@assistant-ui/ai-sdk

Vercel AI SDK runtime hooks, chat transports, and message conversion utilities for assistant-ui applications.

API Reference

AISDKChat

AuiConfig entry that runs the AI SDK chat as the threads scope. Hosts the same orchestration as useChatRuntime inside the client's own resource tree, so it works with any AssistantClient host, React or not. Single thread; the multi-thread and assistant-cloud surface is AISDKThreads. The chat id is captured when the entry mounts, so a later id change in the options has no effect.

AISDKChat props
0?AISDKChatOptions<UI_MESSAGE>

AISDKChat props["0"]
id?string

A unique identifier for the chat. If not provided, a random one will be generated.

messageMetadataSchema?FlexibleSchema$1<UI_MESSAGE['metadata']>

dataPartSchemas?UIDataTypesToSchemas<InferUIMessageData<UI_MESSAGE>>

messages?UI_MESSAGE[]

generateId?IdGenerator$1

A way to provide a function that is going to be used for ids for messages and the chat. If not provided the default AI SDK `generateId` is used.

transport?ChatTransport<UI_MESSAGE>

ChatTransport
sendMessages(options: { /** The type of message submission - either new message or regeneration */ trigger: 'submit-message' | 'regenerate-message'; /** Unique identifier for the chat session */ chatId: string; /** ID of the message to regenerate, or undefined for new messages */ messageId: string | undefined; /** Array of UI messages representing the conversation history */ messages: UI_MESSAGE[]; /** Signal to abort the request if needed */ abortSignal: AbortSignal | undefined; } & ChatRequestOptions) => Promise<ReadableStream<UIMessageChunk>>

Sends messages to the chat API endpoint and returns a streaming response. This method handles both new message submission and message regeneration. It supports real-time streaming of responses through UIMessageChunk events.

reconnectToStream(options: { /** Unique identifier for the chat session to reconnect to */ chatId: string; /** Signal to abort the reconnection request if needed */ abortSignal?: AbortSignal; } & ChatRequestOptions) => Promise<ReadableStream<UIMessageChunk> | null>

Reconnects to an existing streaming response for the specified chat session. This method is used to resume streaming when a connection is interrupted or when resuming a chat session. It's particularly useful for maintaining continuity in long-running conversations or recovering from network issues.

onError?ChatOnErrorCallback

Callback function to be called when an error is encountered.

onToolCall?ChatOnToolCallCallback<UI_MESSAGE>

Optional callback function that is invoked when a tool call is received. Intended for automatic client-side tool execution. To add the tool output, call `addToolOutput` without awaiting it inside this callback. The callback's return value is not used.

onFinish?ChatOnFinishCallback<UI_MESSAGE>

Function that is called when the assistant response has finished streaming.

onData?ChatOnDataCallback<UI_MESSAGE>

Optional callback function that is called when a data part is received.

sendAutomaticallyWhen?(options: { messages: UI_MESSAGE[]; }) => boolean | PromiseLike<boolean>

When provided, this function will be called when the stream is finished or a tool call is added to determine if the current messages should be resubmitted.

suggestions?readonly ThreadSuggestion[] | undefined

isDisabled?boolean | undefined

Whether the entire thread is disabled. When `true`, the composer's input is also disabled (the user cannot type, attach files, or submit). For a narrower gate that keeps the input usable but blocks only sending, use `isSendDisabled`.

isSendDisabled?boolean | undefined

Whether sending new messages is currently disabled. When `true`, the thread composer's input remains usable but `send()` becomes a no-op and the thread composer's `canSend` is `false`. Use this to gate sending on external React state (e.g. while tool config is loading) without disabling the input itself the way `isDisabled` does. Edit composers (saving message edits) intentionally ignore this flag.

unstable_capabilities?AISDKChat props["0"]["unstable_capabilities"]unstable

AISDKChat props["0"]["unstable_capabilities"]
copy?boolean | undefined

throttle?number | undefined

adapters?AISDKRuntimeAdapter["adapters"] | undefined

AISDKChat props["0"]["adapters"]
attachments?AttachmentAdapter | undefined

AISDKChat props["0"]["adapters"]["attachments"]
acceptstring

add(state: { file: File; }) => Promise<PendingAttachment> | AsyncGenerator<PendingAttachment, void>

remove(attachment: Attachment) => Promise<void>

send(attachment: PendingAttachment, options?: { signal?: AbortSignal; }) => Promise<CompleteAttachment>

speech?SpeechSynthesisAdapter | undefined

AISDKChat props["0"]["adapters"]["speech"]
speak(text: string) => SpeechSynthesisAdapter.Utterance

dictation?DictationAdapter | undefined

AISDKChat props["0"]["adapters"]["dictation"]
listen() => DictationAdapter.Session

disableInputDuringDictation?boolean

voice?RealtimeVoiceAdapter | undefined

AISDKChat props["0"]["adapters"]["voice"]
connect(options: { abortSignal?: AbortSignal; }) => RealtimeVoiceAdapter.Session

feedback?FeedbackAdapter | undefined

AISDKChat props["0"]["adapters"]["feedback"]
submit(feedback: FeedbackAdapterFeedback) => void

threadList?ExternalStoreThreadListAdapter | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

AISDKChat props["0"]["adapters"]["threadList"]
threadId?string | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

isLoading?boolean | undefined

threads?readonly ExternalStoreThreadData<"regular">[] | undefined

archivedThreads?readonly ExternalStoreThreadData<"archived">[] | undefined

onSwitchToNewThread?(() => Promise<void> | void) | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

onSwitchToThread?((threadId: string) => Promise<void> | void) | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

onRename?( threadId: string, newTitle: string, ) => (Promise<void> | void) | undefined

onUpdateCustom?(( threadId: string, custom: Record<string, unknown> | undefined, ) => Promise<void> | void) | undefined

onArchive?((threadId: string) => Promise<void> | void) | undefined

onUnarchive?((threadId: string) => Promise<void> | void) | undefined

onDelete?((threadId: string) => Promise<void> | void) | undefined

history?ThreadHistoryAdapter | undefined

AISDKChat props["0"]["adapters"]["history"]
scopeId?string | undefined

Stable identity for the storage scope read by LocalRuntime. Keep it the same when recreating the adapter for one thread, account, or workspace, and change it before loading a different scope. Adapters that omit it are treated as sharing one scope. External-history runtimes do not read it.

unstable_copy?(( branch: readonly ThreadMessage[], messageIds: readonly string[], ) => Promise<void>) | undefinedunstable

Keeps a copy of messages whose source of truth is the runtime's backend. `branch` is the conversation from its first message to its last, and `messageIds` names the ones in it that are new or changed; the adapter stores those, and first any earlier message of the branch it does not hold yet, keyed by each message's own id. It is undefined while the adapter keeps no copies, so the runtime then neither copies nor records tool interactions.

load() => Promise<ExportedMessageRepository & { state?: ReadonlyJSONValue; unstable_resume?: boolean; }>

resume?(options: ChatModelRunOptions) => AsyncGenerator<ChatModelRunResult, void, unknown>

append(item: ExportedMessageRepositoryItem) => Promise<void>

update?(item: ExportedMessageRepositoryItem) => Promise<void>

Rewrites a previously appended message in place, keyed by its message id. Adapters that implement this let a runtime persist a run paused for tool approval, finalize the same message once the run resumes, and record a tool result that arrives after the message settled, which can be a message later turns follow. Without it, a paused run is appended when it ends or when a later turn follows its message while the run is still open; anything that run adds after the append is not stored. An update may arrive for an id whose earlier write failed; treat it as an upsert keyed on the message id rather than assuming the entry exists.

delete?(items: ExportedMessageRepositoryItem[]) => Promise<void>

Deletes messages from history. The runtime may send a second delete for the same message after a write it issued before the delete lands; treat deleting a missing entry as success.

withFormat?<TMessage, TStorageFormat extends Record<string, unknown>>(formatAdapter: MessageFormatAdapter<TMessage, TStorageFormat>) => GenericThreadHistoryAdapter<TMessage>

Required when used with `useAISDKRuntime` / `useChatRuntime`.

suggestion?SuggestionAdapter | undefined

AISDKChat props["0"]["adapters"]["suggestion"]
key?string | number | symbol | undefined

Changes when the adapter should regenerate suggestions for a settled run.

generate( options: SuggestionAdapterGenerateOptions, ) => | Promise<readonly ThreadSuggestion[]> | AsyncGenerator<readonly ThreadSuggestion[], void>

toCreateMessage?CustomToCreateMessageFunction

onResume?AISDKRuntimeAdapter["onResume"]

onResumeToolCall?AISDKRuntimeAdapter["onResumeToolCall"]

onRespondToToolApproval?AISDKRuntimeAdapter["onRespondToToolApproval"]

onResumeError?((error: unknown) => void) | undefined

Called when an automatic resumable stream reconnect fails. Use this to surface a toast, report telemetry, or mark the thread as needing a retry. The failed stream id is cleared after the callback unless a newer id has replaced it.

