feat(replay): Add AI Summary tab for replay details [INTERNAL]

[billyvg]

A new "AI Summary" tab was added to the Replay Details page for internal testing only. It is only visible to the Replay team. This will not be the final design.

• The TabKey enum in static/app/utils/replays/hooks/useActiveReplayTab.tsx was updated to include AI_SUMMARY, and it was added to supportedVideoTabs for mobile compatibility.
• In static/app/views/replays/detail/layout/focusTabs.tsx, the getReplayTabs function was modified to conditionally render the "AI Summary" tab based on the organizations:replay-ai-summaries feature flag, positioning it before the Breadcrumbs tab.
• A new AISummary component was created in static/app/views/replays/detail/aiSummary/index.tsx. This component:
  • Makes a POST request to /organizations/{organization_slug}/replays/summary/ on initial load.
  • Displays a LoadingIndicator while the request is in flight.
  • Shows an error message if the request fails.
  • Renders the summary content upon successful completion.
• The static/app/views/replays/detail/layout/focusArea.tsx file was updated to import and render the new AISummary component when the AI_SUMMARY tab is active.

Closes REPLAY-388

[billyvg]

AI Summary Tab Implementation for Replay Details

Overview

Added a new "AI Summary" tab to the Replay Details page that is visible only when the organizations:replay-ai-summaries feature flag is enabled. The tab is positioned before the Breadcrumbs tab and makes a POST request to /organizations/{organization_slug}/replays/summary/ to fetch AI-generated summaries.

Changes Made

1. Updated TabKey Enum

File: static/app/utils/replays/hooks/useActiveReplayTab.tsx

• Added AI_SUMMARY = 'ai-summary' to the TabKey enum
• Added TabKey.AI_SUMMARY to the supportedVideoTabs array to ensure it works for mobile replays

2. Updated FocusTabs Component

File: static/app/views/replays/detail/layout/focusTabs.tsx

• Added feature flag check for replay-ai-summaries in the getReplayTabs function
• The AI Summary tab appears before Breadcrumbs when the feature flag is enabled
• Added proper TypeScript typing for the Organization parameter

3. Updated FocusArea Component

File: static/app/views/replays/detail/layout/focusArea.tsx

• Added case for TabKey.AI_SUMMARY to render the new AISummary component
• Imported the new AISummary component

4. Created AISummary Component

File: static/app/views/replays/detail/aiSummary/index.tsx

• Loading State: Shows <LoadingIndicator> while the API request is in flight
• Error State: Shows error alert if the request fails
• Success State: Displays the summary text from the API response
• API Integration: Uses Sentry's useApiQuery() hook with proper query key configuration for POST requests
• Caching: Implements 5-minute stale time for efficient caching
• Conditional Fetching: Only makes requests when replay record is available
• Feature Flag: Only accessible when organizations:replay-ai-summaries is enabled
• Styling: Consistent with other replay detail tabs using styled components

API Integration Details

Query Key Structure

function createAISummaryQueryKey(orgSlug: string, replayId: string): ApiQueryKey {
  return [
    `/organizations/${orgSlug}/replays/summary/`,
    {
      method: 'POST',
      data: {
        replayId,
      },
    },
  ];
}

useApiQuery Configuration

const {data, isLoading, isError} = useApiQuery<SummaryResponse>(
  createAISummaryQueryKey(organization.slug, replayRecord?.id ?? ''),
  {
    staleTime: 5 * 60 * 1000, // 5 minutes
    enabled: Boolean(replayRecord?.id),
  }
);

Endpoint

POST /organizations/{organization_slug}/replays/summary/

Request Body

{
  "replayId": "replay-id-here"
}

Expected Response

{
  "summary": "AI-generated summary text here"
}

Benefits of useApiQuery Implementation

1. Automatic Caching: Built-in caching with 5-minute stale time prevents unnecessary requests
2. Request Deduplication: Multiple components requesting the same data will share a single request
3. Background Refetching: Automatic background updates when data becomes stale
4. Conditional Fetching: Only makes requests when replay record is available
5. Error Handling: Built-in error states and retry logic
6. Loading States: Automatic loading state management
7. TypeScript Safety: Full type safety with TypeScript generics

User Experience

1. Feature Flag Disabled: AI Summary tab is completely hidden
2. Feature Flag Enabled: AI Summary tab appears as the first tab before Breadcrumbs
3. First Load: Shows loading indicator while fetching summary
4. Success: Displays the summary text with proper formatting
5. Error: Shows user-friendly error message
6. No Data: Shows informational message when no summary is available
7. Subsequent Visits: Cached data loads instantly within 5-minute window

Testing

The implementation includes:

• Data test ID: replay-details-ai-summary-tab for automated testing
• Proper error handling for API failures
• Loading states for better UX
• Feature flag integration for gradual rollout
• Conditional fetching to prevent unnecessary requests

Dependencies

• Feature flag: organizations:replay-ai-summaries (already defined in src/sentry/features/temporary.py)
• API endpoint: /organizations/{organization_slug}/replays/summary/ (needs backend implementation)
• Components: Uses existing Sentry UI components (LoadingIndicator, Alert, etc.)
• Query Client: Uses Sentry's useApiQuery hook for data fetching and caching

[michellewzhang]

i'm getting some wonky timestamps locally and no breadcrumbs as a result

[michellewzhang]

seems like the LLM returned the timestamp in proper ms format already so we don't need to multiply by 1000?

[billyvg]

I think we have to prompt to have it return the correct timestamp because my results are coming back in seconds vs ms.

[michellewzhang]

ah yeah, that's probably a good call.

[michellewzhang]

should we prompt it to return in MS?

[michellewzhang]

extra style

[michellewzhang]

looks good! when the breadcrumbs are expanded the scroll can get janky but that's fine for now