feat(chat): add vision input with drag-and-drop & paste images (M9.4)

Drag-and-drop or paste images into chat input. Thumbnail preview above
input, max 4 images per message. Images encoded as base64 and sent in
Claude vision format (multimodal content arrays). Image attach button
in chat input toolbar.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

packages/core/src/types/message.ts

@@ -8,6 +8,13 @@ export interface WebSearchResult {
   imgSrc?: string
 }
 
+export interface ImageAttachment {
+  /** Base64-encoded image data (no data: prefix) */
+  data: string
+  /** MIME type e.g. image/jpeg, image/png, image/gif, image/webp */
+  mediaType: string
+}
+
 export interface Message {
   id: string
   role: 'user' | 'assistant' | 'system'
@@ -23,6 +30,8 @@ export interface Message {
   webResults?: WebSearchResult[]
   /** Timestamp of last edit (set when user edits a sent message) */
   editedAt?: number
+  /** Attached images (base64, max 4) for vision-capable models */
+  images?: ImageAttachment[]
 }
 
 export interface Reaction {
@@ -39,4 +48,10 @@ export interface Conversation {
   updatedAt: number
   model?: string
   systemPrompt?: string
+  /** ID of parent conversation this was branched from */
+  parentConversationId?: string
+  /** Message ID in parent where this branch forked */
+  branchPoint?: string
+  /** IDs of child branches forked from this conversation */
+  childBranchIds?: string[]
 }