Files
supabase/apps/studio/components/ui/AIAssistantPanel/AssistantChat.tsx
Saxon FletcherandClaude Opus 5.5 5e59b6047e chore(studio): extend Assistant response time and handle timeouts (#50892)
## Problem

- Assistant responses were capped at 120 seconds and 10 steps, which is
too short for longer reasoning or multi-step tool work.
- When the hosting platform ended a request at that limit, the
connection just dropped. The user got no explanation, and "Thinking…"
and tool rows kept spinning.
- Studio's own tools ignored the request's abort signal, so a stop,
disconnect or deadline couldn't cancel their in-flight requests.
- Aborted responses never closed their Braintrust span. Under TanStack
Start, the remote MCP client was only released on `res.on('close')`,
which the adapter never emits.

## Solution

Uses AI SDK options instead of custom stream handling:

- `maxDuration` goes to 300s and the step limit to 20. `streamText({
timeout: { totalMs } })` stops the response at 270s, leaving time to
finish the stream before the platform cutoff.
- `toUIMessageStream({ messageMetadata })` marks an aborted response
`timedOut: true`. `Chat` ignores `abort` chunks, so the client reads
this flag instead and shows a timeout alert with Retry. The flag is
saved with the message, so the alert survives a reload.
- `toUIMessageStream({ onEnd })` aborts the request whenever the stream
ends, releasing the MCP client on both runtimes. `streamText({ onAbort
})` ends the Braintrust span.
- Studio tools pass the SDK's `abortSignal` to their fetches. MCP tools
already did.
- Reasoning and server-tool rows that never finished show "Response
interrupted" instead of a spinner or "Ran X ✓".

There's no per-tool timeout. Approved SQL and migrations can
legitimately run longer, and aborting the HTTP request doesn't stop the
query in Postgres.

## Review instructions

1. Run the unit tests: `cd apps/studio && pnpm vitest run
lib/api/generate-v4.test.ts lib/ai components/ui/AIAssistantPanel`
2. To see a timeout without waiting 4.5 minutes, temporarily set
`ASSISTANT_TIMEOUT_MS` in `apps/studio/lib/ai/assistant-timeout.ts` to
`15_000` and run `pnpm dev:studio`.
3. Ask the Assistant something that needs several tool calls or long
reasoning, for example "Audit my schema for missing indexes and RLS
gaps, then write the fixes."
4. After 15 seconds, check that:
- the response stops and a "Assistant response timed out" alert appears
with Retry
- any in-progress reasoning or tool row shows "Response interrupted"
instead of spinning
   - Retry starts a new response
   - reloading the page still shows the alert on that chat
5. Stop a response with the Stop button before the deadline. It should
stop without the timeout alert.
6. With the default 270s, confirm that a normal response completes as
before.

## Checklist

Check all before review:

- [ ] I have read
[CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md)
- [ ] If I wrote a new docs topic or edited an existing topic, I used
the `/write-the-docs` or `/edit-the-docs` skill, which references
[WORD_LIST](https://github.com/supabase/supabase/blob/master/apps/docs/WORD_LIST.md)
and the docs
[CONTRIBUTING](https://github.com/supabase/supabase/blob/master/apps/docs/CONTRIBUTING.md)
guide


<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **Improvements**
* AI assistant responses can now run for up to five minutes, supporting
longer requests.
* When a response times out, the assistant displays a message suggesting
you retry or ask for a smaller change.
* Incomplete responses now show a “Response interrupted” notice, and
loading indicators stop when generation ends.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-28 12:44:39 +10:00