joinStrategy?AISDKRuntimeAdapter["joinStrategy"]

messageRepository?AISDKRuntimeAdapter<UI_MESSAGE>["messageRepository"]

AISDKChat props["0"]["messageRepository"]
headId?string | null

messagesMessageFormatItem<TMessage>[]

unstable_onBranchChange?AISDKRuntimeAdapter["unstable_onBranchChange"]unstable

length0 | 1

toString() => string

toLocaleString{ (): string; (locales: string | string[], options?: Intl.NumberFormatOptions & Intl.DateTimeFormatOptions): string; }

pop() => AISDKChatOptions<UI_MESSAGE>

push(...items: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => number

concat{ (...items: ConcatArray<AISDKChatOptions<UI_MESSAGE> | undefined>[]): (AISDKChatOptions<UI_MESSAGE> | undefined)[]; (...items: (AISDKChatOptions<UI_MESSAGE> | ConcatArray<AISDKChatOptions<UI_MESSAGE> | undefined> | undefined)[]): (AISDKChatOptions<UI_MESSAGE> | undefined)[]; }

join(separator?: string) => string

reverse() => (AISDKChatOptions<UI_MESSAGE> | undefined)[]

shift() => AISDKChatOptions<UI_MESSAGE>

slice(start?: number, end?: number) => (AISDKChatOptions<UI_MESSAGE> | undefined)[]

sort(compareFn?: ((a: AISDKChatOptions<UI_MESSAGE> | undefined, b: AISDKChatOptions<UI_MESSAGE> | undefined) => number) | undefined) => [options?: AISDKChatOptions<UI_MESSAGE> | undefined]

splice{ (start: number, deleteCount?: number): (AISDKChatOptions<UI_MESSAGE> | undefined)[]; (start: number, deleteCount: number, ...items: (AISDKChatOptions<UI_MESSAGE> | undefined)[]): (AISDKChatOptions<UI_MESSAGE> | undefined)[]; }

unshift(...items: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => number

indexOf(searchElement: AISDKChatOptions<UI_MESSAGE> | undefined, fromIndex?: number) => number

lastIndexOf(searchElement: AISDKChatOptions<UI_MESSAGE> | undefined, fromIndex?: number) => number

every{ <S>(predicate: (value: AISDKChatOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => value is S, thisArg?: any): this is S[]; (predicate: (value: AISDKChatOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => unknown, thisArg?: any): boolean; }

some(predicate: (value: AISDKChatOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => unknown, thisArg?: any) => boolean

forEach(callbackfn: (value: AISDKChatOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => void, thisArg?: any) => void

map<U>(callbackfn: (value: AISDKChatOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => U, thisArg?: any) => U[]

filter{ <S>(predicate: (value: AISDKChatOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => value is S, thisArg?: any): S[]; (predicate: (value: AISDKChatOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => unknown, thisArg?: any): (AISDKChatOptions<UI_MESSAGE> | undefined)[]; }

reduce{ (callbackfn: (previousValue: AISDKChatOptions<UI_MESSAGE> | undefined, currentValue: AISDKChatOptions<UI_MESSAGE> | undefined, currentIndex: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => AISDKChatOptions<UI_MESSAGE> | undefined): AISDKChatOptions<UI_MESSAGE> | undefined; (callbackfn: (previousValue: AISDKChatOptions<UI_MESSAGE> | undefined, currentValue: AISDKChatOptions<UI_MESSAGE> | undefined, currentIndex: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => AISDKChatOptions<UI_MESSAGE> | undefined, initialValue: AISDKChatOptions<UI_MESSAGE> | undefined): AISDKChatOptions<UI_MESSAGE> | undefined; <U>(callbackfn: (previousValue: U, currentValue: AISDKChatOptions<UI_MESSAGE> | undefined, currentIndex: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => U, initialValue: U): U; }

reduceRight{ (callbackfn: (previousValue: AISDKChatOptions<UI_MESSAGE> | undefined, currentValue: AISDKChatOptions<UI_MESSAGE> | undefined, currentIndex: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => AISDKChatOptions<UI_MESSAGE> | undefined): AISDKChatOptions<UI_MESSAGE> | undefined; (callbackfn: (previousValue: AISDKChatOptions<UI_MESSAGE> | undefined, currentValue: AISDKChatOptions<UI_MESSAGE> | undefined, currentIndex: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => AISDKChatOptions<UI_MESSAGE> | undefined, initialValue: AISDKChatOptions<UI_MESSAGE> | undefined): AISDKChatOptions<UI_MESSAGE> | undefined; <U>(callbackfn: (previousValue: U, currentValue: AISDKChatOptions<UI_MESSAGE> | undefined, currentIndex: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => U, initialValue: U): U; }

find{ <S>(predicate: (value: AISDKChatOptions<UI_MESSAGE> | undefined, index: number, obj: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => value is S, thisArg?: any): S | undefined; (predicate: (value: AISDKChatOptions<UI_MESSAGE> | undefined, index: number, obj: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => unknown, thisArg?: any): AISDKChatOptions<UI_MESSAGE> | undefined; }

findIndex(predicate: (value: AISDKChatOptions<UI_MESSAGE> | undefined, index: number, obj: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => unknown, thisArg?: any) => number

fill(value: AISDKChatOptions<UI_MESSAGE> | undefined, start?: number, end?: number) => [options?: AISDKChatOptions<UI_MESSAGE> | undefined]

copyWithin(target: number, start: number, end?: number) => [options?: AISDKChatOptions<UI_MESSAGE> | undefined]

entries() => ArrayIterator<[number, AISDKChatOptions<UI_MESSAGE> | undefined]>

keys() => ArrayIterator<number>

values() => ArrayIterator<AISDKChatOptions<UI_MESSAGE> | undefined>

includes(searchElement: AISDKChatOptions<UI_MESSAGE> | undefined, fromIndex?: number) => boolean

flatMap<U, This>(callback: (this: This, value: AISDKChatOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => U | readonly U[], thisArg?: This | undefined) => U[]

flat<A, D>(this: A, depth?: D | undefined) => FlatArray<A, D>[]

at(index: number) => AISDKChatOptions<UI_MESSAGE>

findLast{ <S>(predicate: (value: AISDKChatOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => value is S, thisArg?: any): S | undefined; (predicate: (value: AISDKChatOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => unknown, thisArg?: any): AISDKChatOptions<UI_MESSAGE> | undefined; }

findLastIndex(predicate: (value: AISDKChatOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKChatOptions<UI_MESSAGE> | undefined)[]) => unknown, thisArg?: any) => number

toReversed() => (AISDKChatOptions<UI_MESSAGE> | undefined)[]

toSorted(compareFn?: ((a: AISDKChatOptions<UI_MESSAGE> | undefined, b: AISDKChatOptions<UI_MESSAGE> | undefined) => number) | undefined) => (AISDKChatOptions<UI_MESSAGE> | undefined)[]

toSpliced{ (start: number, deleteCount: number, ...items: (AISDKChatOptions<UI_MESSAGE> | undefined)[]): (AISDKChatOptions<UI_MESSAGE> | undefined)[]; (start: number, deleteCount?: number): (AISDKChatOptions<UI_MESSAGE> | undefined)[]; }

with(index: number, value: AISDKChatOptions<UI_MESSAGE> | undefined) => (AISDKChatOptions<UI_MESSAGE> | undefined)[]

AISDKThreads

AuiConfig entry that runs one AI SDK chat per thread. Hosts the same per-thread orchestration as AISDKChat inside the client's own resource tree, so it works with any AssistantClient host, React or not. Without cloud, threads live in memory for the client's lifetime and keep their history across switches; only the visible thread is mounted, and a switched-away chat keeps streaming into its stored state until it settles or the thread is deleted. With cloud, the list is a RemoteThreadList with backgroundThreads: every visited thread stays mounted with its own history, a run continues after a switch and stops on delete, and a freshly created thread titles itself. Model context is registered on every mounted thread.

