ResponseCreateParams interface
Create a new VoiceLive response with these parameters
Properties
| append |
Input items to append to the conversation context before generating a response. |
| cancel |
Whether to cancel any ongoing generation before starting this one. Defaults to true. |
| commit | Whether to commit the response to the conversation. Defaults to true. |
| input |
Input items to be used as the context for this response. An empty array clears previous context. |
| instructions | The default system instructions (i.e. system message) prepended to model calls. This field allows the client to guide the model on desired responses. The model can be instructed on response content and format, (e.g. "be extremely succinct", "act friendly", "here are examples of good responses") and on audio behavior (e.g. "talk quickly", "inject emotion into your voice", "laugh frequently"). The instructions are not guaranteed to be followed by the model, but they provide guidance to the model on the desired behavior. Note that the server sets default instructions which will be used if this
field is not set and are visible in the |
| interim |
Configuration for interim response generation during latency or tool calls. |
| max |
Maximum number of output tokens for a single assistant response,
inclusive of tool calls. Provide an integer between 1 and 4096 to
limit output tokens, or |
| metadata | Set of up to 16 key-value pairs that can be attached to an object. This can be useful for storing additional information about the object in a structured format. Keys can be a maximum of 64 characters long and values can be a maximum of 512 characters long. |
| modalities | The set of modalities the model can respond with. To disable audio, set this to ["text"]. |
| output |
The format of output audio. Options are |
| pre |
Create the response with pre-generated assistant message. The message item would be added into the conversation history and returned with synthesized audio output in the created response. |
| reasoning |
Constrains effort on reasoning for reasoning models. Check model documentation for supported values for each model. Reducing reasoning effort can result in faster responses and fewer tokens used on reasoning in a response. |
| temperature | Sampling temperature for the model, limited to [0.6, 1.2]. Defaults to 0.8. |
| tool |
How the model chooses tools. Options are |
| tools | Tools (functions) available to the model. |
| voice | supported voice identifiers and configurations. |
Property Details
appendInputItems
Input items to append to the conversation context before generating a response.
appendInputItems?: ConversationRequestItemUnion[]Property Value
cancelPrevious
Whether to cancel any ongoing generation before starting this one. Defaults to true.
cancelPrevious?: booleanProperty Value
boolean
commit
Whether to commit the response to the conversation. Defaults to true.
commit?: booleanProperty Value
boolean
inputItems
Input items to be used as the context for this response. An empty array clears previous context.
inputItems?: ConversationRequestItemUnion[]Property Value
instructions
The default system instructions (i.e. system message) prepended to model calls. This field allows the client to guide the model on desired responses. The model can be instructed on response content and format, (e.g. "be extremely succinct", "act friendly", "here are examples of good responses") and on audio behavior (e.g. "talk quickly", "inject emotion into your voice", "laugh frequently"). The instructions are not guaranteed to be followed by the model, but they provide guidance to the model on the desired behavior.
Note that the server sets default instructions which will be used if this
field is not set and are visible in the session.created event at the
start of the session.
instructions?: stringProperty Value
string
interimResponse
Configuration for interim response generation during latency or tool calls.
interimResponse?: InterimResponseConfigProperty Value
maxOutputTokens
Maximum number of output tokens for a single assistant response,
inclusive of tool calls. Provide an integer between 1 and 4096 to
limit output tokens, or inf for the maximum available tokens for a
given model. Defaults to inf.
maxOutputTokens?: number | "inf"Property Value
number | "inf"
metadata
Set of up to 16 key-value pairs that can be attached to an object. This can be useful for storing additional information about the object in a structured format. Keys can be a maximum of 64 characters long and values can be a maximum of 512 characters long.
metadata?: Record<string, string>Property Value
Record<string, string>
modalities
The set of modalities the model can respond with. To disable audio, set this to ["text"].
modalities?: string[]Property Value
string[]
outputAudioFormat
The format of output audio. Options are pcm16, g711_ulaw, or g711_alaw.
outputAudioFormat?: stringProperty Value
string
preGeneratedAssistantMessage
Create the response with pre-generated assistant message. The message item would be added into the conversation history and returned with synthesized audio output in the created response.
preGeneratedAssistantMessage?: AssistantMessageItemProperty Value
reasoningEffort
Constrains effort on reasoning for reasoning models. Check model documentation for supported values for each model. Reducing reasoning effort can result in faster responses and fewer tokens used on reasoning in a response.
reasoningEffort?: stringProperty Value
string
temperature
Sampling temperature for the model, limited to [0.6, 1.2]. Defaults to 0.8.
temperature?: numberProperty Value
number
toolChoice
How the model chooses tools. Options are auto, none, required, or
specify a function, like {"type": "function", "function": {"name": "my_function"}}.
toolChoice?: stringProperty Value
string