Devlog #16 ->
Part 2, continuing from #15

Got retries working on top of the queue from last time. If Gemini comes back with a 429 (rate limited) or 503 (overloaded) instead of an actual answer, the backend now waits and tries again automatically, using exponential backoff (2s, then 4s, then 8s) instead of hammering it again right away, up to a few attempts before it actually gives up and reports failure. It also checks for a retry-after header first and uses that instead of the backoff timer if Gemini actually tells us how long to wait. Then I mirrored the same idea on the Flutter side, so if the backend’s retries still aren’t enough, the app itself will retry the whole request a couple more times before showing the user an error.