709 lines
25 KiB
TypeScript

import type { UIMessage as MessageType } from '@ai-sdk/react'
import { useChat } from '@ai-sdk/react'
import { lastAssistantMessageIsCompleteWithApprovalResponses } from 'ai'
import { LOCAL_STORAGE_KEYS, useFlag } from 'common'
import { useParams, useSearchParamsShallow } from 'common/hooks'
import { AnimatePresence, motion, useReducedMotion } from 'framer-motion'
import { Eraser, Pencil, X } from 'lucide-react'
import { Fragment, useCallback, useEffect, useMemo, useRef, useState, type ReactNode } from 'react'
import { Button, cn, KeyboardShortcut } from 'ui'
import { Admonition } from 'ui-patterns/Admonition'
import { AlertError } from '../AlertError'
import { ButtonTooltip } from '../ButtonTooltip'
import { ErrorBoundary } from '../ErrorBoundary/ErrorBoundary'
import { InlineLinkClassName } from '../InlineLink'
import { ASSISTANT_ERRORS } from './AiAssistant.constants'
import {
containsLogsSnippets,
hasPendingToolApproval,
onErrorChat,
resolvePendingToolApprovalsAsDenied,
} from './AIAssistant.utils'
import { AIOnboarding } from './AIOnboarding'
import { AssistantAgentHarnessFooter } from './AssistantAgentHarnessFooter'
import { AssistantChatForm } from './AssistantChatForm'
import {
Conversation,
ConversationContent,
ConversationScrollButton,
} from './elements/Conversation'
import { Message } from './Message'
import { Markdown } from '@/components/interfaces/Markdown'
import { useCheckOpenAIKeyQuery } from '@/data/ai/check-api-key-query'
import { useRateMessageMutation } from '@/data/ai/rate-message-mutation'
import { useTablesQuery } from '@/data/tables/tables-query'
import { useLatest } from '@/hooks/misc/useLatest'
import { useLocalStorageQuery } from '@/hooks/misc/useLocalStorage'
import { useOrgAiOptInLevel } from '@/hooks/misc/useOrgOptedIntoAi'
import { useSelectedOrganizationQuery } from '@/hooks/misc/useSelectedOrganization'
import { useSelectedProjectQuery } from '@/hooks/misc/useSelectedProject'
import {
isTimedOutMessage,
type AssistantMessageMetadata,
} from '@/lib/ai/assistant-message-metadata'
import { ASSISTANT_TIMEOUT_MESSAGE } from '@/lib/ai/assistant-timeout'
import { getParallelApprovalIdsToReject } from '@/lib/ai/message-utils'
import { IS_PLATFORM } from '@/lib/constants'
import { uuidv4 } from '@/lib/helpers'
import { useTrack } from '@/lib/telemetry/track'
import type { SqlSnippet } from '@/state/ai-assistant-state'
import { useAiAssistantState, useAiAssistantStateSnapshot } from '@/state/ai-assistant-state'
import { SHORTCUT_IDS } from '@/state/shortcuts/registry'
import { useShortcut } from '@/state/shortcuts/useShortcut'
export interface AssistantChatHeaderProps {
isChatLoading: boolean
showMetadataWarning: boolean
updatedOptInSinceMCP: boolean
aiOptInLevel: 'disabled' | 'schema' | 'full' | string | undefined
}
export interface AssistantChatComposerContext {
initialInput?: string
sqlSnippets?: SqlSnippet[]
suggestions?: { title?: string; prompts?: { label: string; description: string }[] }
onSetSqlSnippets?: (snippets: SqlSnippet[]) => void
onClearSqlSnippets?: () => void
}
interface AssistantChatProps {
className?: string
chatId: string
shortcutsEnabled?: boolean
onNewChat: () => void
onSelectChat: (chatId: string) => void
onBranchChat: (messageId: string) => void
composerContext?: AssistantChatComposerContext
renderHeader?: (props: AssistantChatHeaderProps) => ReactNode
onInputChange?: (value: string) => void
}
export const AssistantChat = ({
className,
chatId,
shortcutsEnabled = true,
onNewChat,
onSelectChat,
onBranchChat,
composerContext,
renderHeader,
onInputChange,
}: AssistantChatProps) => {
const { id: entityId } = useParams()
const { data: project } = useSelectedProjectQuery()
const searchParams = useSearchParamsShallow()
const { data: selectedOrganization, isPending: isLoadingOrganization } =
useSelectedOrganizationQuery()
const disablePrompts = useFlag('disableAssistantPrompts')
const shouldReduceMotion = useReducedMotion()
const snap = useAiAssistantStateSnapshot()
const state = useAiAssistantState()