AISDKThreads props
0?AISDKThreadsOptions<UI_MESSAGE>

AISDKThreads props["0"]
suggestions?readonly ThreadSuggestion[] | undefined

onError?ChatOnErrorCallback

Callback function to be called when an error is encountered.

isDisabled?boolean | undefined

Whether the entire thread is disabled. When `true`, the composer's input is also disabled (the user cannot type, attach files, or submit). For a narrower gate that keeps the input usable but blocks only sending, use `isSendDisabled`.

isSendDisabled?boolean | undefined

Whether sending new messages is currently disabled. When `true`, the thread composer's input remains usable but `send()` becomes a no-op and the thread composer's `canSend` is `false`. Use this to gate sending on external React state (e.g. while tool config is loading) without disabling the input itself the way `isDisabled` does. Edit composers (saving message edits) intentionally ignore this flag.

messageRepository?AISDKRuntimeAdapter<UI_MESSAGE>["messageRepository"]

AISDKThreads props["0"]["messageRepository"]
headId?string | null

messagesMessageFormatItem<TMessage>[]

unstable_onBranchChange?AISDKRuntimeAdapter["unstable_onBranchChange"]unstable

onResume?AISDKRuntimeAdapter["onResume"]

onResumeToolCall?AISDKRuntimeAdapter["onResumeToolCall"]

onRespondToToolApproval?AISDKRuntimeAdapter["onRespondToToolApproval"]

adapters?AISDKRuntimeAdapter["adapters"] | undefined

AISDKThreads props["0"]["adapters"]
attachments?AttachmentAdapter | undefined

AISDKThreads props["0"]["adapters"]["attachments"]
acceptstring

add(state: { file: File; }) => Promise<PendingAttachment> | AsyncGenerator<PendingAttachment, void>

remove(attachment: Attachment) => Promise<void>

send(attachment: PendingAttachment, options?: { signal?: AbortSignal; }) => Promise<CompleteAttachment>

speech?SpeechSynthesisAdapter | undefined

AISDKThreads props["0"]["adapters"]["speech"]
speak(text: string) => SpeechSynthesisAdapter.Utterance

dictation?DictationAdapter | undefined

AISDKThreads props["0"]["adapters"]["dictation"]
listen() => DictationAdapter.Session

disableInputDuringDictation?boolean

voice?RealtimeVoiceAdapter | undefined

AISDKThreads props["0"]["adapters"]["voice"]
connect(options: { abortSignal?: AbortSignal; }) => RealtimeVoiceAdapter.Session

feedback?FeedbackAdapter | undefined

AISDKThreads props["0"]["adapters"]["feedback"]
submit(feedback: FeedbackAdapterFeedback) => void

threadList?ExternalStoreThreadListAdapter | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

AISDKThreads props["0"]["adapters"]["threadList"]
threadId?string | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

isLoading?boolean | undefined

threads?readonly ExternalStoreThreadData<"regular">[] | undefined

archivedThreads?readonly ExternalStoreThreadData<"archived">[] | undefined

onSwitchToNewThread?(() => Promise<void> | void) | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

onSwitchToThread?((threadId: string) => Promise<void> | void) | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

onRename?( threadId: string, newTitle: string, ) => (Promise<void> | void) | undefined

onUpdateCustom?(( threadId: string, custom: Record<string, unknown> | undefined, ) => Promise<void> | void) | undefined

onArchive?((threadId: string) => Promise<void> | void) | undefined

onUnarchive?((threadId: string) => Promise<void> | void) | undefined

onDelete?((threadId: string) => Promise<void> | void) | undefined

history?ThreadHistoryAdapter | undefined

AISDKThreads props["0"]["adapters"]["history"]
scopeId?string | undefined

Stable identity for the storage scope read by LocalRuntime. Keep it the same when recreating the adapter for one thread, account, or workspace, and change it before loading a different scope. Adapters that omit it are treated as sharing one scope. External-history runtimes do not read it.

unstable_copy?(( branch: readonly ThreadMessage[], messageIds: readonly string[], ) => Promise<void>) | undefinedunstable

Keeps a copy of messages whose source of truth is the runtime's backend. `branch` is the conversation from its first message to its last, and `messageIds` names the ones in it that are new or changed; the adapter stores those, and first any earlier message of the branch it does not hold yet, keyed by each message's own id. It is undefined while the adapter keeps no copies, so the runtime then neither copies nor records tool interactions.

load() => Promise<ExportedMessageRepository & { state?: ReadonlyJSONValue; unstable_resume?: boolean; }>

resume?(options: ChatModelRunOptions) => AsyncGenerator<ChatModelRunResult, void, unknown>

append(item: ExportedMessageRepositoryItem) => Promise<void>

update?(item: ExportedMessageRepositoryItem) => Promise<void>

Rewrites a previously appended message in place, keyed by its message id. Adapters that implement this let a runtime persist a run paused for tool approval, finalize the same message once the run resumes, and record a tool result that arrives after the message settled, which can be a message later turns follow. Without it, a paused run is appended when it ends or when a later turn follows its message while the run is still open; anything that run adds after the append is not stored. An update may arrive for an id whose earlier write failed; treat it as an upsert keyed on the message id rather than assuming the entry exists.

delete?(items: ExportedMessageRepositoryItem[]) => Promise<void>

Deletes messages from history. The runtime may send a second delete for the same message after a write it issued before the delete lands; treat deleting a missing entry as success.

withFormat?<TMessage, TStorageFormat extends Record<string, unknown>>(formatAdapter: MessageFormatAdapter<TMessage, TStorageFormat>) => GenericThreadHistoryAdapter<TMessage>

Required when used with `useAISDKRuntime` / `useChatRuntime`.

suggestion?SuggestionAdapter | undefined

AISDKThreads props["0"]["adapters"]["suggestion"]
key?string | number | symbol | undefined

Changes when the adapter should regenerate suggestions for a settled run.

generate( options: SuggestionAdapterGenerateOptions, ) => | Promise<readonly ThreadSuggestion[]> | AsyncGenerator<readonly ThreadSuggestion[], void>

unstable_capabilities?AISDKThreads props["0"]["unstable_capabilities"]unstable

AISDKThreads props["0"]["unstable_capabilities"]
copy?boolean | undefined

onFinish?ChatOnFinishCallback<UI_MESSAGE>

Function that is called when the assistant response has finished streaming.

joinStrategy?AISDKRuntimeAdapter["joinStrategy"]

messageMetadataSchema?FlexibleSchema$1<UI_MESSAGE['metadata']>

dataPartSchemas?UIDataTypesToSchemas<InferUIMessageData<UI_MESSAGE>>

generateId?IdGenerator$1

A way to provide a function that is going to be used for ids for messages and the chat. If not provided the default AI SDK `generateId` is used.

onToolCall?ChatOnToolCallCallback<UI_MESSAGE>

Optional callback function that is invoked when a tool call is received. Intended for automatic client-side tool execution. To add the tool output, call `addToolOutput` without awaiting it inside this callback. The callback's return value is not used.

onData?ChatOnDataCallback<UI_MESSAGE>

Optional callback function that is called when a data part is received.

sendAutomaticallyWhen?(options: { messages: UI_MESSAGE[]; }) => boolean | PromiseLike<boolean>

When provided, this function will be called when the stream is finished or a tool call is added to determine if the current messages should be resubmitted.

throttle?number | undefined

toCreateMessage?CustomToCreateMessageFunction

onResumeError?((error: unknown) => void) | undefined

Called when an automatic resumable stream reconnect fails. Use this to surface a toast, report telemetry, or mark the thread as needing a retry. The failed stream id is cleared after the callback unless a newer id has replaced it.

transport?ChatTransport<UI_MESSAGE> | (() => ChatTransport<UI_MESSAGE>) | undefined

The transport threads send through. A factory is invoked once per thread so each thread owns its instance. A plain `AssistantChatTransport` instance is cloned per thread (its assistant-ui wiring is per thread); any other transport instance is shared as-is. Defaults to one `AssistantChatTransport` per thread.

AISDKThreads props["0"]["transport"]
sendMessages(options: { /** The type of message submission - either new message or regeneration */ trigger: 'submit-message' | 'regenerate-message'; /** Unique identifier for the chat session */ chatId: string; /** ID of the message to regenerate, or undefined for new messages */ messageId: string | undefined; /** Array of UI messages representing the conversation history */ messages: UI_MESSAGE[]; /** Signal to abort the request if needed */ abortSignal: AbortSignal | undefined; } & ChatRequestOptions) => Promise<ReadableStream<UIMessageChunk>>

Sends messages to the chat API endpoint and returns a streaming response. This method handles both new message submission and message regeneration. It supports real-time streaming of responses through UIMessageChunk events.

reconnectToStream(options: { /** Unique identifier for the chat session to reconnect to */ chatId: string; /** Signal to abort the reconnection request if needed */ abortSignal?: AbortSignal; } & ChatRequestOptions) => Promise<ReadableStream<UIMessageChunk> | null>

Reconnects to an existing streaming response for the specified chat session. This method is used to resume streaming when a connection is interrupted or when resuming a chat session. It's particularly useful for maintaining continuity in long-running conversations or recovering from network issues.

cloud?AssistantCloud | undefined

When set, the thread list is a `RemoteThreadList` backed by this assistant-cloud. Omit it to keep the in-memory list. Every visited cloud thread stays mounted, so an in-flight run continues after a switch and stops on delete; per-thread history loads once per thread.

