This module is a promise-based client ('inbound' event socket) and server ('outbound' event socket) for FreeSwitch, written entirely in Javascript with no dependencies on the libesl library.
This module is actively maintained and used in production systems.
This is version 11, a new major version of esl
.
It introduces TypeScript support, and gets rid of binding to this
.
This mode of operation is called "inbound" (to FreeSwitch) in the Event Socket FreeSwitch documentation.
A client can be used to trigger calls asynchronously (for example in a click-to-dial application).
A client can also be used to monitor events for known UUIDs or other fields (see the .filter(header,value)
method).
A server will handle calls sent to it using the socket
diaplan application (called "outbound" mode in the Event Socket Outbound FreeSwitch documentation).
The server is available at a pre-defined port which the socket
dialplan application will specify.
The following code does the equivalent of fs_cli -x
: it connects to the Event Socket, runs a single command, then disconnects.
import { FreeSwitchClient, once } from 'esl'
const client = new FreeSwitchClient({
port: 8021
})
const fs_command = async (cmd) => {
const p = once(client,'connect')
await client.connect()
const [ call ] = await p
const res = await call.api(cmd)
await call.exit();
await client.end();
}
fs_command("reloadxml");
Generally speaking though, the client might reconnect multiple times, and your code should handle reconnections:
import { FreeSwitchClient, once } from 'esl'
const client = new FreeSwitchClient({
port: 8021
})
client.on('connect', (call) => {
// Do something here with the API
})
The FreeSwitchClient
constructor takes a single argument, an options object
with the following fields:
host
: defaults to127.0.0.1
port
: defaults to 8021password
: defaults toClueCon
logger
: defaults to theconsole
object
The FreeSwitchClient
class has the following methods.
This method triggers the connection to FreeSWITCH.
The client will automatically reconnect if FreeSWITCH crashes or the connection is lost.
This methods closes the connection to FreeSWITCH and prevents further attempts.
Returns a Promise.
The FreeSwitchClient
class may emit the following events.
Sent when an error is reported.
Sent when connecting to FreeSWITCH. Might be sent multiple times in case disconnections happen.
Sent when disconnected from FreeSWITCH. The retry
value indicates how long the
client will wait until reconnecting to FreeSWITCH.
Sent by the underlying socket when a socket-level warning is triggered.
Sent when the end()
method is called.
You can connect to an Event Socket server from the FreeSwitch XML dialplan, Notice the syntax to specify more than one server if desired.
<action application="set" data="socket_resume=true"/>
<action application="socket" data="127.0.0.1:7000|127.0.0.1:7001 async full"/>
<action application="respond" data="500 socket failure"/>
Another option is to configure a inbound profile to directly use the socket.
This bypasses the XML dialplan; instead, an inline
dialplan is used.
<profile name="my-sofia-profile">
<settings>
<param name="dialplan" value="inline:'socket:127.0.0.1:7000|127.0.0.1:7001 async full'"/>
Here is a simplistic event server:
import { FreeSwitchServer } from 'esl'
const server = new FreeSwitchServer()
server.on('connection', (call) => {
const res = await call.command('playback', 'voicemail/vm-hello')
const foo = res.body.variable_foo
await call.hangup() // hang-up the call
await call.exit() // tell FreeSwitch we're disconnecting
})
await server.listen({ port: 7000 })
The FreeSwitchServer
constructor takes a single argument, an options object
with the following fields:
-
all_events
: boolean, defaults totrue
; indicates whether the FreeSwitchResponse object should request all events from FreeSWITCH (the default), or only the ones required to process commands (all_events:false). Note that the default will negatively impact performance of both FreeSWITCH and your application; it however provides the simplest onboarding. -
my_events
: boolean, defaults totrue
; indicates whether the FreeSwitchResponse object should filter on the Unique-ID of the call. This is generally what one wants, there is generally no reason to set this tofalse
. (If you want to monitor system-wide events you should probably use a FreeSwitchClient instance.) -
logger
: defaults to theconsole
object
The FreeSwitchClient
class has the following methods.
