoffMaxMs: 45000 });

• Updated ws dependency to latest. Only affects npm package because the CDN artifact of twilio.js uses the browser's WebSocket implementation.

• Fixed an issue introduced in 1.6.3 preventing metrics from being published to Insights.

• Added NetworkInformation data to a new Insights event, network-change , which is sent first when a new Connection is established, and then any time thereafter the network information changes. See the support browser chart here .

• Stopped sending Insights events that aren't associated with a CallSid or TempCallSid as they can't be tracked.

• Added Device.setup option { fakeLocalDTMF: true } that uses imitation DTMF sounds instead of the default real DTMF sounds, preventing an issue where DTMF tones would sometimes register twice.

• Updated default RTCConfiguration object to ensure forward compatibility when plan-b is deprecated.
• Added an rtcConfiguration field to IDeviceOptions, which takes an RTCConfiguration object that gets passed to any created RTCPeerConnections. Example:

• Re-assigning the master output device mid-call, particularly when calling device.disconnectAll() , will no longer throw an exception or pause script execution.
• Fixed closeProtection feature
• Added EventEmitter interface ( on , addListener , removeListener , etc...) to Device singleton.

• Added the ability to receive and handle incoming calls while on an active call behind a new flag, allowIncomingWhileBusy . When set to true, Device's default behavior of silently ignoring the incoming call is removed, and the incoming call will instead cause Device to emit an "incoming" event. If accepted, the prior active call will be immediately disconnected, and the incoming call will be accepted, replacing the prior active call.

• Added support for custom incoming parameters from TwiML as Map<string, string> Connection.customParameters . When a TwiML application sends custom parameters using the <Parameter> noun, these parameters will be added to Connection.customParameters . For example:

Note that the following restrictions apply to the Parameter noun:

• Parameter name must be a string, up to 32 bytes.
• Parameter value must be a string, up to 128 bytes.
• Up to 8 parameters can be sent per dial.

• Updated default RTCConfiguration object to ensure forward compatibility when plan-b is deprecated.

• Updated Insights logic to point to new endpoint, fully supporting Insights when using Access Tokens

• Twilio.Device may now be instantiated multiple times via

• Twilio.Device.setup() may now be called with Access Tokens, in addition to Capability Tokens.
• Twilio.Device.destroy() will now completely clear out the Device, allowing Device.setup() to be called with a new set of options

• We now ensure all Audio resources are cleaned up after closing a Connection

• Twilio Client is moving toward the standard EventEmitter interface, meaning events should be managed with .on(eventName, handler) and .removeListener(eventName, handler) , replacing our legacy handlers (such as .accept(handler) , .error(handler) , etc...). The following methods are deprecated:

These have been replaced with the following EventEmitter events:

Note that there is no Connection#ignore event. The Connection.ignore(handler) method is actually a backward-compatible listener for the Connection.on('cancel', handler) event.

• Updated default RTCConfiguration object to ensure forward compatibility when plan-b is deprecated.

• Fixed an issue causing disconnected sound to not play by default.

• Updated AudioPlayer dependency to the newly published @twilio/audioplayer package.

• Fixed a regression introduced by 1.4.31 (rolled back on Apr 18) causing region passed to Device.setup to be ignored.

• Fixed a WebSocket issue occasionally resulting in "Cannot register. Token not validated." errors.

• Added Twilio.Device.isSupported flag which should be true if WebRTC or ORTC is supported. If false , it's an indicator that Device.setup should not be called (it will throw an unsupported exception).
• Fixed an issue where an internal listener wasn't being cleaned up when disconnecting a call, resulting in EventEmitter warnings when opening more than 10 calls in a row.

• Fixed a build issue causing the twilio-client npm package to fail when run through webpack or browserify.

• The JS Twilio Client SDK has been added to npm! https://www.npmjs.com/package/twilio-client
• Added licenses to both artifacts (js and min.js)

• Added a fallback for when posting metrics to Insights fails, so that a warning is logged in the console rather than an exception being thrown.

• Fixed an issue causing Insights timestamps to be undefined and breaking application flow in some environments.
• Fixed an issue causing one-way audio for all calls after the first when using Device.audio.setInputDevice()
• Fixed an issue causing sounds not to play correctly in Chrome when a non-default output device was selected by Device.audio.speakerDevices or Device.audio.ringtoneDevices .

