One of the following:
ResponseCreate { type, access_programs, background, 34 more }

Client event for creating a response over a persistent WebSocket connection. This payload uses the same top-level fields as POST /v1/responses, plus WebSocket-only envelope metadata.

Notes:

  • stream is implicit over WebSocket and should not be sent.
  • background is not supported over WebSocket.
  • stream_id is WebSocket-only and is not part of POST /v1/responses.
type: "response.create"

The type of the client event. Always response.create.

access_programs?: AccessPrograms { cyber }

Domain-specific access programs to use for this request.

cyber?: "standard" | "daybreak_blue" | "daybreak_red"

The Cyber access program to use for this request. Supported values are standard, daybreak_blue, and daybreak_red. If omitted, the API resolves the program from the model’s Cyber tier and your organization and project access, subject to model-specific eligibility restrictions. By default, models without a Cyber tier use Standard. Blue-tier models use Daybreak Blue when authorized; otherwise they fall back to Standard unless the model requires Daybreak access. Red-tier models use Daybreak Red and require authorization. Requests that require unavailable Daybreak access return 403. An implicit Standard fallback is represented by null in the response’s access_programs field, rather than an explicit Standard selection.

One of the following:
"standard"
"daybreak_blue"
"daybreak_red"
background?: boolean | null

Whether to run the model response in the background. Learn more.

context_management?: Array<ContextManagement> | null

Context management configuration for this request.

minItems1
type: string

The context management entry type. Currently only ‘compaction’ is supported.

compact_threshold?: number | null

Token threshold at which compaction should be triggered for this entry.

minimum1000
conversation?: string | BetaResponseConversationParam { id } | null

The conversation that this response belongs to. Items from this conversation are prepended to input_items for this response request. Input items and output items from this response are automatically added to this conversation after this response completes.

One of the following:
string
BetaResponseConversationParam { id }

The conversation that this response belongs to.

id: string

The unique ID of the conversation.

generate?: boolean

Whether to generate output for this response. Defaults to true when omitted. Set to false to prepare request state without model output; the response ID can be used as previous_response_id on a later turn.

image?: Image { context_compression }

Request level image options.

context_compression?: "auto" | "enabled" | "disabled"

Controls image context optimization. Defaults to auto, which uses the model, transport, and input to select a policy. Use enabled to require compression support or disabled to turn it off.

One of the following:
"auto"
"enabled"
"disabled"
include?: Array<BetaResponseIncludable> | null

Specify additional output data to include in the model response. Currently supported values are:

  • web_search_call.action.sources: Include the sources of the web search tool call.
  • code_interpreter_call.outputs: Includes the outputs of python code execution in code interpreter tool call items.
  • computer_call_output.output.image_url: Include image urls from the computer call output.
  • file_search_call.results: Include the search results of the file search tool call.
  • message.input_image.image_url: Include image urls from the input message.
  • message.output_text.logprobs: Include logprobs with assistant messages.
  • reasoning.encrypted_content: Includes an encrypted version of reasoning tokens in reasoning item outputs. This enables reasoning items to be used in multi-turn conversations when using the Responses API statelessly (like when the store parameter is set to false, or when an organization is enrolled in the zero data retention program).
One of the following:
"file_search_call.results"
"web_search_call.results"
"web_search_call.action.sources"
"message.input_image.image_url"
"computer_call_output.output.image_url"
"code_interpreter_call.outputs"
"reasoning.encrypted_content"
"message.output_text.logprobs"
input?: string | BetaResponseInput { , , , 33 more }

Text, image, or file inputs to the model, used to generate a response.

Learn more:

One of the following:
string
BetaResponseInput = Array<BetaResponseInputItem>

A list of one or many input items to the model, containing different content types.

One of the following:
BetaEasyInputMessage { content, role, phase, type }

A message input to the model with a role indicating instruction following hierarchy. Instructions given with the developer or system role take precedence over instructions given with the user role. Messages with the assistant role are presumed to have been generated by the model in previous interactions.

content: string | BetaResponseInputMessageContentList { , , }

Text, image, or audio input to the model, used to generate a response. Can also contain previous assistant responses.

One of the following:
string
BetaResponseInputMessageContentList = Array<BetaResponseInputContent>

A list of one or many input items to the model, containing different content types.

One of the following:
BetaResponseInputText { text, type, prompt_cache_breakpoint }

A text input to the model.

text: string

The text input to the model.

type: "input_text"

The type of the input item. Always input_text.

prompt_cache_breakpoint?: PromptCacheBreakpoint { mode, ttl }

Marks the exact end of a reusable prompt prefix. When ttl is omitted, the breakpoint inherits the request’s prompt_cache_options.ttl; the boundary is not rounded to a token block.

mode: "explicit"

The breakpoint mode. Always explicit.

ttl?: "30m" | "12h"

The minimum lifetime of this cache breakpoint. Defaults to the request TTL.

One of the following:
"30m"
"12h"
BetaResponseInputImage { detail, type, file_id, 2 more }

An image input to the model. Learn about image inputs.

The detail level of the image to be sent to the model. One of high, low, auto, or original. Defaults to auto.

type: "input_image"

The type of the input item. Always input_image.

file_id?: string | null

The ID of the file to be sent to the model.

image_url?: string | null

The URL of the image to be sent to the model. A fully qualified URL or base64 encoded image in a data URL.

formaturi
prompt_cache_breakpoint?: PromptCacheBreakpoint { mode, ttl }

Marks the exact end of a reusable prompt prefix. When ttl is omitted, the breakpoint inherits the request’s prompt_cache_options.ttl; the boundary is not rounded to a token block.

mode: "explicit"

The breakpoint mode. Always explicit.

ttl?: "30m" | "12h"

The minimum lifetime of this cache breakpoint. Defaults to the request TTL.