This method starts accepting connection from FreeSWITCH.
The options are the same as for server.listen
in the Node.js net
package:
port
, host
, backlog
, …
Returns a Promise.
This methods closes the connection to FreeSWITCH and prevents further attempts.
Returns a Promise.
This method returns a Promise for the number of currently opened connections.
const count = await server.getConnectionCount()
console.log(`There are ${count} connections left opened.)
The FreeSwitchServer
class may emit the following events.
Sent when an error is reported.
Sent when an incoming connection is dropped.
Sent when FreeSWITCH connects to Node.js.
This event receives two parameters:
- the first one is a FreeSwitchResponse instance you will use to process the call;
- the second one contains data received during the initial connection.
Both FreeSwitchServer
and FreeSwitchClient
accept a logger
option which
must provide logger.debug
, logger.info
, and logger.error
.
If logger.debug
is not required, it can be set to an no-op function:
const logger = {
debug: () => {},
info: (...args) => console.info(...args),
error: (...args) => console.error(...args)
}
The FreeSwitchResponse
class is the one you will interact most. It allows you
to interact with FreeSWITCH using both low-level (Event Socket) commands and
higher-level (API) commands.
The FreeSwitchResponse
class extends EventEmitter.
Returns the unique identifier used internally to reference this instance.
Send a bgapi
(background API) command to FreeSwitch and wait for completion.
Different FreeSWITCH modules provide different commands, consult the documentation
of each module to know which commands it provides. Inside the FreeSWITCH CLI use
show api
and show application
to get the list of registered commands.
bgapi
will wait until the commands completes before returning its Promise.
This migh be multiple hours if the command initiates a call.
The timeout
parameter has no default. If a timeout is not provided, the
Promise might never get fulfilled.
Might thow FreeSwitchError
.
async api(command: string, timeout?: number) : Promise<{ uuid: string, body: StringMap, headers: StringMap }>
Send an api
command to FreeSwitch.
Different FreeSWITCH modules provide different commands, consult the documentation
of each module to know which commands it provides. Inside the FreeSWITCH CLI use
show api
and show application
to get the list of registered commands.
Returns a Promise that is fulfilled as soon as FreeSwitch sends a reply. Requests are queued and each request is matched with the first-coming response, since there is no way to match between requests and responses.
On the FreeSWITCH side, api
command block the Event Socket until they respond.
This is probably not what you want if using FreeSwitchClient
, you should use
bgapi
in that case.
Also use bgapi
if you need to make sure responses are matched properly, since
it provides the proper semantics.
The timeout defaults to the value of .default_send_timeout()
, i.e. 10s.
Might thow FreeSwitchError
.
These methods are identical; you would typically use command
in a
FreeSwitchServer application, and command_uuid
in a FreeSwitchClient
application.
Execute a dialplan application synchronously — returns a Promise that completes when the command is completed (which may take hours).
// Send the command and wait for completion
await call.command('playback', '/tmp/example.wav')
execute_uuid(uuid:string,app_name:string,app_arg:string,loops?:number,event_uuid?:string) : SendResult
These methods are identical; you would typically use execute
in a
FreeSwitchServer application, and execute_uuid
in a FreeSwitchClient
application.
Execute a dialplan application asynchronously — does not wait for completion.
In most cases you probably want to use command
or command_uuid
instead of
execute
and execute_uuid
.
// Send the command
await call.execute('playback', '/tmp/example.wav')
These methods are identical; you would typically use hangup
in a
FreeSwitchServer application, and hangup_uuid
in a FreeSwitchClient
application.
Hangs up the call.
unicast(args: {'local-ip':string, 'local-port':number, 'remote-ip':string, 'remote-port':number, transport:'tcp'|'udp', flags?:'native'}) : SendResult
unicast_uuid(uuid:string,args:{'local-ip':string, 'local-port':number, 'remote-ip':string, 'remote-port':number, transport:'tcp'|'udp', flags?:'native'}) : SendResult
These methods are identical; you would typically use unicast
in a
FreeSwitchServer application, and unicast_uuid
in a FreeSwitchClient
application.