const currentChat = snap.chats[chatId]
useShortcut(SHORTCUT_IDS.AI_ASSISTANT_CANCEL_EDIT, () => cancelEdit(), {
enabled: shortcutsEnabled,
})
const [updatedOptInSinceMCP] = useLocalStorageQuery(
LOCAL_STORAGE_KEYS.AI_ASSISTANT_MCP_OPT_IN,
false
)
const inputRef = useRef<HTMLTextAreaElement>(null)
const { aiOptInLevel } = useOrgAiOptInLevel()
// Whether attached queries are sent at all. One definition, shared by the chat form
// (which folds them into the message text) and the message metadata (which states
// whether any of them was a logs query), so the two can't disagree.
const includeSnippetsInMessage = aiOptInLevel !== 'disabled'
const showMetadataWarning =
IS_PLATFORM &&
!!selectedOrganization &&
(aiOptInLevel === 'disabled' || aiOptInLevel === 'schema')
// Add a ref to store the last user message
const lastUserMessageRef = useRef<MessageType | null>(null)
const [value, setValue] = useState<string>(composerContext?.initialInput || '')
const [editingMessageId, setEditingMessageId] = useState<string | null>(null)
const [isResubmitting, setIsResubmitting] = useState(false)
const [messageRatings, setMessageRatings] = useState<Record<string, 'positive' | 'negative'>>({})
const { data: check, isSuccess } = useCheckOpenAIKeyQuery()
const isApiKeySet = !!check?.hasKey
const { mutateAsync: rateMessage } = useRateMessageMutation()
const { data: tables } = useTablesQuery(
{
projectRef: project?.ref,
connectionString: project?.connectionString,
schema: 'public',
},
{ enabled: isApiKeySet }
)
const currentTable = tables?.find((t) => t.id.toString() === entityId)
const currentSchema = searchParams?.get('schema') ?? 'public'
// Update context in state
useEffect(() => {
state.setContext({
projectRef: project?.ref,
orgSlug: selectedOrganization?.slug,
connectionString: project?.connectionString ?? '',
})
}, [project?.ref, project?.connectionString, selectedOrganization?.slug, state])
const track = useTrack()
useEffect(() => {
state.ensureChatInstance(chatId)
}, [chatId, state])
const chatInstance = snap.chatInstances[chatId]
const {
messages: chatMessages,
status: chatStatus,
error,
sendMessage,
setMessages,
addToolApprovalResponse,
stop,
regenerate,
} = useChat<MessageType>({
id: chatId,
// Batch token updates without throttling the SDK's tool execution or approval state.
throttle: 50,
...(chatInstance ? { chat: chatInstance } : {}),
sendAutomaticallyWhen: lastAssistantMessageIsCompleteWithApprovalResponses,
onError: onErrorChat,
})
const isChatLoading = chatStatus === 'submitted' || chatStatus === 'streaming'
const hasPendingApproval = hasPendingToolApproval(chatMessages)
const supportMetadata = currentChat?.supportMetadata
const isSupportChat = !!supportMetadata?.isSupportChat
const isSupportChatClosed = isSupportChat && supportMetadata.lifecycleStatus !== 'bot_active'
const supportConversationId = supportMetadata?.frontConversationId
const isChatInputDisabled =
!isApiKeySet || disablePrompts || isLoadingOrganization || isSupportChatClosed
const messagesRef = useLatest(chatMessages)
const isChatLoadingRef = useLatest(isChatLoading)
const branchedFrom = currentChat?.branchedFrom
const branchedConversation = branchedFrom ? snap.chats[branchedFrom.chatId] : undefined
const deleteMessageFromHere = useCallback(
(messageId: string) => {
const messages = messagesRef.current
const messageIndex = messages.findIndex((msg) => msg.id === messageId)
if (messageIndex === -1) return
if (isChatLoadingRef.current) stop()
state.deleteMessagesAfter(messageId, { includeSelf: true, chatId })
const updatedMessages = messages.slice(0, messageIndex)
setMessages(updatedMessages)
},
[state, setMessages, messagesRef, isChatLoadingRef, stop, chatId]
)
const editMessage = useCallback(
(messageId: string) => {
const messages = messagesRef.current
const messageIndex = messages.findIndex((msg) => msg.id === messageId)
if (messageIndex === -1) return
// Target message
const messageToEdit = messages[messageIndex]
// Activate editing mode
setEditingMessageId(messageId)
const textContent =
messageToEdit.parts
