* Add message retry manager for session recreation and caching
Introduces a MessageRetryManager to cache recent sent messages and manage automatic Signal session recreation for failed message retries. Adds new config options to enable these features, integrates retry manager into message send/receive logic, and provides periodic cache cleanup. Improves reliability of message delivery and retry handling.
* Add buffer timeout and cache limit to event-buffer
Introduces a buffer timeout to auto-flush events after 30 seconds and limits the history cache size to prevent memory bloat in event-buffer.ts.
* Add session validation to SignalRepository
Introduces a validateSession method to SignalRepository for checking session existence and state. Updates message retry logic to use the new validation, improving session recreation handling and retry management.
* Improve message retry management and tracking
Refactored message retry manager to use LRU caches for recent messages and session history, added retry counters and statistics tracking, and implemented phone request scheduling. Updated message receive and send logic to mark retry success and pass max retry count. Removed periodic cleanup logic in favor of cache TTLs for better resource management.
* lint fix
* Use validateSession for session checks in message send
Replaces direct session existence checks with the improved validateSession method, which also checks for session staleness. Adds debug logging for failed session validations to aid troubleshooting.
* Delete src/Utils/message-retry.ts
---------
Co-authored-by: Rajeh Taher <rajeh@reforward.dev>
The `logger.info` makes Baileys' log verbose.
The flush is executed constantly and could be more useful during troubleshooting
sessions with log level = debug
* feat: implement state machine for chat synchronization and buffer management
* test: Add Jest configuration and tests for connection deadlock handling
- Created a Jest configuration file to set up testing environment.
- Implemented utility functions for creating isolated authentication sessions and mocking WebSocket connections.
- Added a test case to ensure that the connection does not deadlock when history sync is disabled, verifying the correct handling of message events.
* feat: add GitHub Actions workflow for running tests
* chore: sort import lint
* fix: implement timeout handling for AwaitingInitialSync state in chat socket, maybe fix memory leak
* feat: enhance chat synchronization by refining AwaitingInitialSync handling and adding history sync checks
* feat: interface "ILogger" created
feat: interface "ILogger" used instead of pino logger
feat: "PinoLoggerAdapter" created to implement "ILogger" interface
* feat: PinoLoggerAdapter removed
feat: ILogger mapping the features we're using from pino
* fix: sort imports
---------
Co-authored-by: Mateus Franchini de Freitas <contato.mateusfr@outlook.com>
Co-authored-by: Mateus Franchini de Freitas <mfranchini@domtec.com.br>
Co-authored-by: Rajeh Taher <rajeh@reforward.dev>
1. counter based event buffer keeps track of the number of blocks that request event processing in buffer
2. event buffer only releases events when the last block completes (i.e. counter = 0)
this approach is far simpler than the promised based garbled crap I wrote, should also prevent the deadlock issues it introduced 🙏
This is a breaking change,
1. three events (chats.set, contacts.set, messages.set) are now just one `messaging-history.set` event
2. no need to debounce for app state sync
3. added a new "conditional" chat update to allow for correct app state sync despite not having the chat available on hand
!BREAKING_CHANGE
This was generally a poor design decision on my part to add something like this -- primarily because it could increment the chat unread counter a 100 times for a single message
!BREAKING_CHANGE
1. this allows all offline notifications to be combined into a batch -- to reduce the number of events being sent out
2. to enable the above, the "message.reaction" event has been made an array. Also removes the need for the "operation" field
3. have also now started processing all events under a single queue to prevent state sync problems
Paulo Victor Lund