AISDKThreads props["0"]["cloud"]
threadsAssistantCloudThreads

AssistantCloudThreads
messagesAssistantCloudThreadMessages

cloudAssistantCloudAPI

list(query?: AssistantCloudThreadsListQuery) => Promise<AssistantCloudThreadsListResponse>

get(threadId: string) => Promise<CloudThread>

create(body: AssistantCloudThreadsCreateBody) => Promise<AssistantCloudThreadsCreateResponse>

update(threadId: string, body: AssistantCloudThreadsUpdateBody) => Promise<void>

claim(body: AssistantCloudThreadsClaimBody) => Promise<AssistantCloudThreadsClaimResponse>

Moves every thread of the anonymous identity behind `refresh_token` into the caller's workspace.

delete(threadId: string) => Promise<void>

projectsAssistantCloudProjects

AssistantCloudProjects
threadsAssistantCloudProjectThreads

auth__object

__object
tokensAssistantCloudAuthTokens

invalidate() => void

runsAssistantCloudRuns

AssistantCloudRuns
cloudAssistantCloudAPI

stream(body: AssistantCloudRunsStreamBody) => Promise<AssistantStream>

report(body: AssistantCloudRunReport) => Promise<{ run_id: string; }>

filesAssistantCloudFiles

AssistantCloudFiles
cloudAssistantCloudAPI

pdfToImages(body: PdfToImagesRequestBody) => Promise<PdfToImagesResponse>deprecated

Deprecated: Assistant Cloud has no PDF conversion endpoint, so this request always rejects with a `CloudAPIError` whose `status` is 404.

generatePresignedUploadUrl(body: GeneratePresignedUploadUrlRequestBody) => Promise<GeneratePresignedUploadUrlResponse>

generatePresignedDownloadUrl(body: { key: string; } | { url: string; }) => Promise<GeneratePresignedDownloadUrlResponse>

eventsAssistantCloudEvents

AssistantCloudEvents
bufferAssistantCloudEvent[]

timer?ReturnType<typeof setTimeout> | undefined

flushing?Promise<void> | undefined

retryTimer?ReturnType<typeof setTimeout> | undefined

resolveRetryDelay?(() => void) | undefined

bestEffortRequestedboolean

generationnumber

cloudAssistantCloudAPI

isEnabled() => boolean

listeningboolean

track(event: AssistantCloudEvent) => void

listen() => void

unlisten() => void

dispose() => void

clearPending() => void

onVisibilityChange() => void

flushBestEffort() => Promise<void>

flush(retryFailures: boolean) => Promise<void>

flushPending(retryFailures: boolean) => Promise<void>

waitForRetry(delay: number) => Promise<void>

interruptRetryDelay() => void

scheduleFlush() => void

clearFlushTimer() => void

scoresAssistantCloudScores

AssistantCloudScores
cloudAssistantCloudAPI

create(body: AssistantCloudScoreBody) => Promise<AssistantCloudScoreResponse>

telemetryAssistantCloudTelemetryConfig

AssistantCloudTelemetryConfig
enabled?boolean

Enables Assistant Cloud telemetry. Defaults to `true`. Set to `false` to disable both run reports and engagement events.

events?boolean

Enables Assistant Cloud engagement events. Defaults to `true` when telemetry is enabled. Set to `false` to keep run reports while disabling engagement events.

messages?boolean

Stores the messages of runtimes whose backend keeps the transcript (LangGraph, LangChain, Google ADK, custom external stores), so the dashboard can show them. Defaults to `true` when telemetry is enabled. Set to `false` to keep run reports and events without storing those messages.

release?string

environment?string

tags?string[]

beforeReport?( report: AssistantCloudRunReport, ) => AssistantCloudRunReport | null

Called before each telemetry report is sent. Return a modified report to enrich it (e.g. add `model_id`), or return `null` to skip the report.

registerSdk(sdk: SdkIdentity) => void

scopeId?string | undefined

Stable identity for the account or workspace owning Cloud runtime state. Provide it from the first render and change it when that scope changes.

threadId?string | undefined

Controlled thread id for the cloud list. Ignored without `cloud`.

onThreadIdChange?((threadId: string | undefined) => void) | undefined

Called with the settled remote id when the cloud list changes thread.

length0 | 1

toString() => string

toLocaleString{ (): string; (locales: string | string[], options?: Intl.NumberFormatOptions & Intl.DateTimeFormatOptions): string; }

pop() => AISDKThreadsOptions<UI_MESSAGE>

push(...items: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => number

concat{ (...items: ConcatArray<AISDKThreadsOptions<UI_MESSAGE> | undefined>[]): (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]; (...items: (AISDKThreadsOptions<UI_MESSAGE> | ConcatArray<AISDKThreadsOptions<UI_MESSAGE> | undefined> | undefined)[]): (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]; }

join(separator?: string) => string

reverse() => (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]

shift() => AISDKThreadsOptions<UI_MESSAGE>

slice(start?: number, end?: number) => (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]

sort(compareFn?: ((a: AISDKThreadsOptions<UI_MESSAGE> | undefined, b: AISDKThreadsOptions<UI_MESSAGE> | undefined) => number) | undefined) => [options?: AISDKThreadsOptions<UI_MESSAGE> | undefined]

splice{ (start: number, deleteCount?: number): (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]; (start: number, deleteCount: number, ...items: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]): (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]; }

unshift(...items: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => number

indexOf(searchElement: AISDKThreadsOptions<UI_MESSAGE> | undefined, fromIndex?: number) => number

lastIndexOf(searchElement: AISDKThreadsOptions<UI_MESSAGE> | undefined, fromIndex?: number) => number

every{ <S>(predicate: (value: AISDKThreadsOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => value is S, thisArg?: any): this is S[]; (predicate: (value: AISDKThreadsOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => unknown, thisArg?: any): boolean; }

some(predicate: (value: AISDKThreadsOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => unknown, thisArg?: any) => boolean

forEach(callbackfn: (value: AISDKThreadsOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => void, thisArg?: any) => void

map<U>(callbackfn: (value: AISDKThreadsOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => U, thisArg?: any) => U[]

filter{ <S>(predicate: (value: AISDKThreadsOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => value is S, thisArg?: any): S[]; (predicate: (value: AISDKThreadsOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => unknown, thisArg?: any): (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]; }

reduce{ (callbackfn: (previousValue: AISDKThreadsOptions<UI_MESSAGE> | undefined, currentValue: AISDKThreadsOptions<UI_MESSAGE> | undefined, currentIndex: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => AISDKThreadsOptions<UI_MESSAGE> | undefined): AISDKThreadsOptions<UI_MESSAGE> | undefined; (callbackfn: (previousValue: AISDKThreadsOptions<UI_MESSAGE> | undefined, currentValue: AISDKThreadsOptions<UI_MESSAGE> | undefined, currentIndex: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => AISDKThreadsOptions<UI_MESSAGE> | undefined, initialValue: AISDKThreadsOptions<UI_MESSAGE> | undefined): AISDKThreadsOptions<UI_MESSAGE> | undefined; <U>(callbackfn: (previousValue: U, currentValue: AISDKThreadsOptions<UI_MESSAGE> | undefined, currentIndex: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => U, initialValue: U): U; }

reduceRight{ (callbackfn: (previousValue: AISDKThreadsOptions<UI_MESSAGE> | undefined, currentValue: AISDKThreadsOptions<UI_MESSAGE> | undefined, currentIndex: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => AISDKThreadsOptions<UI_MESSAGE> | undefined): AISDKThreadsOptions<UI_MESSAGE> | undefined; (callbackfn: (previousValue: AISDKThreadsOptions<UI_MESSAGE> | undefined, currentValue: AISDKThreadsOptions<UI_MESSAGE> | undefined, currentIndex: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => AISDKThreadsOptions<UI_MESSAGE> | undefined, initialValue: AISDKThreadsOptions<UI_MESSAGE> | undefined): AISDKThreadsOptions<UI_MESSAGE> | undefined; <U>(callbackfn: (previousValue: U, currentValue: AISDKThreadsOptions<UI_MESSAGE> | undefined, currentIndex: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => U, initialValue: U): U; }

find{ <S>(predicate: (value: AISDKThreadsOptions<UI_MESSAGE> | undefined, index: number, obj: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => value is S, thisArg?: any): S | undefined; (predicate: (value: AISDKThreadsOptions<UI_MESSAGE> | undefined, index: number, obj: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => unknown, thisArg?: any): AISDKThreadsOptions<UI_MESSAGE> | undefined; }

findIndex(predicate: (value: AISDKThreadsOptions<UI_MESSAGE> | undefined, index: number, obj: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => unknown, thisArg?: any) => number

fill(value: AISDKThreadsOptions<UI_MESSAGE> | undefined, start?: number, end?: number) => [options?: AISDKThreadsOptions<UI_MESSAGE> | undefined]

copyWithin(target: number, start: number, end?: number) => [options?: AISDKThreadsOptions<UI_MESSAGE> | undefined]

entries() => ArrayIterator<[number, AISDKThreadsOptions<UI_MESSAGE> | undefined]>

keys() => ArrayIterator<number>

values() => ArrayIterator<AISDKThreadsOptions<UI_MESSAGE> | undefined>

includes(searchElement: AISDKThreadsOptions<UI_MESSAGE> | undefined, fromIndex?: number) => boolean

flatMap<U, This>(callback: (this: This, value: AISDKThreadsOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => U | readonly U[], thisArg?: This | undefined) => U[]

flat<A, D>(this: A, depth?: D | undefined) => FlatArray<A, D>[]

at(index: number) => AISDKThreadsOptions<UI_MESSAGE>

findLast{ <S>(predicate: (value: AISDKThreadsOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => value is S, thisArg?: any): S | undefined; (predicate: (value: AISDKThreadsOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => unknown, thisArg?: any): AISDKThreadsOptions<UI_MESSAGE> | undefined; }

findLastIndex(predicate: (value: AISDKThreadsOptions<UI_MESSAGE> | undefined, index: number, array: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]) => unknown, thisArg?: any) => number

toReversed() => (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]

toSorted(compareFn?: ((a: AISDKThreadsOptions<UI_MESSAGE> | undefined, b: AISDKThreadsOptions<UI_MESSAGE> | undefined) => number) | undefined) => (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]

toSpliced{ (start: number, deleteCount: number, ...items: (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]): (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]; (start: number, deleteCount?: number): (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]; }

with(index: number, value: AISDKThreadsOptions<UI_MESSAGE> | undefined) => (AISDKThreadsOptions<UI_MESSAGE> | undefined)[]

