[MohamedBassem]

No description provided.

[greptile-apps]

Confidence Score: 4/5

The PR should not merge until queued recrawls are represented correctly so completed retries automatically disappear from the broken-links view.

Findings

1.  Recrawls Never Enter Polling ▶

Fix with agent prompt

### Issue 1
packages/trpc/routers/bookmarks.ts:1484
Retrying a broken link only enqueues a job; it does not change `crawlStatus` to `"pending"`. Because `isCrawling` is derived solely from that status and the page polls only while `isCrawling` is true, the row continues to offer another retry and may remain displayed as broken after a successful recrawl until the user manually reloads or revisits the page. Track queued recrawls explicitly or update the persisted status when enqueueing.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

Summary

This PR redesigns broken-link management with filtering, sorting, bulk retry and deletion, richer status classification, and expanded bookmark metadata.

• Adds reusable maintenance-state, search, and selection components.
• Extends getBrokenLinks with titles and crawl-state information.
• Adds confirmation and concurrency-limited handling for destructive bulk actions.
• The queued recrawl state is not observable through the new API field, preventing the intended automatic refresh from working.

_{Reviews (1) · Last reviewed commit: "feat(web): improve broken links manageme..."}

[greptile-apps]

Recrawls Never Enter Polling

Retrying a broken link only enqueues a job; it does not change crawlStatus to "pending". Because isCrawling is derived solely from that status and the page polls only while isCrawling is true, the row continues to offer another retry and may remain displayed as broken after a successful recrawl until the user manually reloads or revisits the page. Track queued recrawls explicitly or update the persisted status when enqueueing.

Knowledge Base Used: Bookmark and organization APIs

Prompt To Fix With AI

This is a comment left during a code review.
Path: packages/trpc/routers/bookmarks.ts
Line: 1484

Comment:
**Recrawls Never Enter Polling**

Retrying a broken link only enqueues a job; it does not change `crawlStatus` to `"pending"`. Because `isCrawling` is derived solely from that status and the page polls only while `isCrawling` is true, the row continues to offer another retry and may remain displayed as broken after a successful recrawl until the user manually reloads or revisits the page. Track queued recrawls explicitly or update the persisted status when enqueueing.

**Knowledge Base Used:** [Bookmark and organization APIs](https://app.greptile.com/karakeep/-/custom-context/knowledge-base/karakeep-app/karakeep/-/docs/bookmark-organization-apis.md)

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

[pullfrog]

Important

The new "Queued" indicator and auto-refresh polling never engage for a retried link — no code path resets a broken link's crawlStatus back to "pending". One inline comment with details. Everything else in the rewrite looks solid.

Reviewed changes

• Richer broken-links UI — apps/web/app/settings/broken-links/page.tsx is rewritten with category cards, search, sort, multi-select, bulk retry/delete with a confirm dialog, error/empty/skeleton states, and a refetch poll while any row reports isCrawling.
• Shared maintenance components — new apps/web/components/settings/Maintenance.tsx exporting SelectionCheckbox (indeterminate support), MaintenanceSearch, MaintenanceEmpty, MaintenanceError.
• Broken-link classification helpers — new apps/web/lib/broken-links.ts with getBrokenLinkIssue (blocked/missing/server/http/failed), getLinkHostname, stripUrlProtocol.
• getBrokenLinks output widened — now returns title (joined from bookmarks/bookmarkLinks) and isCrawling (packages/trpc/routers/bookmarks.ts).
• Test coverage — bookmarks.test.ts asserts the new title/isCrawling fields and per-user isolation.
• Supporting plumbing — MAX_CONCURRENT_BULK_ACTIONS exported for reuse; new settings.broken_links.* / settings.maintenance.* keys in en + en_US.

ℹ️ Nitpicks

• settings.maintenance.loading is added to both locales but never referenced anywhere.
• hasFilters in page.tsx uses the raw search string, so a whitespace-only query still renders "Clear filters" even though filtered is unfiltered; compare against searchQuery instead.

  ^{｜ Fix all ➔ ｜ Fix 👍s ➔ ｜ View workflow run ｜ Using DeepSeek Flash (free via Pullfrog for OSS) ｜ 𝕏}

[pullfrog]

Addressed both review threads and the review-body nitpicks:

• recrawlBookmark now sets crawlStatus to pending after a successful enqueue (leaving crawlStatusCode untouched so the row stays in getBrokenLinks), making the "Queued" state and polling reachable. Updated the test to call recrawlBookmark and assert isCrawling: true / isCrawlingFailure: false.
• Bulk partial failures now keep the failed ids selected (and the delete dialog keeps them) so they remain recoverable; the toast already reports the failure count.
• hasFilters now compares against the trimmed searchQuery, and the unused settings.maintenance.loading key was removed from both locales.

Verified with vitest (all 43 bookmarks.test.ts tests pass), tsc --noEmit for packages/trpc and apps/web, plus oxlint/oxfmt on the changed files. Pushed in 2eb3f531.

Task list (1/6 completed)

• Read PR diff and relevant source files
• Fix recrawlBookmark to reset crawlStatus to pending
• Keep failed ids selected on partial bulk failure
• Fix hasFilters to use searchQuery; remove unused loading i18n key
• Add/update tests for recrawl pending state
• Typecheck/lint, commit, push, reply to threads

  ^{｜ View workflow run ｜ via Pullfrog ｜ Using DeepSeek Flash (free via Pullfrog for OSS) ｜ 𝕏}

[pullfrog]

✅ No new issues found.

Reviewed changes

Changes since the prior pullfrog review (the fix commit for the two findings it raised):

• Retried links now enter the pending state — recrawlBookmark sets bookmarkLinks.crawlStatus = "pending" after a successful enqueue, leaving crawlStatusCode untouched so the row stays in getBrokenLinks and the "Queued" button state plus the 10s refetchInterval become reachable.
• Partial bulk failures stay recoverable — runAction keeps the failed ids selected (and in the delete dialog) rather than clearing the selection, and the toast now reports the failure count.
• Filter/empty-state fix — hasFilters compares the trimmed searchQuery instead of the raw input, and the unused settings.maintenance.loading keys were removed.
• Test coverage — the getBrokenLinks test now sets a failure status and calls recrawlBookmark, asserting isCrawling: true / isCrawlingFailure: false (the LowPriorityCrawlerQueue mock was added). Verified locally: the test passes.

  ^{｜ View workflow run ｜ Using deepseek-v4.1-flash (free via Pullfrog for OSS) ｜ 𝕏}