One of the following:
"30m"
"12h"
BetaResponseInputFile { type, detail, file_data, 4 more }

A file input to the model.

type: "input_file"

The type of the input item. Always input_file.

detail?: "auto" | "low" | "high"

The detail level of the file to be sent to the model. Use auto to let the system select the detail level; for GPT-5.6 and later models, auto uses high-quality rendering, which may increase input token usage. Use low for lower-cost rendering, or high to render the file at higher quality. Defaults to auto.

One of the following:
"auto"
"low"
"high"
file_data?: string

The content of the file to be sent to the model.

file_id?: string | null

The ID of the file to be sent to the model.

file_url?: string

The URL of the file to be sent to the model.

formaturi
filename?: string

The name of the file to be sent to the model.

prompt_cache_breakpoint?: PromptCacheBreakpoint { mode, ttl }

Marks the exact end of a reusable prompt prefix. When ttl is omitted, the breakpoint inherits the request’s prompt_cache_options.ttl; the boundary is not rounded to a token block.

mode: "explicit"

The breakpoint mode. Always explicit.

ttl?: "30m" | "12h"

The minimum lifetime of this cache breakpoint. Defaults to the request TTL.

One of the following:
"30m"
"12h"
role: "user" | "assistant" | "system" | "developer"

The role of the message input. One of user, assistant, system, or developer.

One of the following:
"user"
"assistant"
"system"
"developer"
phase?: "commentary" | "final_answer" | null

Labels an assistant message as intermediate commentary (commentary) or the final answer (final_answer). For models like gpt-5.3-codex and beyond, when sending follow-up requests, preserve and resend phase on all assistant messages — dropping it can degrade performance. Not used for user messages.

One of the following:
"commentary"
"final_answer"
type?: "message"

The type of the message input. Always message.

Message { content, role, agent, 2 more }

A message input to the model with a role indicating instruction following hierarchy. Instructions given with the developer or system role take precedence over instructions given with the user role.

A list of one or many input items to the model, containing different content types.

role: "user" | "system" | "developer"

The role of the message input. One of user, system, or developer.

One of the following:
"user"
"system"
"developer"
agent?: Agent | null

The agent that produced this item.

agent_name: string

The canonical name of the agent that produced this item.

status?: "in_progress" | "completed" | "incomplete"

The status of item. One of in_progress, completed, or incomplete. Populated when items are returned via API.

One of the following:
"in_progress"
"completed"
"incomplete"
type?: "message"

The type of the message input. Always set to message.

BetaResponseOutputMessage { id, content, role, 4 more }

An output message from the model.

id: string

The unique ID of the output message.

content: Array<BetaResponseOutputText { annotations, text, type, logprobs } | BetaResponseOutputRefusal { refusal, type } >

The content of the output message.

One of the following:
BetaResponseOutputText { annotations, text, type, logprobs }

A text output from the model.

annotations: Array<FileCitation { file_id, filename, index, type } | URLCitation { end_index, start_index, title, 2 more } | ContainerFileCitation { container_id, end_index, file_id, 3 more } | FilePath { file_id, index, type } >

The annotations of the text output.

One of the following:
FileCitation { file_id, filename, index, type }

A citation to a file.

file_id: string

The ID of the file.

filename: string

The filename of the file cited.

index: number

The index in the output text at which to insert the file citation.

type: "file_citation"

The type of the file citation. Always file_citation.

URLCitation { end_index, start_index, title, 2 more }

A citation for a web resource used to generate a model response.

end_index: number

The index of the last character of the URL citation in the message.

start_index: number

The index of the first character of the URL citation in the message.

title: string

The title of the web resource.

type: "url_citation"

The type of the URL citation. Always url_citation.

url: string

The URL of the web resource.

formaturi
ContainerFileCitation { container_id, end_index, file_id, 3 more }

A citation for a container file used to generate a model response.

container_id: string

The ID of the container file.

end_index: number

The index of the last character of the container file citation in the message.

file_id: string

The ID of the file.

filename: string

The filename of the container file cited.

start_index: number

The index of the first character of the container file citation in the message.

type: "container_file_citation"

The type of the container file citation. Always container_file_citation.

FilePath { file_id, index, type }

A path to a file.

file_id: string

The ID of the file.

index: number

The index of the file in the list of files.

type: "file_path"

The type of the file path. Always file_path.

text: string

The text output from the model.

type: "output_text"

The type of the output text. Always output_text.

logprobs?: Array<Logprob>
token: string
bytes: Array<number>
logprob: number
top_logprobs: Array<TopLogprob>
token: string
bytes: Array<number>
logprob: number
BetaResponseOutputRefusal { refusal, type }

A refusal from the model.

refusal: string

The refusal explanation from the model.

type: "refusal"

The type of the refusal. Always refusal.

role: "assistant"

The role of the output message. Always assistant.

status: "in_progress" | "completed" | "incomplete"

The status of the message input. One of in_progress, completed, or incomplete. Populated when input items are returned via API.

One of the following:
"in_progress"
"completed"
"incomplete"
type: "message"

The type of the output message. Always message.

agent?: Agent | null

The agent that produced this item.

agent_name: string

The canonical name of the agent that produced this item.

phase?: "commentary" | "final_answer" | null

Labels an assistant message as intermediate commentary (commentary) or the final answer (final_answer). For models like gpt-5.3-codex and beyond, when sending follow-up requests, preserve and resend phase on all assistant messages — dropping it can degrade performance. Not used for user messages.

