RELIABILITY
Daily memory gets sturdier—and a second layer of protection.
Memory is central to Life Tracker, so reliability has two layers. You need to see a correct timeline now, while the service needs a tested recovery path if infrastructure fails later.
Facts and scope come from real product changes, with their impact explained in plain language.
Time behaves consistently
The update adds a shared time formatter and improves daily-memory, voice and timeline state. Edge values and local display now follow the same rules throughout the app.
Synchronization should not erase a newer change or leave the screen in a partial state. Memory-service and timeline-drag tests cover those behaviors independently.
Create a copy without stopping the database
The server uses a single-transaction mysqldump, compresses the result and keeps it temporary until the operation completes. An empty or interrupted archive never becomes a finished backup.
Every archive receives a SHA-256 checksum. After uploading to private R2 storage, the process compares remote and local size before reporting success.
Daily, with an earlier retry
A system timer starts the backup at 03:30 with a small randomized delay. If the run fails, the local file stays in place and another attempt happens sooner than the normal daily interval.
Local retention defaults to seven days. This is service recovery infrastructure, not a new use of personal information: archives remain private and exist only for restoration.
Consistent time
One format across memory workflows.
Verified archive
Checksum and size confirm integrity.
Recovery routine
Nightly execution and retry after failure.
Extra AI access without quota confusion.
Unified usage handling and a transparent rewarded-access flow make free Life AI more predictable.