AISDKToolkit

AISDKToolkit
constructor?(options: AISDKToolkitOptions) => AISDKToolkit

#toolkit?Toolkit

#mcpClients?Map<string, Promise<MCPClient>>

tools?(options: AISDKToolkitToolsOptions = {}) => Promise<ToolSet>

close?() => Promise<void>

#mcpTools?() => Promise<McpToolSet>

#mcpClient?(name: string, config: McpServerConfig, startedAt: number) => Promise<MCPClient>

AssistantChatTransport

AssistantChatTransport
constructor?(initOptions?: AssistantChatTransportInitOptions<UI_MESSAGE>) => AssistantChatTransport

__internal_clone?() => AssistantChatTransport<UI_MESSAGE>

Constructs an unwired copy with the same init options.

setRuntime?(runtime: AssistantRuntime) => void

getResumableAdapter?() => AssistantChatResumableOptions | undefined

__internal_setGetThreadListItem?(getter: () => InitializableThreadListItem | undefined) => void

createResumableSessionStorage

sessionStorage-backed storage for the pending resumable stream id. See the Resumable Streams guide for end-to-end wiring.

createResumableSessionStorage
options?{ key?: string | (() => string | undefined); }

{ key?: string | (() => string | undefined); }
key?string | (() => string | undefined)

Storage key for the pending stream id. A static string namespaces per route or chat surface. A getter is read lazily on every access, so the key can be derived from the active thread's identity; while the getter returns `undefined`, reads report no pending stream and writes are dropped, so a thread whose identity is not known yet never touches another thread's key. Under a remote thread list with more than one thread, scope the key per thread and create one storage instance per thread runtime rather than a single shared one. A shared key is written and cleared by whichever thread acts last, so one conversation's stream can resume inside another.

frontendTools

const frontendTools: (tools: FrontendTools) => ToolSet;

getThreadMessageTokenUsage

getThreadMessageTokenUsage
messageTokenUsageExtractableMessage

TokenUsageExtractableMessage
role?string

metadata?unknown

injectQuoteContext

Injects quote context into messages as markdown blockquotes.

Use this in your route handler before convertToModelMessages so the LLM sees the quoted text that the user is referring to.

import { convertToModelMessages, streamText } from "ai";
import { injectQuoteContext } from "@assistant-ui/ai-sdk";

export async function POST(req: Request) {
  const { messages } = await req.json();
  const result = streamText({
    model: myModel,
    messages: await convertToModelMessages(injectQuoteContext(messages)),
  });
  return result.toUIMessageStreamResponse();
}
injectQuoteContext
messagesUIMessage<unknown, UIDataTypes, UITools>[]

RESUMABLE_STREAM_ID_HEADER

Response header used by the Resumable Streams server and client wiring.

const RESUMABLE_STREAM_ID_HEADER: "x-resumable-stream-id";

unstable_injectInteractableContext

Injects interactable state snapshots into messages as model-visible text.

Mirrors injectQuoteContext: reads the frozen snapshot stamped on a user message's metadata.custom.interactables (by the interactables scope at send time) and prepends a text part. Run this in your route handler before convertToModelMessages, which otherwise ignores metadata.custom.

Wording is consumer-owned — pass format to control how each snapshot reads. A snapshot may originate from a user edit or an agent update_* call, so the default phrasing is neutral. Keep the instance id visible in custom wording: the model needs it to address the update_* tool's id parameter. A custom format must also handle entries with partial: true, whose state carries only the fields that changed since the model's last known state.

import { convertToModelMessages, streamText } from "ai";
import { unstable_injectInteractableContext } from "@assistant-ui/ai-sdk";

export async function POST(req: Request) {
  const { messages } = await req.json();
  const result = streamText({
    model: myModel,
    messages: await convertToModelMessages(unstable_injectInteractableContext(messages)),
  });
  return result.toUIMessageStreamResponse();
}
unstable_injectInteractableContext
messagesUIMessage<unknown, UIDataTypes, UITools>[]

format?(item: Unstable_InteractableSnapshotEntry) => string

useAISDKChat

The underlying useChat helpers object, for advanced views — reach resumeStream, clearError, or the raw sendMessage without forking the runtime hook. undefined when the current thread is not backed by the AI SDK runtime.

const useAISDKChat: <UI_MESSAGE extends UIMessage = UIMessage<unknown, UIDataTypes, UITools>>() => UseChatHelpers<UI_MESSAGE> | undefined;

useAISDKError

Read the last AI SDK chat error object from the runtime extras. undefined when there is no error or the current thread is not backed by the AI SDK runtime.

const useAISDKError: () => Error | undefined;

useAISDKRuntime

useAISDKRuntime
chatHelpersUseChatHelpers<UI_MESSAGE>

UseChatHelpers
idstring

The id of the chat.

setMessages(messages: UI_MESSAGE[] | ((messages: UI_MESSAGE[]) => UI_MESSAGE[])) => void

Update the `messages` state locally. This is useful when you want to edit the messages on the client, and then trigger the `reload` method manually to regenerate the AI response.

error?Error | undefined

UseChatHelpers["error"]
namestring

messagestring

stack?string

cause?unknown

statusChatStatus

Hook status: - `submitted`: The message has been sent to the API and we're awaiting the start of the response stream. - `streaming`: The response is actively streaming in from the API, receiving chunks of data. - `ready`: The full response has been received and processed; a new user message can be submitted. - `error`: An error occurred during the API request, preventing successful completion.

addToolResultChatAddToolOutputFunction<UI_MESSAGE>deprecated

Deprecated: Use addToolOutput

stop() => Promise<void>

Abort the current request immediately, keep the generated tokens if any.

messagesUI_MESSAGE[]

sendMessage(message?: (CreateUIMessage<UI_MESSAGE> & { text?: never; files?: never; messageId?: string; }) | { text: string; files?: FileList | FileUIPart[]; metadata?: InferUIMessageMetadata<UI_MESSAGE>; parts?: never; messageId?: string; } | { files: FileList | FileUIPart[]; metadata?: InferUIMessageMetadata<UI_MESSAGE>; parts?: never; messageId?: string; }, options?: ChatRequestOptions) => Promise<void>

Appends or replaces a user message to the chat list. This triggers the API call to fetch the assistant's response. If a messageId is provided, the message will be replaced.

regenerate({ messageId, ...options }?: { messageId?: string; } & ChatRequestOptions) => Promise<void>

Regenerate the assistant message with the provided message id. If no message id is provided, the last assistant message will be regenerated.

resumeStream(options?: ChatRequestOptions) => Promise<void>

Attempt to resume an ongoing streaming response.

addToolOutputChatAddToolOutputFunction<UI_MESSAGE>

addToolApprovalResponseChatAddToolApproveResponseFunction

clearError() => void

Clear the error state and set the status to ready if the chat is in an error state.

adapter?AISDKRuntimeAdapter<UI_MESSAGE>

AISDKRuntimeAdapter
suggestions?readonly ThreadSuggestion[] | undefined

isDisabled?boolean | undefined

Whether the entire thread is disabled. When `true`, the composer's input is also disabled (the user cannot type, attach files, or submit). For a narrower gate that keeps the input usable but blocks only sending, use `isSendDisabled`.

isSendDisabled?boolean | undefined

Whether sending new messages is currently disabled. When `true`, the thread composer's input remains usable but `send()` becomes a no-op and the thread composer's `canSend` is `false`. Use this to gate sending on external React state (e.g. while tool config is loading) without disabling the input itself the way `isDisabled` does. Edit composers (saving message edits) intentionally ignore this flag.

unstable_capabilities?AISDKRuntimeAdapter["unstable_capabilities"]unstable

AISDKRuntimeAdapter["unstable_capabilities"]
copy?boolean | undefined

adapters?(NonNullable<ExternalStoreAdapter["adapters"]> & { history?: ThreadHistoryAdapter | undefined; suggestion?: SuggestionAdapter | undefined; }) | undefined