One of the following:
"commentary"
"final_answer"
BetaResponseFileSearchToolCall { id, queries, status, 3 more }

The results of a file search tool call. See the file search guide for more information.

id: string

The unique ID of the file search tool call.

queries: Array<string>

The queries used to search for files.

status: "in_progress" | "searching" | "completed" | 2 more

The status of the file search tool call. One of in_progress, searching, incomplete or failed,

One of the following:
"in_progress"
"searching"
"completed"
"incomplete"
"failed"
type: "file_search_call"

The type of the file search tool call. Always file_search_call.

agent?: Agent | null

The agent that produced this item.

agent_name: string

The canonical name of the agent that produced this item.

results?: Array<Result> | null

The results of the file search tool call.

attributes?: Record<string, string | number | boolean> | null

Set of 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, and querying for objects via API or the dashboard. Keys are strings with a maximum length of 64 characters. Values are strings with a maximum length of 512 characters, booleans, or numbers.

One of the following:
string
number
boolean
file_id?: string

The unique ID of the file.

filename?: string

The name of the file.

score?: number

The relevance score of the file - a value between 0 and 1.

formatfloat
text?: string

The text that was retrieved from the file.

BetaResponseComputerToolCall { id, call_id, pending_safety_checks, 5 more }

A tool call to a computer use tool. See the computer use guide for more information.

id: string

The unique ID of the computer call.

call_id: string

An identifier used when responding to the tool call with output.

pending_safety_checks: Array<PendingSafetyCheck>

The pending safety checks for the computer call.

id: string

The ID of the pending safety check.

code?: string | null

The type of the pending safety check.

message?: string | null

Details about the pending safety check.

status: "in_progress" | "completed" | "incomplete"

The status of the item. One of in_progress, completed, or incomplete. Populated when items are returned via API.

One of the following:
"in_progress"
"completed"
"incomplete"
type: "computer_call"

The type of the computer call. Always computer_call.

A click action.

actions?: BetaComputerActionList { , , , 6 more }

Flattened batched actions for computer_use. Each action includes an type discriminator and action-specific fields.

agent?: Agent | null

The agent that produced this item.

agent_name: string

The canonical name of the agent that produced this item.

ComputerCallOutput { call_id, output, type, 4 more }

The output of a computer tool call.

call_id: string

The ID of the computer tool call that produced the output.

minLength1
maxLength64
output: BetaResponseComputerToolCallOutputScreenshot { type, file_id, image_url }

A computer screenshot image used with the computer use tool.

type: "computer_call_output"

The type of the computer tool call output. Always computer_call_output.

id?: string | null

The ID of the computer tool call output.

acknowledged_safety_checks?: Array<AcknowledgedSafetyCheck> | null

The safety checks reported by the API that have been acknowledged by the developer.

id: string

The ID of the pending safety check.

code?: string | null

The type of the pending safety check.

message?: string | null

Details about the pending safety check.

agent?: Agent | null

The agent that produced this item.

agent_name: string

The canonical name of the agent that produced this item.

status?: "in_progress" | "completed" | "incomplete" | null

The status of the message input. One of in_progress, completed, or incomplete. Populated when input items are returned via API.

One of the following:
"in_progress"
"completed"
"incomplete"
One of the following:
One of the following:
BetaResponseFunctionToolCall { arguments, call_id, name, 7 more }

A tool call to run a function. See the function calling guide for more information.

arguments: string

A JSON string of the arguments to pass to the function.

call_id: string

The unique ID of the function tool call generated by the model.

name: string

The name of the function to run.

type: "function_call"

The type of the function tool call. Always function_call.

id?: string

The unique ID of the function tool call.

agent?: Agent | null

The agent that produced this item.

agent_name: string

The canonical name of the agent that produced this item.

async?: boolean

Whether the function tool call runs asynchronously.

caller?: Direct { type } | Program { caller_id, type } | null

The execution context that produced this tool call.

One of the following:
Direct { type }
type: "direct"
Program { caller_id, type }
caller_id: string

The call ID of the program item that produced this tool call.

type: "program"
namespace?: string

The namespace of the function to run.

status?: "in_progress" | "completed" | "incomplete"

The status of the item. One of in_progress, completed, or incomplete. Populated when items are returned via API.

One of the following:
"in_progress"
"completed"
"incomplete"
FunctionCallOutput { output, type, id, 6 more }

The output of a function tool call.

output: string | BetaResponseFunctionCallOutputItemList { , , }

Text, image, or file output of the function tool call.

One of the following:
string
BetaResponseFunctionCallOutputItemList = Array<BetaResponseFunctionCallOutputItem>

An array of content outputs (text, image, file) for the function tool call.

One of the following:
BetaResponseInputTextContent { text, type, prompt_cache_breakpoint }

A text input to the model.

text: string

The text input to the model.

maxLength10485760
type: "input_text"

The type of the input item. Always input_text.

prompt_cache_breakpoint?: PromptCacheBreakpoint | null

Marks the exact end of a reusable prompt prefix. When ttl is omitted, the breakpoint inherits the request’s prompt_cache_options.ttl; the boundary is not rounded to a token block.

mode: "explicit"

The breakpoint mode. Always explicit.

ttl?: "30m" | "12h"

The minimum lifetime of this cache breakpoint. Defaults to the request TTL.

One of the following:
"30m"
"12h"
BetaResponseInputImageContent { type, detail, file_id, 2 more }

An image input to the model. Learn about image inputs

type: "input_image"

The type of the input item. Always input_image.

detail?: BetaImageDetail | null

The detail level of the image to be sent to the model. One of high, low, auto, or original. Defaults to auto.

file_id?: string | null

The ID of the file to be sent to the model.

image_url?: string | null

The URL of the image to be sent to the model. A fully qualified URL or base64 encoded image in a data URL.

maxLength20971520
formaturi
prompt_cache_breakpoint?: PromptCacheBreakpoint | null

Marks the exact end of a reusable prompt prefix. When ttl is omitted, the breakpoint inherits the request’s prompt_cache_options.ttl; the boundary is not rounded to a token block.

mode: "explicit"

The breakpoint mode. Always explicit.

ttl?: "30m" | "12h"

The minimum lifetime of this cache breakpoint. Defaults to the request TTL.