• Fixed an issue causing input audio to be lost after calling Device.audio.setInputDevice('default') and then removing and replacing the USB headset.
• Fixed an issue causing outbound calls to break after the final audio output device was removed and replaced.
• Removed circular references so that the Connection object can be serialized.
• Fixed an issue causing WebSockets to become unrecoverable when the token expired while the socket is offline or disconnected.

• Added support for the answerOnBridge feature of TwiML by adding a new ringing Connection state and Connection#ringing event behind the flag: Twilio.Device.setup(token, { enableRingingState: true }). With both the answerOnBridge property and the enableRingingState flag enabled in the consuming application, the Connection state is now be more granular:
  • ringing state is transitioned to when the callee is notified of the incoming call
  • open state is transitioned to when the callee has accepted the call

• Fixed an issue causing the Device.sounds deprecation warning to be logged every time the library was loaded.
• Fixed an uncommon issue causing "Cannot call register of undefined" errors when destroying the Device.

• Removed babel libraries, which were causing increased file size and clashing with the same libraries included in consuming applications.
• Enabled volume support, as Safari's support of AudioNode is now stable.

• Added experimental support for Safari 11. All functionality has been enabled in Safari 11, with a couple caveats:
  • Audio output selection and availabeOutputDevices are disabled as Safari does not yet support HTMLAudioElement.setSinkId
  • The volume APIs are temporarily disabled in Safari as a related WebAudioAPI feature was sometimes causing tabs to crash in our testing. We will be re-evaluating this and looking to re-enable this feature going forward.

• Fixed a bug causing the ringtones of multiple incoming calls to stack, and then continuing to ring after one is answered.
• Fixed a bug causing Electron applications to report one-way audio.
• Fixed a bug causing additional websockets to be created when Device.setup is called while the existing websocket is temporarily disconnected.
• Fixed a bug causing a tab to play the incoming ringtone indefinitely in the case where an unfocused tab receives an incoming call, and is then focused after that call is canceled.

• Insights token should now correctly update when Device.setup is called with a new token.
• Any changes made to the input/output device selection while a call is incoming should now always be respected.

• Fixed a bug in Chrome causing 0 jitter to be reported as undefined, and MOS to be reported as 1

• Fixed a bug causing the wrong RTCStats timestamps to be used, sometimes resulting in seemingly missing Insights data.

• Fixed a race condition causing the incoming ringtone to continue ringing after a call had already been accepted or rejected.
• Completely removed incoming call sound duration. Once an incoming call is accepted or rejected, the ringtone will stop immediately in all cases.

• Added support for the new interconnect region aliases: ie1-ix (Ireland), us1-ix and us2-ix

• Fixed a bug causing constant audio input level warnings to be raised in newer Chrome versions
• Minimum incoming call sound duration has been reduced from 2000ms to 100ms

• Reverted behavior of Device.activeConnection() to return the first received incoming call when there is no active, ongoing call. This behavior was inadvertently changed in 1.4.15. In the next breaking release, this behavior will be changed so that Device.activeConnection() correctly only returns the active connection if one exists.

• Fixed a bug causing device to ring instead of being ignored when a new call comes in during an ongoing call after re-registering with a new token.

• Removed ws dependency, we instead use the native Websocket now. This should make file size smaller, and make the library fully es5-compatible.
• Implemented support for standard WebRTC statistics, maintaining backward compatibility with google-prefixed stats.

• Added support for German region de1

• Fixed a bug sometimes causing "Twilio not defined" failures in Edge.
• Fixed Connection.mute() so that the call stays muted after switching input devices.

• Updated Connection.sendDigits to use the new RTCDTMFSender in Firefox 52 and above.

• Added connection.getRemoteStream() and connection.getLocalStream() to retrieve the local/remote streams being used in the Connection.

• A disconnected input device should now be usable by Device.audio.setInputDevice() after being reconnected.

