`QuotedTweetSnapshot` drops most fields ft already extracts for outer tweets

[RealADemin]

QuotedTweetSnapshot was introduced in 6ef3f42 (April 2026) as the minimum-viable shape needed to render quoted tweets — handle, name, text, media, post date. PR #134 (merged 2026-05-02) refined preservation across syncs but didn't expand the captured fields. The shape hasn't been revisited despite ft's parser already extracting much richer data for the outer-tweet BookmarkRecord from the same response.

The quoted_status_result payload contains the same legacy.* and core.user_results.result.* paths that convertTweetToRecord parses for outer tweets at src/graphql-bookmarks.ts:255-376. Currently those paths are read for outer and silently dropped when constructing the quoted snapshot at lines 295-320. This issue proposes mirroring the extraction.

The asymmetry

ft extracts the following from the outer tweet's response. None are propagated to QuotedTweetSnapshot:

Outer extracts (BookmarkRecord)	On QuotedTweetSnapshot?
engagement: { likeCount, repostCount, replyCount, quoteCount, bookmarkCount, viewCount }	✗
conversationId	✗
inReplyToStatusId, inReplyToUserId	✗
quotedStatusId (chain detection)	✗
language	✗
sourceApp	✗
possiblySensitive	✗
links: string[] (from legacy.entities.urls) + t.co→display_url expansion of text	✗
Full BookmarkAuthorSnapshot (incl. verification fields once #145 ships)	✗ — has flat authorHandle/authorName/authorProfileImageUrl only

All of these come from legacy.* and core.user_results.result.* paths that exist on quoted_status_result.result.* at the same nesting depth — verified May 2026 against live captures.

Concrete failing cases

Verified against live X API:

1. Article filter cannot detect quote-articles. Bookmark 2049874687069426008 is a quote-tweet whose quoted tweet's text is a single https://t.co/JlZbH7mCPt, resolving via 301 to x.com/i/article/2049760065427574784 (a real X Article). Downstream article-detection relies on links arrays; QuotedTweetSnapshot has none, so the filter is blind to article presence inside quotes.

2. Thread badges absent on quoted tweets. conversationId is the disambiguator for "is this part of a thread?" Outer tweets get it; quoted tweets don't. Quoted tweets that ARE in threads stay un-badged.

3. View counts dropped on quoted tweets. Once view_count always NULL — view_counts_everywhere_api_enabled flag missing from GRAPHQL_FEATURES #144 lands and views.count populates for outer, quoted tweets stay at zero — QuotedTweetSnapshot has no engagement field to receive it.

4. Auto-prepended @-mentions misattributed. Quoted text like "@user1 @user2 hello world" — without display_text_range, downstream consumers can't tell which leading @-handles were typed by the quoted author vs. auto-prepended by Twitter as reply targets. The same gap exists on outer BookmarkRecord currently — this issue subsumes the outer case as well, since the parsing is identical.

Suggested fix

Layer 1 — type + parser (≈25 lines, no schema work)

Extend QuotedTweetSnapshot at src/types.ts:43-53 to mirror BookmarkRecord:

export interface QuotedTweetSnapshot {
  id: string;
  text: string;
  url: string;
  postedAt?: string | null;
  media?: string[];
  mediaObjects?: BookmarkMediaObject[];

  // existing flat author fields can stay (additive, non-breaking)
  authorHandle?: string;
  authorName?: string;
  authorProfileImageUrl?: string;

  // NEW — parity with BookmarkRecord
  conversationId?: string;
  inReplyToStatusId?: string;
  inReplyToUserId?: string;
  quotedStatusId?: string;
  language?: string;
  sourceApp?: string;
  possiblySensitive?: boolean;
  displayTextRange?: [number, number];
  links?: string[];
  engagement?: BookmarkEngagementSnapshot;
  author?: BookmarkAuthorSnapshot;             // additive; existing flat fields stay populated
}

Parser-side: mirror lines 343-376 of the outer-tweet construction inside the quoted-tweet construction at lines 307-330. Same code, applied to qtTweet/qtLegacy. BookmarkEngagementSnapshot (line 34-41 of types.ts) and BookmarkAuthorSnapshot (extended by #145) are reusable as-is.

Adding author alongside the flat fields keeps the change additive — no breaking changes for downstream consumers reading authorHandle/authorName/authorProfileImageUrl. Both populate from the same qtTweet.core.user_results.result.

Also add displayTextRange?: [number, number] to BookmarkRecord (not just QuotedTweetSnapshot) — same field, same parsing. Single line in the outer construction at line 343 area:

displayTextRange: legacy?.display_text_range,

This subsumes any separate proposal to add display_text_range for outer tweets only.

Layer 2 — SQLite columns (optional)

QuotedTweetSnapshot is already serialized as quoted_tweet_json TEXT (added in schema v4 with 6ef3f42). Layer 1 lands automatically without schema changes — the new fields just become additional keys in the existing JSON blob.

For BookmarkRecord.displayTextRange, optionally add a display_text_range TEXT column if per-bookmark queryability is desired. Not required for the fix.

Dependencies

This issue assumes #144 and #145 land:

• view_count always NULL — view_counts_everywhere_api_enabled flag missing from GRAPHQL_FEATURES #144 adds view_counts_everywhere_api_enabled to GRAPHQL_FEATURES. Without it, X strips views from responses and engagement.viewCount stays NULL for both outer and quoted.
• isVerified boolean conflates Twitter Blue, Business, and Government accounts #145 extends BookmarkAuthorSnapshot with verification fields. Reusing that type for quoted tweets' author automatically inherits the disambiguation once isVerified boolean conflates Twitter Blue, Business, and Government accounts #145 ships.

This issue can be drafted/reviewed before either dependency lands but data quality depends on both being resolved.