When you search your contracts, anything sitting in an attachment is invisible — T&Cs, schedules, annexes, price lists. You get results, but not complete ones, and there's no indication that something was skipped. That makes it hard to trust a search that's meant to cover your whole repository.
The same gap affects the AI assistant. Ask it about a contract and it reads the main document only. If the substance is in an attachment, the annex with the actual pricing, the referenced terms, it can't see it, and will tell you the information isn't there. The workaround is opening each attachment and re-uploading it into the chat.
We'd like attachment and stored-file content to be extracted (via OCR where needed) and made available to both search and the assistant, so that searching across all documents actually means all documents.