• Twilio's ORTC adapter will now be used over the native WebRTC components introduced in Edge build 15019. This distinction will remain until Edge's native WebRTC support satisfies all of Twilio Client's dependencies.
• The incoming ringtone sound should now continue playing when one of multiple incoming calls is rejected, ignored or canceled and other incoming calls remain.

• Input Device Fallback - When an active input device is lost, we will now fall back to a backup input device on Chrome and Edge. Firefox does not currently provide a way to detect when an active input is lost.

• [Edge] Switching input devices is now fully supported in Edge.
• [FF] Due to lack of support for opening multiple audio input devices, support for the Device.audio.setInputDevice has been completely disabled in Firefox.
• [FF] Additionally, helpful error messages have been added to Twilio Client's getUserMedia request failures to help identify when user media requests fail due to Firefox's audio input device limitation.
• See the bottom of the 1.4 Overview page for a complete chart of browser compatibility.

• Device.audio.unsetInputDevice() will no longer throw an error if called synchronously inside the Device.incoming() handler.
• Tracks will now always properly be stopped. As long as Device.audio.inputDevice is null and there is no active call, all user media should be properly released.
• Device.audio.setInputDevice() will now work properly when the client is the caller or callee.
• When the incoming sound fails to load or play, the Device.incoming handler will still be fired after a maximum timeout of 2 seconds. Pausing execution (by adding an alert or prompt on Device.incoming) should no longer prevent the incoming ringing sound from playing. Note that in Chrome, the incoming sound will only play once as the paused script execution will prevent looping behavior.
• Exceptions should no longer be logged to the console whenever a new device is found.

• Added a new interconnect region: ie1-tnx

• Removed a legacy dependency that contained eval statements; no more should be present.

• Device.setup now properly updates Device with the new token data when called.
• Connection#warning will now be emitted, instead of Connection#error , when ICE liveliness checks fail. This serves to differentiate between fatal and non-fatal ICE events.

• Audio Output Selection - Using the new Device.audio API , developers may now specify which audio output device(s) to play Client sounds through, in browsers that support the feature (As of writing, Chrome 49 and above). The incoming ringtone sound is separated into its own audio channel, allowing it to be played out of a different set of output devices than in-call audio if desired.
• ** Audio Input Selection** - Additionally, the Device.audio API allows developers to swap the input device during an ongoing call, or set the input device and test it before making a call.
• Volume Indicators - Twilio Client's Connection class now reports the input and output volumes of the call, making it easier for developers to show real-time volume indicators in their UI.
• Custom Sounds - Twilio Client now supports passing custom sounds to Device.setup to be used in place of Twilio's default sounds.
• Twilio Interconnect (TNX) Support - Twilio Client now supports connecting through Twilio Interconnect (TNX) for certain regions. As of writing, connecting to regions us1 and us2 on TNX connection provisioning is supported in Client via us1-tnx and us2-tnx region parameters, respectively.

• Quality Warnings - Twilio Client's Connection class now emits real-time warnings when it detects potential call quality problems, and emits a warning cleared event when that issue clears up.
• User Feedback - Twilio Client's Connection class now accepts user feedback on call quality through the Connection.postFeedback API.

• Device.sounds - We have added deprecation warnings for Device.sounds, and moved its functionality over to the new Device.audio API .

• Close protection will now continue to work after the active token has expired.
• Connection.accept will no longer throw a breaking exception when called twice (Though we still recommend only calling this once!)

• Insights token should now correctly update when Device.setup is called with a new token.

• Fixed a bug in Chrome causing 0 jitter to be reported as undefined, and MOS to be reported as 1
• Fixed a bug causing the wrong RTCStats timestamps to be used, sometimes resulting in seemingly missing Insights data.

• Added support for the new interconnect region aliases: ie1-ix (Ireland), us1-ix and us2-ix
• Added support for German region 'de1'

• Added support for standard RTCStats object

• Updated Connection.sendDigits to use the new RTCDTMFSender in Firefox 52 and above.

• The constant-audio-output-level is now reported as an INFO level event, rather than WARNING in order to reduce false positives. This event may fire if the other side of the call is muted.

• Device.setup now properly updates Device with the new token data when called.
• ICE liveliness checks will now generate an appropriate Insights event when the connection is lost or regained.

