First note the kernel of a discovered attack spread along the long h1-a8 diagonal. If unmasked, the Black bishop could take White’s queen. Can the rook be vacated from f3 with check? Yes; Black can play Rf1+, and White would have to reply RxR (his queen would then be pinned). Then Black could play BxQ, to which White replies KxB, and Black has won a queen for a bishop and a rook.
Don't settle for this, though; think through all of your options. How else could the rook vacate f3, and with what results? He could threaten mate with RxN, looking to play RxR# next move. RxN therefore has about the same effect as a check but picks up a knight along the way. Now what would White do? Stay aware of every piece bearing on the enemy king’s position. Here Black has a second rook he can involve, this one on the open e-file. So if White replies to RxN with RxRd3, Black plays his other rook to e1—and it’s mate. White instead needs to create a flight square for his king, and therefore replies to RxN with QxB. It avoids checkmate but loses the queen. The key thing for Black now is not to play the tempting RxQa8, but to instead use RxR+. It wins a rook, and since it's a check it effectively is a free move; White has to spend his reply saving his king, and so can’t move his queen out of harm’s way on the eighth rank. When Black has finished giving checks and doing whatever damage he can in the process, White’s queen still will be on a8 for the taking.
Capturing the rook on d1 first with check, then taking Black’s queen later, is an example of using the priority of check—the principle that a check must be addressed before any other threat pending on the board. It is a valuable idea that we will revisit often and study in detail later.