?.filter((part) => part.type === 'text')
.map((part) => part.text)
.join('') ?? ''
setValue(textContent)
setTimeout(() => {
if (inputRef.current) {
inputRef?.current?.focus()
// [Joshen] This is just to make the cursor go to the end of the text when focusing
const val = inputRef.current.value
inputRef.current.value = ''
inputRef.current.value = val
}
}, 100)
},
[messagesRef, setValue]
)
const cancelEdit = useCallback(() => {
setEditingMessageId(null)
setValue('')
}, [setValue])
const handleRateMessage = useCallback(
async (messageId: string, rating: 'positive' | 'negative', reason?: string) => {
if (!project?.ref || !selectedOrganization?.slug) return
// Optimistically update UI
setMessageRatings((prev) => ({ ...prev, [messageId]: rating }))
try {
const result = await rateMessage({
rating,
messages: messagesRef.current,
messageId,
projectRef: project.ref,
orgSlug: selectedOrganization.slug,
reason,
spanId: state.messageSpanIds[messageId],
})
track('assistant_message_rating_submitted', {
rating,
category: result.category,
...(reason && { reason }),
chatId,
})
} catch (error) {
console.error('Failed to rate message:', error)
// Rollback on error
setMessageRatings((prev) => {
const { [messageId]: _, ...rest } = prev
return rest
})
}
},
[messagesRef, project?.ref, selectedOrganization?.slug, rateMessage, track, state, chatId]
)
const isContextExceededError =
error &&
(error.message?.includes('context_length_exceeded') ||
error.message?.includes('exceeds the context window'))
const isTimedOut = !error && !isChatLoading && isTimedOutMessage(chatMessages.at(-1))
let displayError = IS_PLATFORM ? ASSISTANT_ERRORS['default'] : error
if (isContextExceededError) displayError = ASSISTANT_ERRORS['context-exceeded']
if (isTimedOut) displayError = { message: ASSISTANT_TIMEOUT_MESSAGE }
const editedMessageIndex = editingMessageId
? chatMessages.findIndex((message) => message.id === editingMessageId)
: -1
const renderedMessages = useMemo(
() =>
chatMessages.map((message, index) => {
const isBeingEdited = editingMessageId === message.id
const isAfterEditedMessage = !!editingMessageId && editedMessageIndex < index
const isLastMessage = index === chatMessages.length - 1
return (
<Fragment key={message.id}>
<Message
id={message.id}
message={message}
isLoading={chatStatus === 'submitted' || chatStatus === 'streaming'}
readOnly={message.role === 'user'}
addToolApprovalResponse={addToolApprovalResponse}
onDelete={deleteMessageFromHere}
onEdit={editMessage}
isAfterEditedMessage={isAfterEditedMessage}
isBeingEdited={isBeingEdited}
onCancelEdit={cancelEdit}
isLastMessage={isLastMessage}
onRate={handleRateMessage}
rating={messageRatings[message.id] ?? null}
onBranch={onBranchChat}
/>
{branchedConversation && branchedFrom?.messageId === message.id && (
<div className="flex items-center gap-2 mt-6">
<div className="flex-1 border-t border-strong" />
<div className="flex items-center gap-1 max-w-[80%] text-xs text-foreground-lighter">
<span className="shrink-0">Branched from</span>
<button
tabIndex={0}
className={cn(InlineLinkClassName, 'cursor-pointer truncate min-w-0')}
onClick={() => onSelectChat(branchedConversation.id)}
>
{branchedConversation.name}
</button>
</div>
<div className="flex-1 border-t border-strong" />
</div>
)}
</Fragment>
)
}),
[
chatMessages,
deleteMessageFromHere,
editMessage,
cancelEdit,
editingMessageId,
editedMessageIndex,
chatStatus,
addToolApprovalResponse,
handleRateMessage,
messageRatings,
branchedConversation,
branchedFrom,
onSelectChat,
onBranchChat,
]
)
const hasMessages = chatMessages.length > 0
const sendMessageToAssistant = (finalContent: string) => {
if (editingMessageId) {
// Handling when the user is in edit mode
// delete the message(s) from the chat just like the delete button
setIsResubmitting(true)
deleteMessageFromHere(editingMessageId)
setEditingMessageId(null)
}
// Read off the attachments this message actually carries, so detaching the
// "Current Query" chip also drops the claim. Gated on the same condition that
// decides whether attachments make it into the text at all: with AI opt-in