• Increased the delay at beginning of a call before enabling monitoring for network issues to counteract the newer versions of Chrome which cause RTT to be reported abnormally high for the first 10-15 seconds of a call, which was causing false positives to be thrown.

• Fixed some internal logic causing ignored-by-local event to be logged multiple times in Insights
• Added / modified other Insights fields to improve logging.

• Added browser specific error information for microphone access failure (31201).

• Close protection will now still work correctly after token has expired, if a call is active.
• Added additional fields to Insights metrics payloads.

• Removed an erroneous field from the payload format that gets sent to Twilio Insights.

• Adjusted call quality warning events that get sent to Twilio Insights.

• Added more lifecycle events to be sent to Twilio Insights. Adjusted existing lifecycle events to new schema.

• Added RTC monitoring and warnings to the events that are sent to Twilio Insights.

• Miscellaneous bug fixes regarding lifecycle event logging.
• Further improvements to internal call quality metrics logging.

• Improvements to internal call quality metrics logging.

• Fixed asset loading in Chrome extension.

• Region selection - Developers can now choose to connect their users to a specific Twilio Client data center, or continue to allow Twilio's Global Low Latency routing to select the lowest-latency path. See the documentation for more information.
• Static IP addresses for Twilio Client media - Twilio Client now establishes media connections from a static range of IP addresses. See the Twilio Client documentation for a list of the IP address ranges that Twilio is currently using. See the documentation for the list of IP address ranges.
• Support for ORTC in Microsoft Edge beta releases - Microsoft Edge beta releases now support ORTC . Now you can use Twilio Client in Microsoft Edge to make and receive calls without Flash.
• DSCP support - Twilio Client 1.3 enables DSCP by default in browsers that support it.

• Presence - We have removed support for Presence in twilio.js 1.3.
• Flash support - We have removed support for Flash in twilio.js 1.3.

• The error passed to error() 's handler function no longer contains the info property. This property previously contained the flash.net.NetConnection object.
• Twilio.Device.setup() no longer accepts rtc in its params argument. Twilio Client will use WebRTC or ORTC if either API is available in the browser, Flash is no longer an option.
• presence() has been removed. Presence is longer supported in Twilio Client.
• getMediaEngine() has been removed, since Flash is no longer supported.

• Presence - Added deprecation warnings to Presence functionality, as these features have been removed in twilio.js 1.3.
• Flash support - Added deprecation warnings to Flash functionality, as this feature has been removed in twilio.js 1.3.

• No longer raises the "MediaStream.stop is deprecated" warning in Google Chrome
• No longer raises the "MediaStream.ended is deprecated" warning in Google Chrome

• Allow twilio.js and other Twilio JavaScript SDKs to be included on the same page
• sendDigits uses RTCDTMFSender where supported

• Resolved an issue where Chrome 45 users were occasionally unable to accept calls.

• Improved remote troubleshooting abilities

• Resolved an issue where Flash users may have been unable to place or accept calls when using the simplePermissionDialog option.
• Compatibility code enabling support for versions of Chrome older than 26 has been rewritten so as not to affect non-standard WebRTC implementations.

• On ICE disconnect, raise an error "ICE liveness checks failed. May be having trouble connecting to Twilio." with error code 31003. This can occur when the browser believes a previously active connection is no longer properly connected. This can happen due to changing network conditions, etc.
• On ICE failure, raise an error "ICE negotiation with Twilio failed. Call will terminate." with error code 31003, and terminate the call. This can occur when the browser is unable to connect to Twilio's WebRTC gateways. Common causes for this error include firewall rules which block connectivity to Twilio's media servers or firewall rules which block outgoing UDP.
• Log additional client-side WebRTC statistics for better troubleshooting call quality issues: audioInputLevel and audioOutputLevel .

• Provide a more useful warning if too many event handlers are registered.
• Fix a bug where internal event handlers were not being unregistered.

• Add a new audioConstraints setting to select a specific microphone, or turn off features like auto-gain control.
• Collect client-side WebRTC statistics for troubleshooting call quality issues.
• Add a new function, Twilio.Device.getMediaEngine() , for checking whether WebRTC or Flash is being used.

• No longer raise an exception when the capability token expires if an error handler has been registered on Twilio.Device .