Module notifications 

API handlers for the persisted notification feed.

• GET /notifications - list notifications (keyset pagination, wallet query)
• POST /notifications/mark-read - mark specific ids read (signed)
• POST /notifications/mark-all-read - mark every unread notification read (signed)

Reads follow the same unsigned-with-wallet-query pattern as /fills and /profile so the frontend bell doesn’t need to sign on every page load. Writes stay behind EIP-712 signature auth; the wallet is extracted from the verified signature, not the body.

Structs

MarkAllReadResponse

MarkReadRequest

MarkReadResponse

NotificationDto

NotificationsListQuery

NotificationsListResponse

Functions

decode_payload 🔒

Decode a stored msgpack payload. A decode failure indicates persisted data corruption (schema mismatch, truncation, etc.), so we surface it loudly (log::error) so operators see it, but return null for the payload instead of failing the whole list — a single poisoned row shouldn’t take down the bell for the wallet’s entire history. The fail-loud log + null payload is the compromise between AGENTS.md’s fail-fast-on-invariant rule and read-path availability.

list_notifications

mark_all_read

mark_read

parse_wallet 🔒

Parse+lowercase the wallet query string, rejecting anything that isn’t a valid 0x-prefixed address. Parsing at the handler boundary keeps the rate limiter’s wallet-based path reachable (it only kicks in when the query parses as WalletAddress) and blocks a trivially-crafted 404-ish DB-load vector against the notifications endpoint.

to_dto 🔒