Expected
When meaning is on, the translation shown for an assistant passage should cover the full caption the learner sees (or clearly disclose a partial translation).
Actual
Both clients send only a suffix of the passage to the meaning helper:
- iOS
ConversationCoordinator: String(request.text.suffix(2200))
- Android
MuralViewModel: request.text.takeLast(2200)
The result is cached under the full revisionKey and shown as the meaning of the whole passage. Long captions (many streamed deltas) lose the beginning with no UI signal.
This is separate from #8 (typed reply / correction caps).
Reproduce
- Enable meanings.
- Produce an assistant passage longer than 2,200 characters (e.g. many close transcript deltas).
- Observe the translation does not cover the start of the caption, with no truncation notice.
Suggested fix
Translate the full request.text, or move the cap into a shared visible limit (same spirit as #8 / PR #42) and do not present a suffix translation as a full-passage meaning.
Expected
When meaning is on, the translation shown for an assistant passage should cover the full caption the learner sees (or clearly disclose a partial translation).
Actual
Both clients send only a suffix of the passage to the meaning helper:
ConversationCoordinator:String(request.text.suffix(2200))MuralViewModel:request.text.takeLast(2200)The result is cached under the full
revisionKeyand shown as the meaning of the whole passage. Long captions (many streamed deltas) lose the beginning with no UI signal.This is separate from #8 (typed reply / correction caps).
Reproduce
Suggested fix
Translate the full
request.text, or move the cap into a shared visible limit (same spirit as #8 / PR #42) and do not present a suffix translation as a full-passage meaning.