Agents using API channel inboxes (e.g., WhatsApp Automate) reported seeing the same conversation appear twice in their conversation list — one showing the last message preview and the other showing "No Messages". Backend investigation confirmed no duplicate conversations exist in the database, making this purely a frontend issue. The root cause is a race condition in WebSocket event delivery. When a conversation is created via the API with auto-assignment, the backend enqueues multiple ActionCable broadcast jobs (`conversation.created`, `assignee.changed`, `team.changed`) within milliseconds of each other. In production with multi-threaded Sidekiq workers, these events can arrive at the frontend out of order. If `assignee.changed` arrives before `conversation.created`, the `UPDATE_CONVERSATION` mutation pushes the conversation into the store (since it doesn't exist yet), and then `ADD_CONVERSATION` blindly pushes it again — resulting in a duplicate entry. The fix adds a uniqueness check in the `ADD_CONVERSATION` mutation to skip the push if a conversation with the same ID already exists in the store, matching the dedup pattern already used by `SET_ALL_CONVERSATION`. |
||
|---|---|---|
| .. | ||
| dashboard | ||
| design-system | ||
| entrypoints | ||
| portal | ||
| sdk | ||
| shared | ||
| superadmin_pages | ||
| survey | ||
| v3 | ||
| widget | ||
| histoire.setup.ts | ||