// disabled the chip is shown but no query is sent, and claiming otherwise would
// have the server prepend ClickHouse context for a message holding no query.
// Rides on the message rather than the request, so a Retry reproduces the context
// the message was asked in.
const metadata: AssistantMessageMetadata = {
containsLogsSnippets:
includeSnippetsInMessage && containsLogsSnippets(composerContext?.sqlSnippets),
}
const payload = {
role: 'user',
createdAt: new Date(),
parts: [{ type: 'text', text: finalContent }],
id: uuidv4(),
metadata,
} as MessageType
composerContext?.onClearSqlSnippets?.()
lastUserMessageRef.current = payload
if (hasPendingApproval && !editingMessageId) {
setMessages(resolvePendingToolApprovalsAsDenied(chatMessages))
}
sendMessage(payload, {
body: {
schema: currentSchema,
table: currentTable?.name,
},
})
setValue('')
if (finalContent.includes('Help me to debug')) {
track('assistant_debug_submitted', { chatId })
} else {
track('assistant_prompt_submitted', { chatId })
}
}
const handleClearMessages = () => {
if (isChatLoading) stop()
snap.clearMessages(chatId)
setMessages([])
lastUserMessageRef.current = null
setEditingMessageId(null)
}
useEffect(() => {
// Keep "Thinking" visible while stopping and resubmitting during edit
// Only clear once the new response actually starts streaming (or errors)
if (isResubmitting && (chatStatus === 'streaming' || !!error)) {
setIsResubmitting(false)
}
}, [isResubmitting, chatStatus, error])
useEffect(() => {
// Approval-required tools can't run in parallel. Auto-deny extras so the model reissues them sequentially.
for (const id of getParallelApprovalIdsToReject(chatMessages)) {
addToolApprovalResponse?.({
id,
approved: false,
reason:
'Only one approval-required tool call is allowed per turn. Please reissue this tool call after the current one completes.',
})
}
}, [chatMessages, addToolApprovalResponse])
useEffect(() => {
setValue(composerContext?.initialInput || '')
if (inputRef.current && composerContext?.initialInput) {
inputRef.current.focus()
inputRef.current.setSelectionRange(
composerContext.initialInput.length,
composerContext.initialInput.length
)
}
}, [composerContext?.initialInput])
let placeholder: string
if (hasMessages && isSupportChat) {
placeholder = 'Share details so the assistant can help with your support request...'
} else if (hasMessages) {
placeholder = 'Ask a follow up question...'
} else if ((composerContext?.sqlSnippets ?? []).length > 0) {
placeholder = 'Ask a question or make a change...'
} else if (isSupportChat) {
placeholder = 'Describe your support issue...'
} else {
placeholder = 'Ask about your data, troubleshoot an issue, or explore your project...'
}
const composer = (
<div className={cn('relative z-20 w-full', hasMessages && 'px-7 pb-3')}>
<div className="mx-auto flex w-full max-w-3xl flex-col gap-y-3">
{isSupportChat && !isSupportChatClosed && (
<div>
<div className="mb-3 border-t" />
<div className="flex items-center gap-2">
<Button
variant="outline"
size="tiny"
disabled={!supportConversationId}
onClick={() => state.setSupportLifecycleStatus(chatId, 'escalated')}
>
Escalate to human
</Button>
<Button
variant="outline"
size="tiny"
disabled={!supportConversationId}
onClick={() => state.setSupportLifecycleStatus(chatId, 'user_resolved')}
>
Resolve
</Button>
</div>
</div>
)}
{disablePrompts && (
<Admonition
showIcon={false}
type="default"
title="Assistant has been temporarily disabled"
description="We're currently looking into getting it back online"
/>
)}
{isSuccess && !isApiKeySet && (
<Admonition
type="default"
title="OpenAI API key not set"
description={
<Markdown
content={
'Add your `OPENAI_API_KEY` to your environment variables to use the AI Assistant.'
}
/>
}
/>
)}
<div>
<AssistantChatForm
textAreaRef={inputRef}
className={cn('z-20', !hasMessages && 'bg')}
loading={isChatLoading}
isEditing={!!editingMessageId}
disabled={isChatInputDisabled}
placeholder={placeholder}
value={value}
onValueChange={(e) => {
setValue(e.target.value)
onInputChange?.(e.target.value)
