fix: a11y continuous improvements

[isekovanic]

🎯 Goal

Cross platform a11y fixes across the message surface (message row, footer, reactions, replies) and the channel preview list. VoiceOver / TalkBack now read each surface with proper labels, no misleading affordances, and correct focus behavior.

🛠 Implementation details

• Poll message rotor actions - iOS announced custom rotor actions that couldn't actually be activated. Removed the composite rotor; poll bubbles now drill in so each option is a focus stop with proper radio/checkbox role.
• Reply composer focus - initiating a reply landed VoiceOver focus on the navigation back button. Reply now focuses the composer's text input when a screen reader is active, and announces "Replying to {user}" + "Editing message" via useAnnounceOnShow so the user knows the composer entered reply/edit mode.
• Channel preview verbosity - list rows previously announced "Avatar of X" per member, bare timestamps ("12:14 AM"), bare unread counts ("3"). Each leaf now carries a curated label: avatar reads "Direct chat with {name}" / "Channel with N members", date reads "Last message {date}", unread badge reads "{count} unread messages", muted indicator reads "Muted", delivery-status icon reads "Read, sent by you" / "Delivered, sent by you" / etc.
• Channel preview delivery status not read - the status View had accessibilityLabel but no accessible={true}, so VO walked past it. Fixed by adding accessible so the composed delivery-status label is
  announced.
• Message footer delivery state silent - the checkmark icon was a silent SVG; VO read only "11:05 AM". Footer is now accessible + accessibilityRole='text' (cross-platform compose), with MessageStatus
  rendering a HiddenA11yText for the localized status string. Footer announces as one stop: "Read, 11:05 AM".
• Reaction list "actions available" misleading affordance — VO promised rotor actions that didn't exist. Clustered reactions are now one labeled button via the Pressable's auto compose (emoji names + "you reacted" tag for own + "and N more reactions" suffix), with hint "Double tap to view reactions". Segmented reactions stay as per-item Pressables with the same "you reacted" tagging.
• Message rich content can't be activated - message rows with images, files, quoted replies, or shared location couldn't drill into the inner content. Generalized the poll only accessible={false} to cover any interactive children (attachments, quoted_message, poll, shared_location). Plain text messages keep their single focus stop behavior; rich messages drill in.

New shared components

• CompositeAccessibilityProbe (package/src/components/Accessibility/) - wraps non Text visual content (avatars, icons, badges) so it announces as one focus stop with a curated label. Internally renders a hidden Text sibling + a decorative marked View around children. Used by ChannelAvatar, ChannelPreviewMutedStatus, ChannelMessagePreviewDeliveryStatus. Solves the iOS vs Android composition asymmetry where Android's importantForAccessibility='no-hide-descendants' leaks through nested accessible={true} descendants.
• HiddenA11yText (package/src/components/Accessibility/) - visually invisible Text whose only job is to splice extra info into a parent's compose loop (e.g. "you reacted", "and N more reactions", localized delivery status). Different concern from the probe - this is for adding text to a composition chain, not for collapsing a subtree into one stop.

Both are exported from the SDK so consumers building their own custom message / preview UIs can reuse the patterns.

🎨 UI Changes

iOS

Before	After

Android

Before	After

🧪 Testing

☑️ Checklist

• I have signed the Stream CLA (required)
• PR targets the develop branch
• Documentation is updated
• New code is tested in main example apps, including all possible scenarios
  • SampleApp iOS and Android
  • Expo iOS and Android

[Stream-SDK-Bot]

SDK Size

title	develop	branch	diff	status
js_bundle_size	1725 KB	1727 KB	+2373 B	🔴