One of the following:
"30m"
"12h"
BetaResponseInputFileContent { type, detail, file_data, 4 more }

A file input to the model.

type: "input_file"

The type of the input item. Always input_file.

detail?: "auto" | "low" | "high"

The detail level of the file to be sent to the model. Use auto to let the system select the detail level; for GPT-5.6 and later models, auto uses high-quality rendering, which may increase input token usage. Use low for lower-cost rendering, or high to render the file at higher quality. Defaults to auto.

One of the following:
"auto"
"low"
"high"
file_data?: string | null

The base64-encoded data of the file to be sent to the model.

maxLength73400320
file_id?: string | null

The ID of the file to be sent to the model.

file_url?: string | null

The URL of the file to be sent to the model.

formaturi
filename?: string | null

The name of the file to be sent to the model.

prompt_cache_breakpoint?: PromptCacheBreakpoint | null

Marks the exact end of a reusable prompt prefix. When ttl is omitted, the breakpoint inherits the request’s prompt_cache_options.ttl; the boundary is not rounded to a token block.

mode: "explicit"

The breakpoint mode. Always explicit.

ttl?: "30m" | "12h"

The minimum lifetime of this cache breakpoint. Defaults to the request TTL.

One of the following:
"30m"
"12h"
type: "function_call_output"

The type of the function tool call output. Always function_call_output.

id?: string | null

The unique ID of the function tool call output. Populated when this item is returned via API.

agent?: Agent | null

The agent that produced this item.

agent_name: string

The canonical name of the agent that produced this item.

call_id?: string | null

The unique ID of the function tool call generated by the model.

minLength1
maxLength64
caller?: Direct { type } | Program { caller_id, type } | null

The execution context that produced this tool call.

One of the following:
Direct { type }
type: "direct"

The caller type. Always direct.

Program { caller_id, type }
caller_id: string

The call ID of the program item that produced this tool call.

minLength1
maxLength64
type: "program"

The caller type. Always program.

name?: string | null

The name of the tool that produced the output.

minLength1
maxLength128
namespace?: string | null

The namespace of the tool that produced the output.

minLength1
maxLength128
status?: "in_progress" | "completed" | "incomplete" | null

The status of the item. One of in_progress, completed, or incomplete. Populated when items are returned via API.

One of the following:
"in_progress"
"completed"
"incomplete"
AgentMessage { author, content, recipient, 3 more }

A message routed between agents.

author: string

The sending agent identity.

content: Array<BetaResponseInputTextContent { text, type, prompt_cache_breakpoint } | BetaResponseInputImageContent { type, detail, file_id, 2 more } | EncryptedContent { encrypted_content, type } >

Plaintext, image, or encrypted content sent between agents.

One of the following:
BetaResponseInputTextContent { text, type, prompt_cache_breakpoint }

A text input to the model.

text: string

The text input to the model.

maxLength10485760
type: "input_text"

The type of the input item. Always input_text.

prompt_cache_breakpoint?: PromptCacheBreakpoint | null

Marks the exact end of a reusable prompt prefix. When ttl is omitted, the breakpoint inherits the request’s prompt_cache_options.ttl; the boundary is not rounded to a token block.

mode: "explicit"

The breakpoint mode. Always explicit.

ttl?: "30m" | "12h"

The minimum lifetime of this cache breakpoint. Defaults to the request TTL.

One of the following:
"30m"
"12h"
BetaResponseInputImageContent { type, detail, file_id, 2 more }

An image input to the model. Learn about image inputs

type: "input_image"

The type of the input item. Always input_image.

detail?: BetaImageDetail | null

The detail level of the image to be sent to the model. One of high, low, auto, or original. Defaults to auto.

file_id?: string | null

The ID of the file to be sent to the model.

image_url?: string | null

The URL of the image to be sent to the model. A fully qualified URL or base64 encoded image in a data URL.

maxLength20971520
formaturi
prompt_cache_breakpoint?: PromptCacheBreakpoint | null

Marks the exact end of a reusable prompt prefix. When ttl is omitted, the breakpoint inherits the request’s prompt_cache_options.ttl; the boundary is not rounded to a token block.

mode: "explicit"

The breakpoint mode. Always explicit.

ttl?: "30m" | "12h"

The minimum lifetime of this cache breakpoint. Defaults to the request TTL.

One of the following:
"30m"
"12h"
EncryptedContent { encrypted_content, type }

Opaque encrypted content that Responses API decrypts inside trusted model execution.

encrypted_content: string

Opaque encrypted content.

maxLength10485760
type: "encrypted_content"

The type of the input item. Always encrypted_content.

recipient: string

The destination agent identity.

type: "agent_message"

The item type. Always agent_message.

id?: string | null

The unique ID of this agent message item.

agent?: Agent | null

The agent that produced this item.

agent_name: string

The canonical name of the agent that produced this item.

MultiAgentCall { action, arguments, call_id, 3 more }
action: "spawn_agent" | "interrupt_agent" | "list_agents" | 3 more

The multi-agent action that was executed.

One of the following:
"spawn_agent"
"interrupt_agent"
"list_agents"
"send_message"
"followup_task"
"wait_agent"
arguments: string

The action arguments as a JSON string.

call_id: string

The unique ID linking this call to its output.

minLength1
maxLength64
type: "multi_agent_call"

The item type. Always multi_agent_call.

id?: string | null

The unique ID of this multi-agent call.

agent?: Agent | null

The agent that produced this item.

agent_name: string

The canonical name of the agent that produced this item.

MultiAgentCallOutput { action, call_id, output, 3 more }
action: "spawn_agent" | "interrupt_agent" | "list_agents" | 3 more

The multi-agent action that produced this result.