AISDKRuntimeAdapter["adapters"]
attachments?AttachmentAdapter | undefined

AISDKRuntimeAdapter["adapters"]["attachments"]
acceptstring

add(state: { file: File; }) => Promise<PendingAttachment> | AsyncGenerator<PendingAttachment, void>

remove(attachment: Attachment) => Promise<void>

send(attachment: PendingAttachment, options?: { signal?: AbortSignal; }) => Promise<CompleteAttachment>

speech?SpeechSynthesisAdapter | undefined

AISDKRuntimeAdapter["adapters"]["speech"]
speak(text: string) => SpeechSynthesisAdapter.Utterance

dictation?DictationAdapter | undefined

AISDKRuntimeAdapter["adapters"]["dictation"]
listen() => DictationAdapter.Session

disableInputDuringDictation?boolean

voice?RealtimeVoiceAdapter | undefined

AISDKRuntimeAdapter["adapters"]["voice"]
connect(options: { abortSignal?: AbortSignal; }) => RealtimeVoiceAdapter.Session

feedback?FeedbackAdapter | undefined

AISDKRuntimeAdapter["adapters"]["feedback"]
submit(feedback: FeedbackAdapterFeedback) => void

threadList?ExternalStoreThreadListAdapter | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

AISDKRuntimeAdapter["adapters"]["threadList"]
threadId?string | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

isLoading?boolean | undefined

threads?readonly ExternalStoreThreadData<"regular">[] | undefined

archivedThreads?readonly ExternalStoreThreadData<"archived">[] | undefined

onSwitchToNewThread?(() => Promise<void> | void) | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

onSwitchToThread?((threadId: string) => Promise<void> | void) | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

onRename?( threadId: string, newTitle: string, ) => (Promise<void> | void) | undefined

onUpdateCustom?(( threadId: string, custom: Record<string, unknown> | undefined, ) => Promise<void> | void) | undefined

onArchive?((threadId: string) => Promise<void> | void) | undefined

onUnarchive?((threadId: string) => Promise<void> | void) | undefined

onDelete?((threadId: string) => Promise<void> | void) | undefined

history?ThreadHistoryAdapter | undefined

AISDKRuntimeAdapter["adapters"]["history"]
scopeId?string | undefined

Stable identity for the storage scope read by LocalRuntime. Keep it the same when recreating the adapter for one thread, account, or workspace, and change it before loading a different scope. Adapters that omit it are treated as sharing one scope. External-history runtimes do not read it.

unstable_copy?(( branch: readonly ThreadMessage[], messageIds: readonly string[], ) => Promise<void>) | undefinedunstable

Keeps a copy of messages whose source of truth is the runtime's backend. `branch` is the conversation from its first message to its last, and `messageIds` names the ones in it that are new or changed; the adapter stores those, and first any earlier message of the branch it does not hold yet, keyed by each message's own id. It is undefined while the adapter keeps no copies, so the runtime then neither copies nor records tool interactions.

load() => Promise<ExportedMessageRepository & { state?: ReadonlyJSONValue; unstable_resume?: boolean; }>

resume?(options: ChatModelRunOptions) => AsyncGenerator<ChatModelRunResult, void, unknown>

append(item: ExportedMessageRepositoryItem) => Promise<void>

update?(item: ExportedMessageRepositoryItem) => Promise<void>

Rewrites a previously appended message in place, keyed by its message id. Adapters that implement this let a runtime persist a run paused for tool approval, finalize the same message once the run resumes, and record a tool result that arrives after the message settled, which can be a message later turns follow. Without it, a paused run is appended when it ends or when a later turn follows its message while the run is still open; anything that run adds after the append is not stored. An update may arrive for an id whose earlier write failed; treat it as an upsert keyed on the message id rather than assuming the entry exists.

delete?(items: ExportedMessageRepositoryItem[]) => Promise<void>

Deletes messages from history. The runtime may send a second delete for the same message after a write it issued before the delete lands; treat deleting a missing entry as success.

withFormat?<TMessage, TStorageFormat extends Record<string, unknown>>(formatAdapter: MessageFormatAdapter<TMessage, TStorageFormat>) => GenericThreadHistoryAdapter<TMessage>

Required when used with `useAISDKRuntime` / `useChatRuntime`.

suggestion?SuggestionAdapter | undefined

AISDKRuntimeAdapter["adapters"]["suggestion"]
key?string | number | symbol | undefined

Changes when the adapter should regenerate suggestions for a settled run.

generate( options: SuggestionAdapterGenerateOptions, ) => | Promise<readonly ThreadSuggestion[]> | AsyncGenerator<readonly ThreadSuggestion[], void>

toCreateMessage?CustomToCreateMessageFunction

unstable_messageRepositoryInstance?MessageRepository | undefinedunstable

AISDKRuntimeAdapter["unstable_messageRepositoryInstance"]
messagesMap<string, RepositoryMessage>

Map
clear() => void

delete(key: string) => boolean

forEach(callbackfn: (value: { children: string[]; next: (RepositoryParent & { prev: (RepositoryParent & any) | null; current: ThreadMessage; level: number; }) | null; } & { prev: ({ children: string[]; next: (RepositoryParent & any) | null; } & any) | null; current: ThreadMessage; level: number; }, key: string, map: Map<string, { children: string[]; next: (RepositoryParent & { prev: (RepositoryParent & any) | null; current: ThreadMessage; level: number; }) | null; } & { prev: ({ children: string[]; next: (RepositoryParent & any) | null; } & any) | null; current: ThreadMessage; level: number; }>) => void, thisArg?: any) => void

get(key: string) => ({ children: string[]; next: (RepositoryParent & { prev: (RepositoryParent & any) | null; current: ThreadMessage; level: number; }) | null; } & { prev: ({ children: string[]; next: (RepositoryParent & any) | null; } & any) | null; current: ThreadMessage; level: number; })

has(key: string) => boolean

set(key: string, value: { children: string[]; next: (RepositoryParent & { prev: (RepositoryParent & any) | null; current: ThreadMessage; level: number; }) | null; } & { prev: ({ children: string[]; next: (RepositoryParent & any) | null; } & any) | null; current: ThreadMessage; level: number; }) => Map<string, { children: string[]; next: (RepositoryParent & { prev: (RepositoryParent & any) | null; current: ThreadMessage; level: number; }) | null; } & { prev: ({ children: string[]; next: (RepositoryParent & any) | null; } & any) | null; current: ThreadMessage; level: number; }>

sizenumber

entries() => MapIterator<[string, { children: string[]; next: (RepositoryParent & { prev: (RepositoryParent & any) | null; current: ThreadMessage; level: number; }) | null; } & { prev: ({ children: string[]; next: (RepositoryParent & any) | null; } & any) | null; current: ThreadMessage; level: number; }]>

keys() => MapIterator<string>

values() => MapIterator<{ children: string[]; next: (RepositoryParent & { prev: (RepositoryParent & any) | null; current: ThreadMessage; level: number; }) | null; } & { prev: ({ children: string[]; next: (RepositoryParent & any) | null; } & any) | null; current: ThreadMessage; level: number; }>

headRepositoryMessage | null

AISDKRuntimeAdapter["unstable_messageRepositoryInstance"]["head"]
childrenstring[]

nextRepositoryMessage | null

prevRepositoryMessage | null

currentThreadMessage

levelnumber

rootRepositoryParent

RepositoryParent
childrenstring[]

nextRepositoryMessage | null

updateLevels(message: RepositoryMessage, newLevel: number) => void

selectPathTo(message: RepositoryMessage) => void

performOp(newParent: RepositoryMessage | null, child: RepositoryMessage, operation: "cut" | "link" | "relink") => void

_messagesCachedValue<readonly ThreadMessage[]>

CachedValue
_valueT | null

func() => T

valuereadonly ThreadMessage[]

dirty() => void

headIdstring | null

canonicalHeadIdstring | null

getMessages(headId?: string) => readonly ThreadMessage[]

addOrUpdateMessage(parentId: string | null, message: ThreadMessage) => void

getMessage(messageId: string) => { parentId: string | null; message: ThreadMessage; index: number; }

deleteMessage(messageId: string, replacementId?: string | null | undefined) => void

hasChildren(messageId: string) => boolean

getBranches(messageId: string) => string[]

evictOffBranchOptimisticMessages(previousHead: RepositoryMessage | null, currentHead: RepositoryMessage | null) => void

Evicts optimistic messages (`metadata.isOptimistic`) on the branch the head just moved away from. Only that branch is walked, so an optimistic message added off the head branch (such as the server-id copy that replaces a client-id placeholder before the head moves to it) is kept until a `switchToBranch` or `resetHead` moves the head off the branch it is on. Keeps a client→server id swap from leaving a phantom sibling.

switchToBranch(messageId: string) => void

resetHead(messageId: string | null) => void

clear() => void

export() => ExportedMessageRepository

import({ headId, messages }: ExportedMessageRepository) => void

unstable_hostApprovalOwner?object | undefinedunstable

The object a host answer belongs to, normally the `Chat` the runtime renders. A host answer never reaches the `useChat` messages, so without an owner it lives only as long as this runtime. With an owner, it survives runtime remounts over the same chat.

