IraVoice selects an available IraVoice cluster based on configured capacity allocation and queues the call request in that cluster. The call_params object in the request payload controls CPA behavior, media streaming, webhook delivery, inactivity handling, and recording.

Call Params

Example

Code
{
    "stream_on_cpa_events": ["LV"],
    "drop_on_cpa_events": ["AM", "FX"],

    "websocket_host": "bot.example.com",
    "websocket_port": 8443,
    "websocket_app": "voicebot",

    "streaming_useraudio": true,
    "stream_frame_size_ms": 200,

    "stt_sampling_rate": 16000,
    "tts_sampling_rate": 16000,

    "denoise": "WebRTC NS",

    "bot_inactivity_limit": 30,

    "event_url": "https://events.example.com/calls",

    "disable_recording": false
}

Call Progress Analysis (CPA)

Applicable only for outbound calls when cpa_config is included in the makecall request. IraCPA analyzes the initial call audio to classify the answer type and detect machine-generated responses, tones, silence, and screening behavior. The first CPA event is typically detected within the first 1.75 seconds of answered audio.

Primary CPA Events

Code	Meaning
LV	Live human
AM	Answering machine
FX	Fax machine
SIT	Special Information Tones
IBP	Immediate beep detected immediately after answer

Depending on the configured CPA break_event behavior, analysis may continue after the initial classification to detect additional events.

Extended CPA Events

Code	Meaning
SL	Silence
BP	Beep detected after initial answer classification
EAM	Extended Answering Machine detection after ringtone progression
CS	Call screening detected

Automatic actions can be triggered using:

• stream_on_cpa_events
• drop_on_cpa_events

To learn more about IraCPA, click here

Parameter	Type	Description
stream_on_cpa_events	array<string>	Starts WebSocket media streaming when the detected CPA event matches any configured value. Default: ["LV"]
drop_on_cpa_events	array<string>	Terminates the call when the detected CPA event matches any configured value. Default: ["FX", "AM"]

WebSocket Streaming

Streams bidirectional audio between IraVoice and the configured WebSocket endpoint.

• Audio from the call is streamed to the WebSocket server.
• Audio received from the WebSocket server is injected into the live call.

The WebSocket message format is documented in the WebSocket Message Model section.

Parameter	Type	Description
websocket_host	string	Hostname of the WebSocket server. Requires websocket_port and websocket_app.
websocket_port	integer	WebSocket server port. Used only when websocket_host is configured.
websocket_app	string	Path component.
streaming_useraudio	boolean	Controls upstream audio delivery mode. true → streams audio continuously. false → buffers audio till speech ends and sends it as a WAV file.
stream_frame_size_ms	integer	Duration of each audio chunk in milliseconds. Default: 200 Minimum: 20 Chunking is time-based, not RTP packet-based.
stt_sampling_rate	integer	Sampling rate (Hz) for audio streamed IraVoice → AI orchestration layer. Range: 8000-48000
tts_sampling_rate	integer	Sampling rate (Hz) for audio streamed AI orchestration layer → IraVoice. Range: 8000-48000

Example WebSocket endpoint construction

Input	Result
websocket_host = bot.example.com websocket_port = 8443 websocket_app = voicebot	wss://bot.example.com:8443/voicebot

Bot Inactivity Handling

Terminates the call if no audio is received from the WebSocket server for a continuous duration defined by bot_inactivity_limit. Disabled when:

• bot_inactivity_limit is omitted, or
• bot_inactivity_limit = 0

Parameter	Type	Description
bot_inactivity_limit	integer	Maximum allowed inactivity duration in seconds before the call is dropped. Default: 0 (disabled)

Webhooks & Events

Parameter	Type	Description
event_url	string	HTTP webhook endpoint that receives call lifecycle events via POST. Events include started, ringing, answered, hangup, and related state transitions. If configured, events are always emitted.

Recording

Parameter	Type	Description
disable_recording	boolean	Disables call recording when set to true. Default: false (recording enabled).

Noise Suppression

Parameter	Type	Description
denoise	string	WebRTC NS - Lightweight, standard noise suppression optimized for high-volume scaling. Hush - Premium, AI-driven speaker isolation offering superior clarity at a higher compute cost.

Critical Constraints

Topic	Details
Transport Security	Only wss:// endpoints are supported. ws:// connections are rejected. The WebSocket server must present a valid SSL certificate.
WebSocket Protocol	Refer to the WebSocket Message Model section for the complete message schema.
Webhook Payload	Refer to the Webhook Event URL Payload section for webhook request structure.

Similar to makecall. But only to be used if you know exactly which cluster you want the call to go through.

Example usecase: Transfering / Bridging voicebot call to a human agent. When you trigger the first outbound call, the response JSON will have the cluster and call_uuid. Save the cluster value for later use.
If the voicebot requires to send call to human agent, call makecalldirect API with the cluster set to that of the original call.
The campaign_name will be the campaign exclusively created to send calls to human agents through suitable SIP trunk.

Drops the call.
Checks if the call_uuid supplied is of a call initiated via the Distributor.
If yes, it looks up which cluster the call was placed in and sends the dropcall request to that cluster.

Stops the websocket stream and disconnects the websocket connection.
Checks if the call_uuid supplied is of a call initiated via the Distributor.
If yes, it looks up which cluster the call was placed in and sends the stop stream request to that cluster.

Example usecase: Stop the stream to voicebot before transfering or bridging the call with human agent.

Bridges two calls in the same cluster.
Checks if the first_uuid and second_uuid supplied are of a calls initiated via the Distributor and if both calls are in the same cluster.
If yes, LD sends the new bridge request to that cluster.

Transfers the call to the destination specified.
If the deflect is set to true, then SIP REFER will be sent to the trunk through which the call was made.
If the deflect is set to false, then the call transfer is internal within IraSwitch, for example to an extension or to a queue.