One of the following:
"spawn_agent"
"interrupt_agent"
"list_agents"
"send_message"
"followup_task"
"wait_agent"
call_id: string

The unique ID of the multi-agent call.

minLength1
maxLength64
output: Array<Output>

Text output returned by the multi-agent action.

text: string

The text content.

maxLength10485760
type: "output_text"

The content type. Always output_text.

annotations?: Array<FileCitation { file_id, filename, index, type } | URLCitation { end_index, start_index, title, 2 more } | ContainerFileCitation { container_id, end_index, file_id, 3 more } >

Citations associated with the text content.

One of the following:
FileCitation { file_id, filename, index, type }
file_id: string

The ID of the file.

filename: string

The filename of the file cited.

index: number

The index of the file in the list of files.

minimum0
type: "file_citation"

The citation type. Always file_citation.

URLCitation { end_index, start_index, title, 2 more }
end_index: number

The index of the last character of the citation in the message.

minimum0
start_index: number

The index of the first character of the citation in the message.

minimum0
title: string

The title of the cited resource.

type: "url_citation"

The citation type. Always url_citation.

url: string

The URL of the cited resource.

formaturi
ContainerFileCitation { container_id, end_index, file_id, 3 more }
container_id: string

The ID of the container.

end_index: number

The index of the last character of the citation in the message.

minimum0
file_id: string

The ID of the container file.

filename: string

The filename of the container file cited.

start_index: number

The index of the first character of the citation in the message.

minimum0
type: "container_file_citation"

The citation type. Always container_file_citation.

type: "multi_agent_call_output"

The item type. Always multi_agent_call_output.

id?: string | null

The unique ID of this multi-agent call output.

agent?: Agent | null

The agent that produced this item.

agent_name: string

The canonical name of the agent that produced this item.

ToolSearchCall { arguments, type, id, 4 more }
arguments: unknown

The arguments supplied to the tool search call.

type: "tool_search_call"

The item type. Always tool_search_call.

id?: string | null

The unique ID of this tool search call.

agent?: Agent | null

The agent that produced this item.

agent_name: string

The canonical name of the agent that produced this item.

call_id?: string | null

The unique ID of the tool search call generated by the model.

minLength1
maxLength64
execution?: "server" | "client"

Whether tool search was executed by the server or by the client.

One of the following:
"server"
"client"
status?: "in_progress" | "completed" | "incomplete" | null

The status of the tool search call.

One of the following:
"in_progress"
"completed"
"incomplete"
BetaResponseToolSearchOutputItemParam { tools, type, id, 4 more }
tools: Array<BetaFunctionTool { name, parameters, strict, 6 more } | BetaFileSearchTool { type, vector_store_ids, filters, 2 more } | BetaComputerTool { type } | 13 more>

The loaded tool definitions returned by the tool search output.

One of the following:
BetaFunctionTool { name, parameters, strict, 6 more }

Defines a function in your own code the model can choose to call. Learn more about function calling.

name: string

The name of the function to call.

parameters: Record<string, unknown> | null

A JSON schema object describing the parameters of the function.

strict: boolean | null

Whether strict parameter validation is enforced for this function tool.

type: "function"

The type of the function tool. Always function.

allowed_callers?: Array<"direct" | "programmatic"> | null

The tool invocation context(s).

One of the following:
"direct"
"programmatic"
async?: boolean
defer_loading?: boolean

Whether this function is deferred and loaded via tool search.

description?: string | null

A description of the function. Used by the model to determine whether or not to call the function.

output_schema?: Record<string, unknown> | null

A JSON schema object describing the JSON value encoded in string outputs for this function.

BetaFileSearchTool { type, vector_store_ids, filters, 2 more }

A tool that searches for relevant content from uploaded files. Learn more about the file search tool.

type: "file_search"

The type of the file search tool. Always file_search.

vector_store_ids: Array<string>

The IDs of the vector stores to search.

filters?: ComparisonFilter { key, type, value } | CompoundFilter { filters, type } | null

A filter to apply.

One of the following:
ComparisonFilter { key, type, value }

A filter used to compare a specified attribute key to a given value using a defined comparison operation.

key: string

The key to compare against the value.

type: "eq" | "ne" | "gt" | 5 more

Specifies the comparison operator: eq, ne, gt, gte, lt, lte, in, nin.

  • eq: equals
  • ne: not equal
  • gt: greater than
  • gte: greater than or equal
  • lt: less than
  • lte: less than or equal
  • in: in
  • nin: not in
One of the following:
"eq"
"ne"
"gt"
"gte"
"lt"
"lte"
"in"
"nin"
value: string | number | boolean | Array<string | number>

The value to compare against the attribute key; supports string, number, or boolean types.

One of the following:
string
number
boolean
Array<string | number>
string
number
CompoundFilter { filters, type }

Combine multiple filters using and or or.

filters: Array<ComparisonFilter { key, type, value } | unknown>

Array of filters to combine. Items can be ComparisonFilter or CompoundFilter.

One of the following:
ComparisonFilter { key, type, value }

A filter used to compare a specified attribute key to a given value using a defined comparison operation.

key: string

The key to compare against the value.

type: "eq" | "ne" | "gt" | 5 more

Specifies the comparison operator: eq, ne, gt, gte, lt, lte, in, nin.

  • eq: equals
  • ne: not equal
  • gt: greater than
  • gte: greater than or equal
  • lt: less than
  • lte: less than or equal
  • in: in
  • nin: not in
One of the following:
"eq"
"ne"
"gt"
"gte"
"lt"
"lte"
"in"
"nin"
value: string | number | boolean | Array<string | number>

The value to compare against the attribute key; supports string, number, or boolean types.

unknown
type: "and" | "or"

Type of operation: and or or.

max_num_results?: number

The maximum number of results to return. This number should be between 1 and 50 inclusive.

ranking_options?: RankingOptions { hybrid_search, ranker, score_threshold }

Ranking options for search.

