The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
docs	Skipped		May 19, 2026 7:02pm

PR Summary

Medium Risk
Changes deletion criteria and adds storage-key cleanup during log retention, plus rewires plan/workspace resolution for dispatching cleanup jobs; mistakes could cause unintended data retention or premature deletion. Uses DB text search and additional subscription lookups, which may impact performance at scale.

Overview
Extends cleanup-logs to also delete cloud-stored “large execution values” referenced from workflowExecutionLogs.executionData, while avoiding deletion when the same storage key is still referenced by non-deleted logs; it also skips deleting logs for executions in resumable paused states and adds detailed deletion stats logging.

Reworks retention scoping in cleanup-dispatcher to derive workspace sets from the same effective plan lookup used elsewhere (handling personal vs organization workspaces), and tightens enterprise cleanup eligibility by verifying the workspace is org-mode and the org’s current subscription resolves to enterprise before applying org-configured retention.

Adds onError options to subscription lookup helpers (getOrganizationSubscription, getHighestPrioritySubscription, and new getHighestPriorityPersonalSubscription) so cleanup dispatch can fail fast per-tenant and skip problematic rows rather than silently misclassifying plans.

^{Reviewed by Cursor Bugbot for commit b9ffe66. Configure here.}

Greptile Summary

This PR extends the log cleanup job to also delete orphaned large-value storage refs (S3 keys embedded in executionData) when their associated execution logs are purged, and refactors the cleanup dispatcher to use getHighestPriorityPersonalSubscription — fixing a bug where personal workspaces owned by enterprise org members were incorrectly placed in the enterprise cleanup bucket instead of the free-plan bucket.

• Large-value key cleanup (cleanup-logs.ts): on each batch, collectLargeValueKeys extracts storage keys from executionData, then filterLargeValueKeysWithoutRetainedReferences checks whether any surviving log row still references each key before deleting from storage. The cross-batch reference guard is correct — onBatch runs before the DB DELETE, so the notInArray exclusion of current-batch IDs is sufficient.
• Dispatcher refactor (cleanup-dispatcher.ts): replaces the old SQL JOIN against the subscription table with per-user calls to getHighestPriorityPersonalSubscription, isolating personal-workspace plan resolution from org-level subscriptions. Individual per-user errors are now caught and logged without aborting the entire dispatch.
• Billing utilities (billing.ts, plan.ts, subscription.ts): adds onError: 'throw' option to getOrganizationSubscription and getHighestPrioritySubscription, and extracts getHighestPriorityPersonalSubscription as a separate, personal-subscription-only lookup.

Confidence Score: 5/5

Safe to merge — the core cleanup correctness is sound and the dispatcher bug fix is well-reasoned. The two comments are background-job performance observations that do not affect data integrity.

The cross-batch reference guard in filterLargeValueKeysWithoutRetainedReferences is logically correct: onBatch runs before the DELETE, so excluding current-batch IDs via notInArray accurately represents which logs will survive. The dispatcher correctly isolates personal-workspace plan resolution from org-level subscriptions, fixing the previously identified enterprise-member bug. The per-key sequential full-table scan and redundant workspace queries in dispatchCleanupJobs are worth addressing but do not produce wrong results.

apps/sim/background/cleanup-logs.ts — the per-key position() scan loop warrants a closer look before this runs at enterprise scale.

Important Files Changed

Sequence Diagram

sequenceDiagram
    participant CJ as CleanupLogsJob
    participant CBD as chunkedBatchDelete
    participant DB as Database
    participant Filter as filterLargeValueKeysWithoutRetainedRefs
    participant S3 as StorageService

    CJ->>CBD: chunkedBatchDelete(workflowExecutionLogs)
    loop per workspace chunk x batch
        CBD->>DB: "SELECT rows WHERE workspaceId IN AND startedAt < retentionDate"
        DB-->>CBD: rows[]
        CBD->>CJ: onBatch(rows)
        CJ->>Filter: filterLargeValueKeysWithoutRetainedRefs(keys, deletedLogIds)
        loop per unique key
            Filter->>DB: SELECT id WHERE id NOT IN deletedLogIds AND position(key in executionData) LIMIT 1
            DB-->>Filter: referencingLog?
        end
        Filter-->>CJ: unreferencedKeys[]
        CJ->>S3: deleteFile for each unreferencedKey
        CBD->>DB: DELETE WHERE id IN batch ids
    end

_{Reviews (3): Last reviewed commit: "cleanup code" | Re-trigger Greptile}

bugbot run

@greptile

bugbot run

@greptile