What triggers an automatic retry?

Retries are triggered by transient failures — network timeouts, 5xx server errors, WebSocket disconnections, and rate limit responses (429). Permanent failures like 4xx client errors or validation errors are not retried because they would fail again.

How does exponential backoff work?

The first retry happens after 1 second. Each subsequent retry doubles the wait time: 1s, 2s, 4s, 8s, up to a maximum of 30 seconds. A small random jitter (0-500ms) is added to each delay to prevent thundering herd problems when many clients retry simultaneously.

How many times does WISEROWS retry a failed request?

By default, requests retry up to 3 times with exponential backoff. For critical operations like data writes, the system may retry up to 5 times. If all retries fail, the user sees a clear error message with a manual retry option.

What are optimistic updates?

Optimistic updates apply changes to the UI immediately before the server confirms them. When you move a kanban card or edit a field, the change appears instantly. If the server request fails and retries are exhausted, the optimistic update is rolled back and the UI reverts to the previous state.

Does WISEROWS queue operations when I go offline?

Yes. When the connection drops, mutations are queued locally. The queue is processed in order when connectivity returns. Convex handles this natively through its reactive sync protocol — pending mutations are replayed automatically.

How does WebSocket reconnection work?

Convex maintains a persistent WebSocket connection for real-time sync. If the connection drops, it reconnects automatically with exponential backoff. During reconnection, the UI continues to work with cached data. Once reconnected, any missed updates are synced immediately.

Will I see error messages during retries?

No. Retries happen silently in the background. You will not see any error toasts, loading spinners, or disruptions during the retry window. Only if all retries fail (which is rare) will the system show a non-intrusive error notification with a manual retry option.

Does Convex have built-in retry semantics?

Yes. Convex automatically retries transient failures at the database level. Mutations are atomic and idempotent — if a mutation is retried, it produces the same result. This means WISEROWS gets infrastructure-level retry guarantees in addition to application-level retries.

What happens to in-flight data during a network drop?

In-flight mutations are held in a local queue and replayed when the connection returns. Queries are served from the local cache during the outage. The Convex sync protocol ensures consistency — you will never see partial or out-of-order updates after reconnection.

Can retries cause duplicate operations?

No. Convex mutations are designed to be idempotent. Each mutation has a unique identifier that prevents duplicate execution. If a retry sends the same mutation twice, the server recognizes it and returns the result of the original execution.

Automatic Retry

Failed requests fix themselves

Network hiccups happen. WISEROWS retries failed requests silently in the background with exponential backoff. You keep working — the system catches up.

Network TimelineTotal: 3.1s (transparent to user)

0ms

Request sent

200ms

Network timeout

1.0s

Retry 1 (backoff)

1.2s

Network timeout

3.0s

Retry 2 (backoff)

3.1s

Success — 200 OK

User saw: instant optimistic update

No error shown

Exponential backoff

Each retry waits longer than the last — 1s, 2s, 4s, 8s — giving the server time to recover. A small random jitter prevents all clients from retrying at the same instant. Smart, patient, invisible.

First retry after 1 second — fast recovery for transient issues
Doubling delay on each attempt: 1s → 2s → 4s → 8s
Maximum backoff capped at 30 seconds
Random jitter (0-500ms) prevents thundering herd
Up to 3 retries for standard ops, 5 for critical writes
Only transient failures retry — 4xx errors fail immediately

Exponential Backoff

Retry 1

Retry 2

Retry 3

Retry 4

Retry 5

16s

+ random jitter (0-500ms) to prevent thundering herd

Optimistic updates

Changes apply to the UI the moment you make them — before the server confirms. Drag a kanban card, edit a field, toggle a status — it happens instantly. If the server rejects the change, the UI rolls back smoothly.

UI updates before server confirmation — zero perceived latency
Failed mutations roll back gracefully with visual feedback
Convex handles optimistic state natively via reactive sync
No loading spinners for inline edits and status changes
Concurrent edits merge automatically via conflict resolution
Rollbacks are rare — most operations succeed on first attempt

Optimistic Update Flow

User action0ms

Move card to "Published"

Optimistic update0ms

UI shows card in Published

Server request5ms

Mutation sent to Convex

Confirmed50ms

Server confirms, state locked

Offline queue

Lost your connection on a train? Keep working. WISEROWS queues your mutations locally and replays them in order when connectivity returns. Convex's reactive sync protocol handles the rest.

Mutations queued locally during network outages
Queue processed in-order on reconnection
Queries served from local reactive cache
WebSocket reconnects with exponential backoff
No "connection lost" banner for brief outages (<5s)
Missed real-time updates synced immediately on reconnect

Connection States

Connected

Real-time sync active

Offline

Mutations queued, cache served

Reconnecting

Backoff retry in progress

Reconnected

Queue replayed, updates synced

Built on Convex's retry semantics

WISEROWS runs on Convex, which provides infrastructure-level retry guarantees. Mutations are atomic and idempotent. Retried operations never duplicate. Consistency is guaranteed.

Atomic

Mutations succeed completely or not at all

Idempotent

Retried mutations produce the same result

Consistent

No partial or out-of-order state

Default retries

30s

Max backoff

Duplicate mutations

<50ms

Optimistic latency

How it feels

The network is not your problem.

You drag a card, edit a field, save a note. It happens instantly. Behind the scenes, the request might fail and retry twice — but you will never know. The UI stays responsive, your data stays consistent, and the only thing you notice is that everything just works. Even on a train. Even on conference Wi-Fi.

Frequently asked questions

Your data is incomplete. Fix it.

Define your schema. Import your data. Let AI enrich every record. See quality scores improve in real time. Free to start, no credit card.