BetaComputerTool { type }

A tool that controls a virtual computer. Learn more about the computer tool.

BetaComputerUsePreviewTool { display_height, display_width, environment, type }

A tool that controls a virtual computer. Learn more about the computer tool.

BetaWebSearchTool { type, external_web_access, filters, 2 more }

Search the Internet for sources related to the prompt. Learn more about the web search tool.

Mcp { server_label, type, allowed_callers, 9 more }

Give the model access to additional tools via remote Model Context Protocol (MCP) servers. Learn more about MCP.

CodeInterpreter { container, type, allowed_callers }

A tool that runs Python code to help generate a response to a prompt.

ProgrammaticToolCalling { type }
ImageGeneration { type, action, background, 13 more }

A tool that generates images using the GPT image models.

LocalShell { type }

A tool that allows the model to execute shell commands in a local environment.

BetaFunctionShellTool { type, allowed_callers, environment }

A tool that allows the model to execute shell commands.

BetaCustomTool { name, type, allowed_callers, 4 more }

A custom tool that processes input using a specified format. Learn more about custom tools

BetaToolSearchOutputNamespaceTool { description, name, tools, type }

Groups function/custom tools under a shared namespace.

BetaToolSearchTool { type, description, execution, parameters }

Hosted or BYOT tool search configuration for deferred tools.

BetaWebSearchPreviewTool { type, search_content_types, search_context_size, user_location }

This tool searches the web for relevant results to use in a response. Learn more about the web search tool.

BetaApplyPatchTool { type, allowed_callers }

Allows the assistant to create, delete, or update files using unified diffs.

type: "tool_search_output"

The item type. Always tool_search_output.

id?: string | null

The unique ID of this tool search output.

agent?: Agent | null

The agent that produced this item.

call_id?: string | null

The unique ID of the tool search call generated by the model.

minLength1
maxLength64
execution?: "server" | "client"

Whether tool search was executed by the server or by the client.

status?: "in_progress" | "completed" | "incomplete" | null

The status of the tool search output.

AdditionalTools { role, tools, type, 2 more }
BetaResponseConfigurationUpdateItemParam { type, id, agent, 2 more }

An update to the conversation’s response configuration. Supply reasoning, text, or both. Each setting stays in effect until a later update replaces it.

BetaResponseReasoningItem { id, summary, type, 4 more }

A description of the chain of thought used by a reasoning model while generating a response. Be sure to include these items in your input to the Responses API for subsequent turns of a conversation if you are manually managing context.

BetaResponseCompactionItemParam { encrypted_content, type, id, agent }

A compaction item generated by the v1/responses/compact API.

ImageGenerationCall { id, result, status, 10 more }

An image generation request made by the model.

BetaResponseCodeInterpreterToolCall { id, code, container_id, 4 more }

A tool call to run code.

LocalShellCall { id, action, call_id, 3 more }

A tool call to run a command on the local shell.

LocalShellCallOutput { id, output, type, 2 more }

The output of a local shell tool call.

ShellCall { action, call_id, type, 5 more }

A tool representing a request to execute one or more shell commands.

ShellCallOutput { call_id, output, type, 5 more }

The streamed output items emitted by a shell tool call.

ApplyPatchCall { call_id, operation, status, 4 more }

A tool call representing a request to create, delete, or update files using diff patches.

ApplyPatchCallOutput { call_id, status, type, 4 more }

The streamed output emitted by an apply patch tool call.

McpListTools { id, server_label, tools, 3 more }

A list of tools available on an MCP server.

McpApprovalRequest { id, arguments, name, 3 more }

A request for human approval of a tool invocation.

McpApprovalResponse { approval_request_id, approve, type, 3 more }

A response to an MCP approval request.

McpCall { id, arguments, name, 7 more }

An invocation of a tool on an MCP server.

BetaResponseCustomToolCallOutput { call_id, output, type, 3 more }

The output of a custom tool call from your code, being sent back to the model.

BetaResponseCustomToolCall { call_id, input, name, 6 more }

A call to a custom tool created by the model.

CompactionTrigger { type, id, agent }

Compacts the current context. Must be the final input item.

ItemReference { id, agent, type }

An internal identifier for an item to reference.

Program { id, call_id, code, 3 more }
ProgramOutput { id, call_id, result, 3 more }
instructions?: string | null

A system (or developer) message inserted into the model’s context.

When using along with previous_response_id, the instructions from a previous response will not be carried over to the next response. This makes it simple to swap out system (or developer) messages in new responses.

max_output_tokens?: number | null

An upper bound for the number of tokens that can be generated for a response, including visible output tokens and reasoning tokens.

minimum16
max_tool_calls?: number | null

The maximum number of total calls to built-in tools that can be processed in a response. This maximum number applies across all built-in tool calls, not per individual tool. Any further attempts to call a tool by the model will be ignored.

metadata?: Record<string, string> | null

Set of 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, and querying for objects via API or the dashboard.

Keys are strings with a maximum length of 64 characters. Values are strings with a maximum length of 512 characters.

model?: "gpt-6-astra" | "gpt-6.1-sol" | "gpt-6.1-terra" | 109 more | (string & {})

Model ID used to generate the response, like gpt-6-astra. OpenAI offers a wide range of models with different capabilities, performance characteristics, and price points. Refer to the model guide to browse and compare available models.

