Agent Config

Overview of all configuration options available for setting up and customizing agents in Millis. Learn how to manage voice settings, conversation flows, metadata inclusion, voicemail handling, and more.

prompt

string

required

The primary prompt used to guide the agent's behavior.

voice

object

required

Detailed voice configuration for the agent.

Show child attributes

voice.provider

enum<string>

required

The TTS provider for generating the agent's voice.

Available options:

elevenlabs,

openai,

cartesia,

playht,

deepgram,

rime,

neets,

openai_realtime,

dubverse

voice.voice_id

string

required

The unique identifier for the voice used by the agent.

voice.model

string | null

The model name provided by ElevenLabs (e.g., 'eleven_turbo_v2_5') for voice generation.

voice.settings

object | null

Additional voice_settings for ElevenLabs voice, such as stability and similarity boost values.

app_functions

object[] | null

Predefined functions tied to app integrations.

Show child attributes

app_functions.name

string

required

The name of the app-specific function.

app_functions.credentials

object | null

Credentials required for app integration.

Show child attributes

app_functions.credentials.{key}

string

call_settings

object | null

General settings for calls.

Show child attributes

call_settings.enable_recording

boolean

default:

false

Specifies whether call recording is enabled.

custom_llm_websocket

string | null

WebSocket URL for custom LLM integration.

custom_vocabulary

object | null

Custom vocabulary for improving recognition.

Show child attributes

custom_vocabulary.keywords

object

required

Custom keywords with their associated weights for ASR prioritization.

Show child attributes

custom_vocabulary.keywords.{key}

integer

extra_prompt_webhook

Webhook for providing extra prompts and metadata at the start of a conversation.

first_message

string | null

The agent's opening message for conversations.

flow

object | null

Settings for conversation flow management.

Show child attributes

flow.agent_terminate_call

object | null

Configuration for enabling the agent to terminate calls autonomously.

Show child attributes

flow.agent_terminate_call.enabled

boolean

required

Specifies whether the agent can end the call autonomously.

flow.agent_terminate_call.instruction

string | null

Custom instructions for the agent on when to terminate the call.

flow.agent_terminate_call.messages

string[] | null

List of messages the agent may use before ending a call.

flow.auto_fill_responses

object | null

Settings for predefined agent responses when there are response gaps.

Show child attributes

flow.auto_fill_responses.messages

string[]

required

Predefined messages for the agent to use when there are response gaps.

flow.auto_fill_responses.response_gap_threshold

integer

required

The time (in milliseconds) before the agent uses an auto-fill response.

flow.call_transfer

object | null

Configuration for transferring calls to another phone number.

Show child attributes

flow.call_transfer.phone

string

required

The phone number to transfer the call to.

flow.call_transfer.instruction

string | null

Instructions for the agent on how to handle the call transfer.

flow.call_transfer.messages

string[] | null

Predefined messages for the agent to say before initiating the call transfer.

flow.inactivity_handling

object | null

Settings for handling user inactivity during the call.

Show child attributes

flow.inactivity_handling.idle_time

number

required

Time in seconds of user inactivity before the agent prompts the user.

flow.inactivity_handling.message

string

required

Message the agent says after the initial idle time.

flow.interruption

object | null

Settings related to handling interruptions during the agent's responses.

Show child attributes

flow.interruption.allowed

boolean | null

Indicates if user interruptions are allowed during the agent's response.

flow.interruption.first_messsage

boolean | null

Defines if interruptions are allowed during the agent's initial message.

flow.interruption.keep_interruption_message

boolean | null

Specifies whether to keep the interrupted message in the response.

flow.response_delay

Configures delays for agent responses.

flow.user_start_first

boolean | null

Determines if the user speaks first to initiate the conversation.

flow.voicemail

object | null

Settings related to handling voicemail scenarios.

Show child attributes

flow.voicemail.action

enum<string>

required

The action to take when a voicemail is detected, e.g., hang up or leave a message.

Available options:

hangup,

message

flow.voicemail.continue_on_voice_activity

boolean | null

Determines if the agent continues the conversation on voice activity during voicemail handling.

flow.voicemail.message

string | null

The message to leave as voicemail if the agent opts to do so.

knowledge_base

object | null

Configuration for the agent's knowledge base.

Show child attributes

knowledge_base.files

string[] | null

List of files used as knowledge sources for the agent.

knowledge_base.messages

string[] | null

Predefined messages for agent to say while looking into knowledge base.

language

enum<string> | null

The default language for the agent.