Interface media with the specified IP and port.
local-ip
: default to 127.0.0.1local-port
: default to 8025remote-ip
: default to 127.0.0.1remote-port
: default to 8026- flags:
native
— do not transcode audio to/from the FreeSWITCH internal format (L16)
Add the specified events to the list of events forwarded to Node.js.
By default this module already executes
call.event_json('CHANNEL_EXECUTE_COMPLETE', 'BACKGROUND_JOB')
, or, with the
all_events
flag of FreeSwitchServer, call.event_json('ALL')
.
call.event_json('CHANNEL_HANGUP_COMPLETE','DTMF')
Remove the specified events from the list of events forwarded to Node.js.
Removing CHANNEL_EXECUTE_COMPLETE
and BACKGROUND_JOB
will break
command
/command_uuid
and bgapi
, respectively.
Stop receiving events.
Using this method will prevent command
/command_uuid
and bgapi
from
working.
Add an event filter for the specified event header and value.
Remove an event filter for the specified event header and value.
Enqueue an event in the FreeSWITCH event queue.
Requires the full
flag when sending to FreeSwitchServer.
Used in server mode, requests FreeSwitch to not close the socket as soon as the call is over, allowing us to do some post-processing on the call (mainly, receiving call termination events).
By default, FreeSwitchServer
with call exit()
for you after 4 seconds.
You must capture the cleanup_linger
event if you want to handle things differently.
Enable logging on the socket, optionally setting the log level.
Disable logging.
Send a message on the socket.
The command is one of the low-level call-command
documented for the Event
Socket interface.
In most cases you should use one of the provided methods (api
, bgapi
, etc.) rather than try to implement this.
Write a command to the Event Socket and wait for the (low-level) reply.
In most cases you should use one of the provided methods (api
, bgapi
, etc.) rather than try to implement this.
The FreeSwitchResponse
class may emit different events.
By default in FreeSwitchServer, all_events
is true
and your code will
receive the different events for the call.
You might also activate additional events in FreeSwitchClient using
the event_json()
method.
The event callback will receive a single argument, an object with two fields:
headers
: the headers of the Event Socket eventbody
: the content sent by FreeSWITCH
Both are Object
.
import { FreeSwitchServer } from 'esl'
const server = new FreeSwitchServer()
server.on('connection', (call) => {
// Only triggered once. `onceAsync` returns a Promise and might throw.
call.onceAsync('CHANNEL_ANSWER').then( function ({headers,body}) {
console.log('Call was answered');
});
// Might be triggered multiple times.
call.on('CHANNEL_ANSWER', function({headers,body}) {
console.log('Call was answered');
});
// By default `all_events` is true and we do not need to use `event_json`.
})
await server.listen({ port: 7000 })
Emitted when the underlying network socket is closed.
Emitted when the unerlying network socket has an error.
Emitted when a write on the underlying network socket has an error.
Emitted when the underlying socket was terminated due to an error.
Emitted when FreeSWITCH did not provide a Content-Type header.
Should normally not happen, most probably a bug in FreeSWITCH if this happens.
Emitted when the parser received an unsupported Content-Type header from FreeSWITCH.
Should normally not happen, report these as bug!
Emitted when the JSON received from FreeSWITCH could not be parsed.
Emitted when the FreeSWITCH response could be parsed but no Event-Name is found.
Emitted when you activated .linger()
and it's time for your code to call
.exit()
.
Emitted when you activated .log()
and a log event is received.
Emitted by FreeSWITCH to indicate imminent disconnection of the socket.
Undocumented rejection from FreeSWITCH.
Add the module to your project using npm
, yarn
, pnpm
.
npm install esl
The test suite provides many examples.
Please use GitHub issues for community support.
Commercial support is available as well from the maintainer.
- creating client and server now uses
new
and theFreeSwitchClient
,FreeSwitchServer
classes this
is no longer used; thecall
object is passed as a parameter.