}}
onSubmit={(finalMessage) => {
sendMessageToAssistant(finalMessage)
}}
onStop={() => {
stop()
// to save partial responses from the AI
// Read the live SDK state: the rendered snapshot may trail the stream by 50ms.
const messages = chatInstance?.messages ?? chatMessages
const lastMessage = messages[messages.length - 1]
if (lastMessage && lastMessage.role === 'assistant') {
state.updateMessage(lastMessage, chatId)
}
}}
sqlSnippets={composerContext?.sqlSnippets}
onRemoveSnippet={(index) => {
const newSnippets = [...(composerContext?.sqlSnippets ?? [])]
newSnippets.splice(index, 1)
composerContext?.onSetSqlSnippets?.(newSnippets)
}}
includeSnippetsInMessage={includeSnippetsInMessage}
/>
</div>
</div>
</div>
)
return (
<ErrorBoundary
message="Something went wrong with the AI Assistant"
sentryContext={{
component: 'AIAssistant',
feature: 'AI Assistant Panel',
projectRef: project?.ref,
organizationSlug: selectedOrganization?.slug,
}}
actions={[
{
label: 'Clear messages and refresh',
onClick: () => {
handleClearMessages()
window.location.reload()
},
},
]}
>
<div className={cn('flex bg-card flex-col h-full w-full md:h-full max-h-dvh', className)}>
{renderHeader?.({
isChatLoading,
showMetadataWarning,
updatedOptInSinceMCP,
aiOptInLevel,
})}
{hasMessages ? (
<Conversation className={cn('flex-1')}>
<ConversationContent className="w-full py-8 mb-10">
{renderedMessages}
<div className="w-full max-w-3xl mx-auto">
{(error || isTimedOut) && (
<AlertError
error={displayError}
showErrorPrefix={false}
showInstructions={false}
subject={
isTimedOut
? 'Assistant response timed out'
: "Sorry, I'm having trouble responding right now."
}
additionalActions={
<div className="flex items-center gap-x-2 mr-auto">
{isContextExceededError ? (
<Button size="tiny" onClick={onNewChat} className="text-xs">
New chat
</Button>
) : (
<>
<Button size="tiny" onClick={() => regenerate()} className="text-xs">
Retry
</Button>
<ButtonTooltip
size="tiny"
onClick={handleClearMessages}
className="w-7 h-7"
icon={<Eraser />}
tooltip={{ content: { side: 'bottom', text: 'Clear messages' } }}
/>
</>
)}
</div>
}
/>
)}
{isChatLoading && (
<motion.span
animate={shouldReduceMotion ? { opacity: 1 } : { opacity: [1, 0] }}
transition={
shouldReduceMotion
? { duration: 0 }
: { duration: 1, repeat: Infinity, ease: 'linear' }
}
className="inline-block w-1.5 h-4 bg-foreground-lighter mt-4"
/>
)}
<p className="text-center text-xs text-foreground-muted mt-6">
The Assistant can make mistakes. Double check responses.
</p>
</div>
</ConversationContent>
<ConversationScrollButton />
</Conversation>
) : null}
<AnimatePresence>
{editingMessageId && (
<motion.div
initial={{ opacity: 0 }}
animate={{ opacity: 1 }}
exit={{ opacity: 0 }}
className="pointer-events-none z-10 -mt-24"
>
<div className="h-24 w-full bg-linear-to-t from-background to-transparent relative">
<motion.div
className="absolute left-1/2 z-20 bottom-8 pointer-events-auto"
variants={{
hidden: { y: 5, opacity: 0 },
show: { y: 0, opacity: 1 },
}}
transition={{ duration: 0.1 }}
initial="hidden"
animate="show"
exit="hidden"
>
<div className="-translate-x-1/2 bg-alternative dark:bg-muted border rounded-md px-3 py-2 min-w-[180px] flex items-center justify-between gap-x-2">
<div className="flex items-center gap-x-2 text-sm text-foreground">
<Pencil size={14} />
<span>Editing message</span>
</div>
<ButtonTooltip
variant="outline"
size="tiny"
icon={<X size={14} />}
onClick={cancelEdit}
className="w-6 h-6 p-0"
title="Cancel editing"
aria-label="Cancel editing"
tooltip={{
content: { side: 'top', text: <KeyboardShortcut keys={['Meta', 'Esc']} /> },
}}
/>
</div>
</motion.div>
</div>
</motion.div>
)}
</AnimatePresence>
{hasMessages ? (
composer
) : (
<AIOnboarding
key={chatId}
sqlSnippets={composerContext?.sqlSnippets}
suggestions={composerContext?.suggestions}
onValueChange={(prompt) => {
setValue(prompt)
onInputChange?.(prompt)
}}
onFocusInput={() => inputRef.current?.focus()}
>
{composer}
{!isSupportChat && <AssistantAgentHarnessFooter />}
</AIOnboarding>
)}
</div>
</ErrorBoundary>
)
}