cancelPendingToolCallsOnSendboolean | undefined= true

Whether to automatically cancel pending interactive tool calls when the user sends a new message. When enabled (default), the pending tool calls will be marked as failed with an error message indicating the user cancelled the tool call by sending a new message.

onResume?ExternalStoreAdapter["onResume"]

Called when `runtime.thread.resumeRun(config)` is invoked. When omitted, `resumeRun` throws `"Runtime does not support resuming runs."`. Provide this to bridge resume invocations into a custom replay channel (for example, an SSE reconnect endpoint keyed by turn id).

onResumeToolCall?ExternalStoreAdapter["onResumeToolCall"]

Called when `runtime.thread.resumeToolCall(options)` is invoked for a tool call the in-process tracker does not own. When omitted, `resumeToolCall` throws `"Tool call ${toolCallId} is not waiting for resume."`. Provide this to bridge resume-tool-call invocations into a custom handler.

onRespondToToolApproval?(( response: RespondToToolApprovalOptions, context: { toolCallId: string; toolName: string; /** Sends this response through the AI SDK's `addToolApprovalResponse`, which carries only `approved` and `reason`. */ respondViaAISDK: () => Promise<void>; }, ) => Promise<void> | void) | undefined

Answers tool approval requests through a host-owned channel instead of the AI SDK's `addToolApprovalResponse`. Called for every approval request in the thread with the complete response, including option and free-form answers. Hand requests the host does not own to `respondViaAISDK`, which is what runs when this option is omitted. The answer applies to the approval when the handler starts and is removed if it throws. It is never written into the `useChat` messages, so `sendAutomaticallyWhen` cannot forward it. With a history adapter, the answer is stored with its message once the handler resolves and returns on reload. With `unstable_hostApprovalOwner`, in-memory answers survive runtime remounts over the same chat; without an owner or history adapter, they last only as long as the runtime. Bring the resumed run back into the chat, for example with `resumeStream`, and have the endpoint refuse a second resume. While a handler is set, an approval's `display`, `allowFreeform`, `dismissible` and `options` reach the renderer, because the handler can receive answers the AI SDK cannot carry. A stream declares them through the `approvalDescriptor` of its `tool-approval-request` chunk, the one approval field the AI SDK keeps opaque; the converter reads the request and answer fields from that descriptor when the approval itself lacks them.

joinStrategy?JoinStrategy | undefined

How consecutive assistant messages are rendered. `"concat-content"` (the default) merges them into a single thread message. `"none"` keeps each assistant message as its own thread message, which is useful when a backend persists proactive or consecutive assistant messages as separate entries.

messageRepository?MessageFormatRepository<UI_MESSAGE>

A branch-aware AI SDK message tree seeded once when `useChat` is empty. After that seed, live updates come only from `useChat`. A later empty chat or a new object identity does not reload the tree.

MessageFormatRepository
headId?string | null

messagesMessageFormatItem<TMessage>[]

unstable_onBranchChange?ExternalStoreAdapter["unstable_onBranchChange"]deprecatedunstable

Called after an explicit `switchToBranch` (for example a BranchPicker click). Complements `setMessages` and does not enable switching by itself.

Deprecated: This API is still under active development and might change without notice.

useChatRuntime

useChatRuntime
options?UseChatRuntimeOptions<UI_MESSAGE>

UseChatRuntimeOptions
suggestions?readonly ThreadSuggestion[] | undefined

id?string

A unique identifier for the chat. If not provided, a random one will be generated.

onError?ChatOnErrorCallback

Callback function to be called when an error is encountered.

messages?UI_MESSAGE[]

isDisabled?boolean | undefined

Whether the entire thread is disabled. When `true`, the composer's input is also disabled (the user cannot type, attach files, or submit). For a narrower gate that keeps the input usable but blocks only sending, use `isSendDisabled`.

isSendDisabled?boolean | undefined

Whether sending new messages is currently disabled. When `true`, the thread composer's input remains usable but `send()` becomes a no-op and the thread composer's `canSend` is `false`. Use this to gate sending on external React state (e.g. while tool config is loading) without disabling the input itself the way `isDisabled` does. Edit composers (saving message edits) intentionally ignore this flag.

messageRepository?AISDKRuntimeAdapter<UI_MESSAGE>["messageRepository"]

UseChatRuntimeOptions["messageRepository"]
headId?string | null

messagesMessageFormatItem<TMessage>[]

unstable_onBranchChange?AISDKRuntimeAdapter["unstable_onBranchChange"]unstable

onResume?AISDKRuntimeAdapter["onResume"]

onResumeToolCall?AISDKRuntimeAdapter["onResumeToolCall"]

onRespondToToolApproval?AISDKRuntimeAdapter["onRespondToToolApproval"]

adapters?AISDKRuntimeAdapter["adapters"] | undefined

UseChatRuntimeOptions["adapters"]
attachments?AttachmentAdapter | undefined

UseChatRuntimeOptions["adapters"]["attachments"]
acceptstring

add(state: { file: File; }) => Promise<PendingAttachment> | AsyncGenerator<PendingAttachment, void>

remove(attachment: Attachment) => Promise<void>

send(attachment: PendingAttachment, options?: { signal?: AbortSignal; }) => Promise<CompleteAttachment>

speech?SpeechSynthesisAdapter | undefined

UseChatRuntimeOptions["adapters"]["speech"]
speak(text: string) => SpeechSynthesisAdapter.Utterance

dictation?DictationAdapter | undefined

UseChatRuntimeOptions["adapters"]["dictation"]
listen() => DictationAdapter.Session

disableInputDuringDictation?boolean

voice?RealtimeVoiceAdapter | undefined

UseChatRuntimeOptions["adapters"]["voice"]
connect(options: { abortSignal?: AbortSignal; }) => RealtimeVoiceAdapter.Session

feedback?FeedbackAdapter | undefined

UseChatRuntimeOptions["adapters"]["feedback"]
submit(feedback: FeedbackAdapterFeedback) => void

threadList?ExternalStoreThreadListAdapter | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

UseChatRuntimeOptions["adapters"]["threadList"]
threadId?string | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

isLoading?boolean | undefined

threads?readonly ExternalStoreThreadData<"regular">[] | undefined

archivedThreads?readonly ExternalStoreThreadData<"archived">[] | undefined

onSwitchToNewThread?(() => Promise<void> | void) | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

onSwitchToThread?((threadId: string) => Promise<void> | void) | undefineddeprecated

Deprecated: This API is still under active development and might change without notice.

onRename?( threadId: string, newTitle: string, ) => (Promise<void> | void) | undefined

onUpdateCustom?(( threadId: string, custom: Record<string, unknown> | undefined, ) => Promise<void> | void) | undefined

onArchive?((threadId: string) => Promise<void> | void) | undefined

onUnarchive?((threadId: string) => Promise<void> | void) | undefined

onDelete?((threadId: string) => Promise<void> | void) | undefined

history?ThreadHistoryAdapter | undefined

UseChatRuntimeOptions["adapters"]["history"]
scopeId?string | undefined

Stable identity for the storage scope read by LocalRuntime. Keep it the same when recreating the adapter for one thread, account, or workspace, and change it before loading a different scope. Adapters that omit it are treated as sharing one scope. External-history runtimes do not read it.

unstable_copy?(( branch: readonly ThreadMessage[], messageIds: readonly string[], ) => Promise<void>) | undefinedunstable

Keeps a copy of messages whose source of truth is the runtime's backend. `branch` is the conversation from its first message to its last, and `messageIds` names the ones in it that are new or changed; the adapter stores those, and first any earlier message of the branch it does not hold yet, keyed by each message's own id. It is undefined while the adapter keeps no copies, so the runtime then neither copies nor records tool interactions.

load() => Promise<ExportedMessageRepository & { state?: ReadonlyJSONValue; unstable_resume?: boolean; }>

resume?(options: ChatModelRunOptions) => AsyncGenerator<ChatModelRunResult, void, unknown>

append(item: ExportedMessageRepositoryItem) => Promise<void>

update?(item: ExportedMessageRepositoryItem) => Promise<void>

Rewrites a previously appended message in place, keyed by its message id. Adapters that implement this let a runtime persist a run paused for tool approval, finalize the same message once the run resumes, and record a tool result that arrives after the message settled, which can be a message later turns follow. Without it, a paused run is appended when it ends or when a later turn follows its message while the run is still open; anything that run adds after the append is not stored. An update may arrive for an id whose earlier write failed; treat it as an upsert keyed on the message id rather than assuming the entry exists.

delete?(items: ExportedMessageRepositoryItem[]) => Promise<void>

Deletes messages from history. The runtime may send a second delete for the same message after a write it issued before the delete lands; treat deleting a missing entry as success.

withFormat?<TMessage, TStorageFormat extends Record<string, unknown>>(formatAdapter: MessageFormatAdapter<TMessage, TStorageFormat>) => GenericThreadHistoryAdapter<TMessage>

