rmyndharis/OpenWA v0.7.11

on GitHub

Added

• Disappearing-messages support (Baileys engine). Outbound messages now honor a chat's disappearing-messages timer and set it on each send (text, media, and replies), so recipients no longer see "This message won't disappear — the sender may be using an older version of WhatsApp." The timer is learned from inbound messages — the reliable source, since the cached chat setting is often absent for a long-standing timer — and resolved across both phone and @lid chat identifiers so it applies on LID-migrated 1:1 chats, with a fallback to the chat's cached setting. It is applied only when a positive value is known; when it's unknown or disabled, the per-message expiration is omitted, exactly as before. Reactions, deletes/revokes, and status posts are unaffected. Thanks @ulises2k. (#473, #513)
• Selective skip for disappearing messages. New STORE_EPHEMERAL_MESSAGES env var (default true). Set to false to skip persisting and dispatching incoming disappearing messages (those with ephemeralDuration > 0) — no DB insert, no webhook dispatch, no websocket event. Backward compatible; existing deployments are unaffected. The ephemeralDuration field is also surfaced on IncomingMessage for consumers that want to handle it themselves. Thanks @spidgrou. (#506)
• Durable dead-letter record for failed webhook deliveries. A webhook delivery that permanently fails — exhausting its retries or being rejected before it is sent — is now persisted to a new webhook_delivery_failures table instead of disappearing when its job is evicted from the queue. Operators can review the recorded failures (endpoint, event, status, error, attempts) through a new admin endpoint, GET /webhooks/delivery-failures. (#520)

Fixed

• Deleting a session now removes its message history and bulk batches. The messages and message_batches tables had no cascade from sessions, so a deleted session left its rows behind — growing the largest tables without bound and skewing dashboard statistics. They are now removed in the same transaction as the session. (#504)
• Deleting a session while it is reconnecting no longer leaks its engine. A delete that landed during the multi-second engine initialization of an in-flight reconnect (or start) could leave the freshly-launched browser/socket registered under the now-deleted session, still counting toward the concurrent-session limit. The post-init guard now re-checks that the session still exists before keeping the engine. (#521)
• Inbound media downloads are bounded by a wall-clock timeout. A slow or stalled inbound media transfer could hold a download slot — and, on the Baileys engine, the entire inbound-message pipeline — open indefinitely. Downloads now time out (MEDIA_DOWNLOAD_TIMEOUT_MS, default 30s) and the message is delivered with the media omitted. (#510)
• Webhook delivery identifiers stay consistent with the signed body. The X-OpenWA-Idempotency-Key / X-OpenWA-Delivery-Id headers could diverge from the signed payload when a webhook:before plugin returned a modified payload, and all webhooks for an event shared one data object. Each webhook now receives an isolated copy of the data and the server-generated identifiers are authoritative. (#512)
• POST /auth/validate no longer double-counts key usage and now validates IP-restricted keys correctly (it previously reported a valid IP-pinned key as invalid). (#507)
• ⚠️ GET /settings now requires an ADMIN key (behavior change) — matching the rest of the configuration surface; it was previously readable by any authenticated key. A client that read settings with a non-admin key must switch to an ADMIN key. (#514)
• Bulk-message batchId uniqueness is scoped per session, so two sessions can reuse a batch id and neither can probe the other's id namespace. (#515)
• ⚠️ Boot-time configuration validation now rejects 0 for the rate-limit limits and the webhook timeout (behavior change) — values that silently disabled throttling or aborted every delivery. A deployment that set 0 to disable these must remove the override or use a positive value. (#516)
• SSRF protection now blocks the RFC6052 IPv4-translatable IPv6 form (::ffff:0:a.b.c.d), closing a gap where an internal address could be reached behind a NAT64/SIIT translator. (#518)
• Per-key IP allowlist now uses the shared, hardened IP matcher and rejects a malformed client address instead of coercing it into an allowed range. (#519)
• Dashboard: the Infrastructure page is no longer rendered for non-admin roles, and image-attachment preview object URLs are released after use. (#508)
• Released a small in-memory leak: a deleted session's stored failure reason is now cleared. (#505)
• The webhook worker now connects to the configured Redis. Configuration from .env and the dashboard-saved file is loaded before the application modules are evaluated, so the webhook delivery worker reads its Redis host/port/password from the configured values instead of falling back to a local default when those are supplied by file rather than the process environment. (#523)

Performance

• Configurable webhook worker concurrency (WEBHOOK_WORKER_CONCURRENCY, default 10): a single slow or unresponsive receiver no longer head-of-line-blocks delivery for every other webhook. (#511)
• Dropped a redundant single-column index on messages(sessionId) already covered by the existing composite indexes, reducing write-time overhead on a high-volume table. (#509)