The Assistant works within limits on how much of a file or conversation it can read at once. Until now, only the Assistant knew when it had cut something off, so an answer could look like it covered a whole file or conversation when it didn't. Now the Assistant tells you.
Details
- Word, CSV and Excel attachments that were only partly read show a "Partly read" chip that names the limit applied, both before you send and after.
- In long conversations, the thread marks where the Assistant starts working from a summary of earlier messages and tags each later answer that used it.
- Answers for which long earlier replies were shortened are marked too.
- Conversations within the Assistant's limits now keep their earlier replies in full. Long earlier replies are shortened only when a conversation is over its limit.