Add durable EvalOps memory sync queue

[haasonsaas]

Summary

• Add a local SQLite queue for failed EvalOps Memory sync writes.
• Retry queued chat, meeting-summary, and journal syncs with bounded backoff from a background drain loop.
• Force-flush queued syncs after successful EvalOps login and from the Settings retry button.
• Surface pending/failed queue status in EvalOps Settings.

Refs #15.

Validation

• npm ci --ignore-scripts
• npm run build
• git diff --check

Validation caveat

• npm test initially failed because the fresh clone did not have the native ContextKit binary.
• After npm run contextkit:build, npm test got through ContextKit and meeting detection, then failed in audio-data-size.test.mjs because the local WAV duration reported 0.0s despite receiving mic buffers and writing a 158 KB mic file. That appears unrelated to this memory-sync queue change.

[cursor]

PR Summary

Medium Risk
Introduces new persistent DB state and a background retry loop that can affect app lifecycle and network call patterns. Failures/backoff and forced flush on login could surface edge cases (duplicate attempts, contention, or unexpected load) if the queue logic misbehaves.

Overview
Adds a durable local queue (evalops_memory_sync_queue) for failed EvalOps Memory sync writes, including an index for efficient scheduling and a bounded exponential backoff retry policy.

Memory sync background jobs now enqueue on failure and are drained periodically from the main process (plus force-flushed after successful EvalOps login), with new IPC endpoints (evalops:memorySync:status/flush) and settings UI surfacing pending/last-error state and a manual Retry action.

^{Reviewed by Cursor Bugbot for commit 55de74d. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Bugbot Autofix is ON, but it could not run because the branch was deleted or merged before autofix could start.}

^{Reviewed by Cursor Bugbot for commit 55de74d. Configure here.}

[cursor]

Force flush flag silently dropped during concurrent execution

Medium Severity

flushEvalOpsMemorySyncQueue returns the existing queueFlushPromise when a flush is already in progress, completely ignoring the force flag. If the periodic (non-force) drain is running when the user logs in or clicks Retry, the force: true call silently joins the non-force drain — which only processes items past their nextAttemptAt. Items backing off (e.g., with a 30-minute delay) won't be retried as expected.

Additional Locations (1)

• src/main/evalops/handlers.ts#L42-L45

 

^{Reviewed by Cursor Bugbot for commit 55de74d. Configure here.}

[cursor]

Queued jobs silently deleted when sync category disabled

Medium Severity

runMemorySyncJob delegates to syncChatThreadMemory/syncMeetingSummaryMemory/syncJournalEntryMemory, each of which returns early (no error) when isMemorySyncEnabled is false. On success, removeQueuedMemorySync deletes the job from the durable queue. So if a sync category is disabled between initial failure and retry, the drain loop silently discards queued items without actually syncing them.

Additional Locations (1)

• src/main/evalops/memory-sync.ts#L206-L213

 

^{Reviewed by Cursor Bugbot for commit 55de74d. Configure here.}