Deprecated values:

  • “gpt-5.4-nano”, “gpt-5.4-nano-2026-03-17”, “gpt-5.1”, “gpt-5.1-2025-11-13”: Announced shutdown date: 2027-04-01. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “gpt-5.3-chat-latest”, “gpt-5.2-chat-latest”: Announced shutdown date: 2026-08-10. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “gpt-5.1-codex”, “gpt-5.1-chat-latest”, “gpt-5-chat-latest”, “gpt-4o-search-preview”, “gpt-4o-mini-search-preview”, “gpt-4o-search-preview-2025-03-11”, “gpt-4o-mini-search-preview-2025-03-11”, “o3-deep-research”, “o3-deep-research-2025-06-26”, “o4-mini-deep-research”, “o4-mini-deep-research-2025-06-26”, “computer-use-preview”, “computer-use-preview-2025-03-11”, “gpt-5-codex”, “gpt-5.1-codex-max”: Announced shutdown date: 2026-07-23. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “gpt-5.1-mini”: Not a supported model ID. Retained for SDK compatibility.
  • “gpt-5”, “gpt-5-mini”, “gpt-5-nano”, “gpt-5-2025-08-07”, “gpt-5-mini-2025-08-07”, “gpt-5-nano-2025-08-07”, “o3”, “o3-2025-04-16”, “o3-pro”, “o3-pro-2025-06-10”, “gpt-5-pro”, “gpt-5-pro-2025-10-06”: Announced shutdown date: 2026-12-11. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “gpt-4.1”, “gpt-4.1-mini”, “gpt-4.1-2025-04-14”, “gpt-4.1-mini-2025-04-14”, “gpt-4o”, “gpt-4o-2024-11-20”, “gpt-4o-2024-08-06”, “gpt-4o-mini”, “gpt-4o-mini-2024-07-18”: Announced shutdown date: 2027-01-14. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “gpt-4.1-nano”, “gpt-4.1-nano-2025-04-14”, “o4-mini”, “o4-mini-2025-04-16”, “o3-mini”, “o3-mini-2025-01-31”, “o1”, “o1-2024-12-17”, “gpt-4o-2024-05-13”, “gpt-4-turbo”, “gpt-4-turbo-2024-04-09”, “gpt-4-1106-preview”, “gpt-4”, “gpt-4-0613”, “gpt-3.5-turbo”, “gpt-3.5-turbo-0125”, “o1-pro”, “o1-pro-2025-03-19”: Announced shutdown date: 2026-10-23. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “o1-preview”, “o1-preview-2024-09-12”: Announced shutdown date: 2025-07-28. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “o1-mini”, “o1-mini-2024-09-12”: Announced shutdown date: 2025-10-27. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “gpt-audio-mini”, “gpt-audio-mini-2025-12-15”: Announced shutdown date: 2027-01-20. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “gpt-4o-audio-preview”, “gpt-4o-audio-preview-2024-12-17”, “gpt-4o-audio-preview-2025-06-03”, “gpt-4o-mini-audio-preview”, “gpt-4o-mini-audio-preview-2024-12-17”: Announced shutdown date: 2026-05-07. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “gpt-4o-audio-preview-2024-10-01”: Announced shutdown date: 2025-10-10. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “chatgpt-4o-latest”: Announced shutdown date: 2026-02-17. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “codex-mini-latest”: Announced shutdown date: 2026-02-12. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “gpt-4-0125-preview”, “gpt-4-turbo-preview”, “gpt-4-0314”: Announced shutdown date: 2026-03-26. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “gpt-4-vision-preview”: Announced shutdown date: 2024-12-06. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “gpt-4-32k”, “gpt-4-32k-0314”, “gpt-4-32k-0613”: Announced shutdown date: 2025-06-06. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “gpt-3.5-turbo-16k”, “gpt-3.5-turbo-0301”, “gpt-3.5-turbo-0613”, “gpt-3.5-turbo-16k-0613”: Announced shutdown date: 2024-09-13. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
  • “gpt-3.5-turbo-1106”: Announced shutdown date: 2026-09-28. See https://developers.openai.com/api/docs/deprecations for details and recommended replacements.
moderation?: Moderation | null

Configuration for running moderation on the input and output of this response.

multi_agent?: MultiAgent | null

Configuration for server-hosted multi-agent execution.

parallel_tool_calls?: boolean | null

Whether to allow the model to run tool calls in parallel.

previous_response_id?: string | null

The unique ID of the previous response to the model. Use this to create multi-turn conversations. Learn more about conversation state. Cannot be used in conjunction with conversation.

prompt?: BetaResponsePrompt { id, variables, version } | null

Reference to a prompt template and its variables. Learn more.

prompt_cache_key?: string | null

Used by OpenAI to cache responses for similar requests to optimize your cache hit rates. Replaces the user field. Learn more.

prompt_cache_options?: PromptCacheOptions { comparison_response_id, mode, prewarm, ttl }

Options for prompt caching. Supported for gpt-5.6 and later models. By default, OpenAI automatically chooses one implicit cache breakpoint. You can add explicit breakpoints to content blocks with prompt_cache_breakpoint. Each request can write up to four breakpoints. For cache matching, OpenAI considers up to the latest 80 breakpoints in the conversation, without a content-block lookback limit. Set mode to explicit to disable the implicit breakpoint. The ttl defaults to 30m. Supported lifetimes depend on the API. See the prompt caching guide for current details.

prompt_cache_retention?: "in_memory" | "24h" | null

Deprecated.

Deprecated. Use prompt_cache_options.ttl instead.

The retention policy for the prompt cache. Set to 24h to enable extended prompt caching, which keeps cached prefixes active for longer, up to a maximum of 24 hours. Learn more. This field expresses a maximum retention policy, while prompt_cache_options.ttl expresses a minimum cache lifetime. The two fields are independent and do not interact. For gpt-5.5, gpt-5.5-pro, and future models, only 24h is supported.