Required when used with `useAISDKRuntime` / `useChatRuntime`.

suggestion?SuggestionAdapter | undefined

UseChatRuntimeOptions["adapters"]["suggestion"]
key?string | number | symbol | undefined

Changes when the adapter should regenerate suggestions for a settled run.

generate( options: SuggestionAdapterGenerateOptions, ) => | Promise<readonly ThreadSuggestion[]> | AsyncGenerator<readonly ThreadSuggestion[], void>

unstable_capabilities?UseChatRuntimeOptions["unstable_capabilities"]unstable

UseChatRuntimeOptions["unstable_capabilities"]
copy?boolean | undefined

onFinish?ChatOnFinishCallback<UI_MESSAGE>

Function that is called when the assistant response has finished streaming.

joinStrategy?AISDKRuntimeAdapter["joinStrategy"]

messageMetadataSchema?FlexibleSchema$1<UI_MESSAGE['metadata']>

dataPartSchemas?UIDataTypesToSchemas<InferUIMessageData<UI_MESSAGE>>

generateId?IdGenerator$1

A way to provide a function that is going to be used for ids for messages and the chat. If not provided the default AI SDK `generateId` is used.

onToolCall?ChatOnToolCallCallback<UI_MESSAGE>

Optional callback function that is invoked when a tool call is received. Intended for automatic client-side tool execution. To add the tool output, call `addToolOutput` without awaiting it inside this callback. The callback's return value is not used.

onData?ChatOnDataCallback<UI_MESSAGE>

Optional callback function that is called when a data part is received.

sendAutomaticallyWhen?(options: { messages: UI_MESSAGE[]; }) => boolean | PromiseLike<boolean>

When provided, this function will be called when the stream is finished or a tool call is added to determine if the current messages should be resubmitted.

throttle?number | undefined

toCreateMessage?CustomToCreateMessageFunction

onResumeError?((error: unknown) => void) | undefined

Called when an automatic resumable stream reconnect fails. Use this to surface a toast, report telemetry, or mark the thread as needing a retry. The failed stream id is cleared after the callback unless a newer id has replaced it.

transport?ChatTransport<UI_MESSAGE> | undefined

The transport threads send through. `AssistantChatTransport` instances are cloned per thread through `__internal_clone()` so their assistant-ui wiring remains isolated. Other transport instances are shared as-is.

UseChatRuntimeOptions["transport"]
sendMessages(options: { /** The type of message submission - either new message or regeneration */ trigger: 'submit-message' | 'regenerate-message'; /** Unique identifier for the chat session */ chatId: string; /** ID of the message to regenerate, or undefined for new messages */ messageId: string | undefined; /** Array of UI messages representing the conversation history */ messages: UI_MESSAGE[]; /** Signal to abort the request if needed */ abortSignal: AbortSignal | undefined; } & ChatRequestOptions) => Promise<ReadableStream<UIMessageChunk>>

Sends messages to the chat API endpoint and returns a streaming response. This method handles both new message submission and message regeneration. It supports real-time streaming of responses through UIMessageChunk events.

reconnectToStream(options: { /** Unique identifier for the chat session to reconnect to */ chatId: string; /** Signal to abort the reconnection request if needed */ abortSignal?: AbortSignal; } & ChatRequestOptions) => Promise<ReadableStream<UIMessageChunk> | null>

Reconnects to an existing streaming response for the specified chat session. This method is used to resume streaming when a connection is interrupted or when resuming a chat session. It's particularly useful for maintaining continuity in long-running conversations or recovering from network issues.

cloud?AssistantCloud | undefined

UseChatRuntimeOptions["cloud"]
threadsAssistantCloudThreads

AssistantCloudThreads
messagesAssistantCloudThreadMessages

cloudAssistantCloudAPI

list(query?: AssistantCloudThreadsListQuery) => Promise<AssistantCloudThreadsListResponse>

get(threadId: string) => Promise<CloudThread>

create(body: AssistantCloudThreadsCreateBody) => Promise<AssistantCloudThreadsCreateResponse>

update(threadId: string, body: AssistantCloudThreadsUpdateBody) => Promise<void>

claim(body: AssistantCloudThreadsClaimBody) => Promise<AssistantCloudThreadsClaimResponse>

Moves every thread of the anonymous identity behind `refresh_token` into the caller's workspace.

delete(threadId: string) => Promise<void>

projectsAssistantCloudProjects

AssistantCloudProjects
threadsAssistantCloudProjectThreads

auth__object

__object
tokensAssistantCloudAuthTokens

invalidate() => void

runsAssistantCloudRuns

AssistantCloudRuns
cloudAssistantCloudAPI

stream(body: AssistantCloudRunsStreamBody) => Promise<AssistantStream>

report(body: AssistantCloudRunReport) => Promise<{ run_id: string; }>

filesAssistantCloudFiles

AssistantCloudFiles
cloudAssistantCloudAPI

pdfToImages(body: PdfToImagesRequestBody) => Promise<PdfToImagesResponse>deprecated

Deprecated: Assistant Cloud has no PDF conversion endpoint, so this request always rejects with a `CloudAPIError` whose `status` is 404.

generatePresignedUploadUrl(body: GeneratePresignedUploadUrlRequestBody) => Promise<GeneratePresignedUploadUrlResponse>

generatePresignedDownloadUrl(body: { key: string; } | { url: string; }) => Promise<GeneratePresignedDownloadUrlResponse>

eventsAssistantCloudEvents

AssistantCloudEvents
bufferAssistantCloudEvent[]

timer?ReturnType<typeof setTimeout> | undefined

flushing?Promise<void> | undefined

retryTimer?ReturnType<typeof setTimeout> | undefined

resolveRetryDelay?(() => void) | undefined

bestEffortRequestedboolean

generationnumber

cloudAssistantCloudAPI

isEnabled() => boolean

listeningboolean

track(event: AssistantCloudEvent) => void

listen() => void

unlisten() => void

dispose() => void

clearPending() => void

onVisibilityChange() => void

flushBestEffort() => Promise<void>

flush(retryFailures: boolean) => Promise<void>

flushPending(retryFailures: boolean) => Promise<void>

waitForRetry(delay: number) => Promise<void>

interruptRetryDelay() => void

scheduleFlush() => void

clearFlushTimer() => void

scoresAssistantCloudScores

AssistantCloudScores
cloudAssistantCloudAPI

create(body: AssistantCloudScoreBody) => Promise<AssistantCloudScoreResponse>

telemetryAssistantCloudTelemetryConfig

AssistantCloudTelemetryConfig
enabled?boolean

Enables Assistant Cloud telemetry. Defaults to `true`. Set to `false` to disable both run reports and engagement events.

events?boolean

Enables Assistant Cloud engagement events. Defaults to `true` when telemetry is enabled. Set to `false` to keep run reports while disabling engagement events.

messages?boolean

Stores the messages of runtimes whose backend keeps the transcript (LangGraph, LangChain, Google ADK, custom external stores), so the dashboard can show them. Defaults to `true` when telemetry is enabled. Set to `false` to keep run reports and events without storing those messages.

release?string

environment?string

tags?string[]

beforeReport?( report: AssistantCloudRunReport, ) => AssistantCloudRunReport | null

Called before each telemetry report is sent. Return a modified report to enrich it (e.g. add `model_id`), or return `null` to skip the report.

registerSdk(sdk: SdkIdentity) => void

scopeId?string | undefined

Stable identity for the account or workspace owning Cloud runtime state. Provide it from the first render and change it when that scope changes.

onThreadIdChange?((threadId: string | undefined) => void) | undefined

useThreadTokenUsage

Reads token usage from the newest assistant message that reports any.

A route attaches usage through the AI SDK's messageMetadata option. Because a thread message carries a fixed metadata shape, the converter moves every other key the route returns into metadata.custom, which is where this hook looks.

function useThreadTokenUsage(): ThreadTokenUsage | undefined;

generativeTools

Warning

Deprecated. Use AISDKToolkit instead: new AISDKToolkit({ toolkit }).tools({ frontend }). It is a strict superset (it also opens MCP server connections), so it replaces generativeTools everywhere. The frontendTools option is named frontend on .tools(), and .tools() is async. generativeTools will be removed in a future version.

Builds an AI SDK ToolSet for server-side use with streamText / generateText from a generative toolkit and the frontend-uploaded tools.

Each toolkit tool's execute runs on the server. Pair this with the "use generative" compiler: import the toolkit in a server route (where it resolves to the server build — schema + execute, with render stripped) and pass it here. Tools without an execute are still exposed to the model but left for the client to fulfill. frontendTools lets the client contribute tools that aren't in the static toolkit.

// Define once at module scope so any MCP connections pool across requests.
const aiToolkit = new AISDKToolkit({ toolkit: docsToolkit });

// In your route handler:
const { tools } = await req.json();
streamText({
  model,
  messages,
  tools: await aiToolkit.tools({ frontend: tools }),
});
const generativeTools: (options: GenerativeToolsOptions) => ToolSet;

On this page

View as MarkdownEdit on GitHub