Bookmarks

Core Feature

Learn how URL and PDF ingestion, collection assignment, and real-time updates work.

Overview

When you save a URL or upload a PDF in Relink, the item is inserted immediately and then enriched in the background. Extraction preserves the full available content, and the AI pipeline uses that material to generate summaries, sections, tags, related context, collection assignment, and embeddings for semantic search.

Saving URLs and PDFs

You can save URLs or PDFs from the Omnibar on desktop or mobile. Both flows support assigning a collection up front, and if you do not choose one manually, AI can place the item into a single collection folder during processing.

PDF uploads are plan-limited. The Free plan allows files up to 5 MB each with 100 MB total PDF storage, Pro allows up to 30 MB per PDF, and Team allows up to 50 MB per PDF.

Processing States

processingThe card appears immediately while extraction, summarization, tagging, and collection assignment finish in the background.
readyProcessing complete. The card updates in place with final type, summary, tags, collection, and related-memory support.
failedCould not extract content (e.g., paywalled or anti-bot protections).

Collections

Collections are the long-term folder structure for your archive. The dashboard sidebar renders them as a tree, supports nested folders, and counts items recursively so parent folders include everything in descendant subfolders.

On desktop, the collection tree lives beside the grid. On mobile, the same selector is available from a dedicated dialog above the grid.

The archive also supports a dedicated selection mode for bulk actions. You can select multiple bookmarks from the grid, move them into a collection in one step, or send the whole selection to Trash without opening each item individually.

Archive filters stack with collections. You can narrow the visible set by tags and content type, and the collection counts update against the currently filtered archive instead of the full library.

Notes

Every bookmark has a private scratchpad where you can write notes. Notes are auto-saved instantly. They are not factored into semantic search indexing, but are included as context when chatting with a specific article.

Trash & Soft Deletion

Deleting a bookmark moves it to the Trash (soft delete — sets a deletedAt timestamp). Trashed links are hidden from all views and excluded from search, chat, and synthesis. They can be restored or permanently deleted from /dashboard/trash.

Trash Lifecycle

DeleteSets deletedAt. Link disappears from dashboard and all queries. On Pro and Team, PDF-backed bookmarks keep the original uploaded file while they remain in Trash.
RestoreClears deletedAt. Link reappears in the dashboard immediately, including the original uploaded PDF if it has not been permanently deleted from Trash.
PurgePermanently removes the row, embeddings, and all AI metadata. Irreversible.
Auto-purgeA cron job at /api/cron/purge-trash automatically purges items older than 30 days.

Real-time Updates (SSE)

The dashboard automatically watches processing items using Server-Sent Events (SSE). When AI finishes, the card updates in place without a page refresh, including status, content type, collection assignment, and other UI-critical metadata.

SSE Subscription Example (Internal usage)

const sse = new EventSource('/api/links/stream?ids=uuid1,uuid2');

sse.onmessage = (event) => {
  const updatedLink = JSON.parse(event.data);
  console.log("Link is ready:", updatedLink.tldr);
};