:hedgehog: PostHog is the leading platform for building self-driving products. Our developer tools – AI observability, analytics, session replay, flags, experiments, error tracking, logs, and more – capture all the context agents need to diagnose problems, uncover opportunities, and ship fixes. Steer it all from Slack, web, desktop, or the MCP.
Python ab-testingai-analyticsanalyticscdpdata-warehouseexperimentsfeature-flagsjavascriptproduct-analyticspythonreactsession-replaysurveystypescriptweb-analytics
import { useValues } from 'kea'
import { MaxErrorTrackingSearchResponse } from '~/queries/schema/schema-assistant-error-tracking'
import { AssistantTool } from '~/queries/schema/schema-assistant-messages'
import { RecordingUniversalFilters } from '~/types'
import { ErrorTrackingFiltersWidget } from 'products/error_tracking/frontend/posthogAi/ErrorTrackingFiltersWidget'
import { RecordingsWidget } from 'products/posthog_ai/frontend/api/dataWidgets'
import { MessageTemplate } from 'products/posthog_ai/frontend/api/primitives'
import { isDangerousOperationResponse, normalizeDangerousOperationResponse } from '../approvalOperationUtils'
import { DangerousOperationApprovalCard } from '../DangerousOperationApprovalCard'
import { maxLogic } from '../maxLogic'
import { AnnotatedScreenshotView, parseAnnotatedScreenshot } from './AnnotatedScreenshotView'
export const RENDERABLE_UI_PAYLOAD_TOOLS: AssistantTool[] = [
'search_session_recordings',
'search_error_tracking_issues',
'create_form',
]
export function isRenderableUIPayloadTool(toolName: string, toolPayload: unknown): boolean {
return (
(RENDERABLE_UI_PAYLOAD_TOOLS as readonly string[]).includes(toolName) ||
isDangerousOperationResponse(toolPayload)
)
}
// Renders rich UI for a small set of tools that return structured payloads (recordings search,
// error-tracking search, create_form, upsert_dashboard) and for dangerous-operation approval cards.
export function UIPayloadAnswer({
toolCallId,
toolName,
toolPayload,
embedded = false,
}: {
toolCallId: string
toolName: string
toolPayload: any
embedded?: boolean
}): JSX.Element | null {
const { conversationId, onAcceptSessionFilters } = useValues(maxLogic)
if (toolName === 'search_session_recordings') {
const filters = toolPayload as RecordingUniversalFilters
return (
<RecordingsWidget
toolCallId={toolCallId}
filters={filters}
embedded={embedded}
onAcceptFilters={onAcceptSessionFilters}
/>
)
}
if (toolName === 'search_error_tracking_issues') {
const filters = toolPayload as MaxErrorTrackingSearchResponse
return <ErrorTrackingFiltersWidget toolCallId={toolCallId} filters={filters} embedded={embedded} />
}
if (toolName === 'summarize_website_interactions') {
const data = parseAnnotatedScreenshot((toolPayload as { screenshot?: unknown } | null)?.screenshot)
if (!data) {
return null
}
return embedded ? (
<div className="overflow-hidden rounded border bg-surface-primary p-2">
<AnnotatedScreenshotView data={data} />
</div>
) : (
<MessageTemplate type="ai" wrapperClassName="w-full">
<AnnotatedScreenshotView data={data} />
</MessageTemplate>
)
}
// Check if this is a dangerous operation requiring approval
if (isDangerousOperationResponse(toolPayload)) {
if (!conversationId) {
return null
}
const normalizedOperation = normalizeDangerousOperationResponse(toolPayload)
return <DangerousOperationApprovalCard operation={normalizedOperation} />
}
// It's not expected to hit the null branch below, because such a case SHOULD have already been filtered out
// in maxThreadLogic.selectors.threadGrouped, but better safe than sorry - there can be deployments mismatches etc.
return null
}