For older models that support both in_memory and 24h, the default depends on your organization’s data retention policy:

  • Organizations without ZDR enabled default to 24h.
  • Organizations with ZDR enabled default to in_memory when prompt_cache_retention is not specified.
reasoning?: Reasoning | null

Configuration options for reasoning models.

safety_identifier?: string | null

A stable identifier used to help detect users of your application that may be violating OpenAI’s usage policies. The IDs should be a string that uniquely identifies each user, with a maximum length of 128 characters. We recommend hashing their username or email address, in order to avoid sending us any identifying information. Learn more.

maxLength128
service_tier?: BetaServiceTier | null

Specifies the processing type used for serving the request.

  • If set to ‘auto’, then the request will be processed with the service tier configured in the Project settings. Unless otherwise configured, the Project will use ‘default’.
  • If set to ‘default’, then the request will be processed with the standard pricing and performance for the selected model.
  • If set to ‘flex’, then the request will be processed with the Flex Processing service tier.
  • To opt-in to Fast mode at the request level, include the service_tier=fast or service_tier=priority parameter for Responses or Chat Completions. The response will show service_tier=priority regardless of if you specify service_tier=fast or priority in your request.
  • If set to ‘ultrafast’, then the request will be processed with the access-controlled Ultrafast Processing service tier. This tier is currently available for gpt-5.6-sol; a response served through it will show service_tier=ultrafast.
  • When not set, the default behavior is ‘auto’.

When the service_tier parameter is set, the response body will include the service_tier value based on the processing mode actually used to serve the request. This response value may be different from the value set in the parameter.

store?: boolean | null

Whether to store the generated model response for later retrieval via API. Defaults to true when omitted. If set to true, response data will be stored for at least 30 days, subject to the data retention exceptions.

stream?: boolean | null

If set to true, the model response data will be streamed to the client as it is generated using server-sent events. See the Streaming section below for more information.

stream_id?: string

The WebSocket lane for this response. Requests with the same stream_id are processed FIFO, and events for the response echo the same stream_id.

stream_id controls routing; previous_response_id controls conversation lineage, so a new lane can fork from a response created on another lane.

minLength1
maxLength256
stream_options?: StreamOptions | null

Options for streaming responses. Only set this when you set stream: true.

temperature?: number | null

What sampling temperature to use, between 0 and 2. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic. We generally recommend altering this or top_p but not both.

minimum0
maximum2
text?: BetaResponseTextConfig { format, verbosity }

Configuration options for a text response from the model. Can be plain text or structured JSON data. Learn more:

tool_choice?: BetaToolChoiceOptions | BetaToolChoiceAllowed { mode, tools, type } | BetaToolChoiceTypes { type } | 6 more

How the model should select which tool (or tools) to use when generating a response. See the tools parameter to see how to specify which tools the model can call.

tools?: Array<BetaTool>

An array of tools the model may call while generating a response. You can specify which tool to use by setting the tool_choice parameter.

We support the following categories of tools:

  • Built-in tools: Tools that are provided by OpenAI that extend the model’s capabilities, like web search or file search. Learn more about built-in tools.
  • MCP Tools: Integrations with third-party systems via custom MCP servers or predefined connectors such as Google Drive and SharePoint. Learn more about MCP Tools.
  • Function calls (custom tools): Functions that are defined by you, enabling the model to call your own code with strongly typed arguments and outputs. Learn more about function calling. You can also use custom tools to call your own code.
top_logprobs?: number | null

An integer between 0 and 20 specifying the maximum number of most likely tokens to return at each token position, each with an associated log probability. In some cases, the number of returned tokens may be fewer than requested.

minimum0
maximum20
top_p?: number | null

An alternative to sampling with temperature, called nucleus sampling, where the model considers the results of the tokens with top_p probability mass. So 0.1 means only the tokens comprising the top 10% probability mass are considered.

We generally recommend altering this or temperature but not both.

minimum0
maximum1
truncation?: "auto" | "disabled" | null

Deprecated.

The truncation strategy to use for the model response.

  • auto: If the input to this Response exceeds the model’s context window size, the model will truncate the response to fit the context window by dropping items from the beginning of the conversation. On some models, trimming can begin earlier because the API subtracts the model’s maximum output allowance from its input limit, even if you set a lower max_output_tokens.
  • disabled (default): If the input size will exceed the context window size for a model, the request will fail with a 400 error.
user?: string

Deprecated.

This field is being replaced by safety_identifier and prompt_cache_key. Use prompt_cache_key instead to maintain caching optimizations. A stable identifier for your end-users. Used to boost cache hit rates by better bucketing similar requests and to help OpenAI detect and prevent abuse. Learn more.

BetaResponseSteerEvent { input, previous_response_id, type }

Queues user input to steer a response on this WebSocket connection. Input can contain text, images, and files. Steering is supported only for single-agent responses on models and execution modes that support steering. Responses bound to a conversation or using automatic compaction do not support steering.

A response.steer.accepted event acknowledges that the server owns the queued input, not that it has been applied. The successor’s response.created event is the commit point. Input that cannot be committed is returned in response.steer.failed.

Steering may cause the active response to finish at a safe output boundary with response.incomplete and incomplete_details.reason set to steered, followed automatically by a successor response.created. Normal completion can also be followed by an automatic successor. Automatic successors inherit the previous response’s settings and continue from it with the queued input.

If the response stops for client-owned tool output or approval, accepted steering input remains queued and response.steer.pending is emitted after response.completed. Fill the required_input stubs from that event with saved tool results or approval decisions, and send one explicit response.create per parent with the same previous_response_id and WebSocket lane. Do not rerun tools or resend accepted steering input. The queued input is prepended in submission order to that request’s input, and the explicit request retains its own settings.

This event accepts only type, previous_response_id, and input. Do not send stream_id; the target response determines the WebSocket lane.

BetaResponseInjectEvent { input, response_id, type }

Injects input items into an active response over a WebSocket connection. The items are validated and committed atomically. Currently, the server accepts client-owned tool outputs that resume a waiting agent.