Available options:

ln,

cy,

wo,

cs,

hy,

nl,

ceb,

la,

en-NZ,

ss,

lt,

ne,

fy,

oc,

yi,

tt,

bn,

fr-CA,

sw,

th,

ms,

ka,

pa,

fi,

ig,

af,

multi,

eu,

bs,

ps,

en-AU,

si,

ff,

de,

mt,

km,

pl,

nl-BE,

fr,

en-IN,

no,

gu,

lo,

sk,

ht,

tk,

bg,

mn,

su,

ko,

it,

id,

tl,

zu,

pt-BR,

jv,

et,

ca,

sl,

ja,

lb,

gl,

zh-TW,

hi,

ha,

th-TH,

lv,

ast,

sq,

tr,

ar,

sv,

haw,

tg,

lg,

ba,

mk,

vi,

hu,

el,

mi,

en-US,

kn,

az,

he,

hr,

ml,

kk,

es,

br,

sr,

xh,

uk,

as,

ilo,

ru,

gd,

yo,

da,

sa,

fa,

es-419,

bo,

ur,

my,

nn,

mo,

so,

mr,

tn,

sd,

sn,

uz,

de-CH,

fo,

en-GB,

zh,

ro,

pt,

or,

ga,

ta,

mg,

en,

be,

am,

te,

is

llm

object | null

Configuration for the LLM used by the agent.

Show child attributes

llm.model

enum<string>

required

The language model provider used for generating responses.

Available options:

gpt-3.5-turbo,

gpt-4-turbo,

gpt-4o-realtime-preview,

gpt-4o,

gpt-4o-mini,

llama-3-70b,

llama-3-1-8b,

llama-3-1-70b,

llama-3-1-405b,

mistral-large-2407,

l3.1-70b-euryale-v2.2

llm.temperature

number

default:

0

Temperature setting for response variability.

millis_functions

object[] | null

Millis-specific functions for agent customization.

Show child attributes

millis_functions.data

object

required

Data required for the function, including parameters.

Show child attributes

millis_functions.data.param

object

required

Details of a parameter for a Webform function.

Show child attributes

millis_functions.data.param.description

string

required

A description of the parameter's use in the function.

millis_functions.data.param.name

string

required

The name of the parameter.

millis_functions.data.param.required

boolean

required

Specifies if the parameter is mandatory.

millis_functions.data.param.type

string

required

The data type of the parameter.

millis_functions.description

string

required

A description of what the function does.

millis_functions.name

string

required

The name of the Webform function.

millis_functions.type

string

required

Type of the function (e.g., webform, webhook).

millis_functions.messages

string[] | null

Predefined messages for agent to say while executing the function.

privacy_settings

object | null

Settings for user privacy and data collection.

Show child attributes

privacy_settings.opt_out_data_collection

boolean

required

Indicates if the user wants to opt out of data collection.

privacy_settings.do_not_call_detection

boolean

default:

false

Specifies whether Do Not Call detection is enabled. When enabled, the agent will detect if the user wants to opt-out of future calls or be added to a Do Not Call list.

session_data_webhook

Webhook for receiving end of session data.

session_timeout

object | null

Settings related to session duration and idle time.

Show child attributes

session_timeout.max_duration

number | null

Maximum call duration (in seconds) before ending the session.

session_timeout.max_idle

number | null

Maximum idle time (in seconds) with no user input before ending the session.

session_timeout.message

string | null

Message to play when the session times out.

speech_to_text

object | null

Speech-to-text configuration for transcription.

Show child attributes

speech_to_text.multilingual

boolean | null

Indicates if the STT supports multiple languages at the same time.

speech_to_text.provider

enum<string> | null

The STT provider for transcribing speech into text.

Available options:

deepgram,

gladia,

talkscriber

switch_language

object | null

Settings for dynamically switching agent language.

Show child attributes

switch_language.languages

string[]

required

List of languages supported for dynamic switching during the conversation.

timezone

string | null

The timezone for the agent's operation.

tools

object[] | null

List of tools/functions the agent can use.

Show child attributes

tools.description

string

required

A description of the function's purpose.

tools.name

string

required

The name of the function.

tools.exclude_session_id

boolean | null

Indicates if the session ID should not be included in the webhook request.

tools.execute_after_message

boolean | null

Indicates if the function starts executing only after a predefined message is played.

tools.header

object | null

Headers to include in the webhook request.

Show child attributes

tools.header.{key}

string

tools.messages

string[] | null

Predefined messages for agent to say while executing the function.

tools.method

string | null

HTTP method to use for the webhook (e.g., POST, GET).

tools.params

object[] | null

List of parameters required by the function.

Show child attributes

tools.params.description

string

required

A description of the parameter.

tools.params.name

string

required

The parameter name for a function.

tools.params.required

boolean

required

Indicates if the parameter is required.

tools.params.type

string

required

The data type of the parameter.

tools.run_after_call

boolean | null

Specifies if the function should only run after the call ends.

tools.webhook

string | null

The webhook URL to trigger the function.

vad_threshold

number | null

Voice activity detection threshold for detecting speech.