> ## Documentation Index > Fetch the complete documentation index at: https://help.memoryplugin.com/llms.txt > Use this file to discover all available pages before exploring further. # Memory Suggestions > AI-suggested cleanups for your memories: remove duplicates, combine related ones, update stale ones export const ArticleInfo = ({author, lastUpdated}) => { const authorAvatar = author === 'Alara' ? '/images/author-alara.jpg' : author === 'asad' ? '/images/author-alara.jpg' : null; const formatDate = dateInput => { if (!dateInput) return ''; if (typeof dateInput === 'string' && !dateInput.match(/^\d{4}-\d{2}-\d{2}/)) { return dateInput; } try { const date = new Date(dateInput); const now = new Date(); const diffTime = Math.abs(now - date); const diffDays = Math.floor(diffTime / (1000 * 60 * 60 * 24)); if (diffDays === 0) return 'today'; if (diffDays === 1) return '1 day ago'; if (diffDays < 7) return `${diffDays} days ago`; if (diffDays < 30) return `${Math.ceil(diffDays / 7)} week${Math.ceil(diffDays / 7) > 1 ? 's' : ''} ago`; if (diffDays < 365) return `${Math.ceil(diffDays / 30)} month${Math.ceil(diffDays / 30) > 1 ? 's' : ''} ago`; return `${Math.ceil(diffDays / 365)} year${Math.ceil(diffDays / 365) > 1 ? 's' : ''} ago`; } catch { return dateInput; } }; return
{authorAvatar ? {author} : author?.[0]?.toUpperCase()}
Written by {author}
Last updated {formatDate(lastUpdated)}
; }; Add memories for a few months and a bucket gets messy. You save "I use TypeScript" three times in slightly different words. Two memories say almost the same thing and could be one. A fact that used to be true has drifted. Memory Suggestions is the cleanup pass: MemoryPlugin scans a bucket, finds these cases, and proposes fixes for you to approve or reject. Think of it as a friend helping you clear out a junk drawer, holding up two near-identical things and asking "do you need both of these?" You still decide. Where the analogy breaks: it only looks for a few specific kinds of clutter (duplicates, mergeable pairs, outdated wording), not a full reorganization of the drawer. Memory Suggestions is available on all paid plans. It is not a Pro-only feature. ## The three kinds of suggestion Every suggestion falls into one of three operations, and each card gives you a plain accept-or-reject choice: Duplicate or redundant memories it recommends deleting. Accept with **Remove**, reject with **Keep**. Related memories that read better merged into a single cleaner entry. Accept with **Combine**, reject with **Keep separate**. A memory rewritten to be clearer or more current. Accept with **Save update**, reject with **Keep original**. Accepting shows a short confirmation ("Memory removed", "Memories combined", or "Memory updated"). Rejecting keeps things exactly as they were ("Kept as-is"). Nothing changes without your say-so. ## Running it Go to **Suggestions** in the sidebar. It sits under the memory section. Choose a bucket and click **Find suggestions**. MemoryPlugin analyzes the bucket ("Looking through your memories..."), which usually takes a minute or two. Suggestions appear as cards. Go through them one at a time, accepting or rejecting each. Once you've handled the current batch, use **Check for new** to scan memories not already covered by existing suggestions. Diagram: an AI curator proposes delete, combine, or update edits; a human approves or rejects before anything is written Diagram: an AI curator proposes delete, combine, or update edits; a human approves or rejects before anything is written Suggestions page showing a Remove card, a Combine card, and an Update card, each with accept and reject actions Suggestions page showing a Remove card, a Combine card, and an Update card, each with accept and reject actions ## The pending badge in the sidebar You don't have to remember to visit the page. When the selected bucket has suggestions waiting, the **Suggestions** item in the sidebar shows an **amber count badge**. That number is your pile of pending cleanups. When it's gone, you're caught up. ## What it skips Memory Suggestions is built for the short, factual memories that make up most of a bucket. It **skips any memory over 10,000 tokens**. Very large memories are unlikely to be duplicates of anything and are expensive to compare, so the analysis leaves them alone rather than burning time on them. Everything under that limit is fair game. ## Next steps Group a bucket into categories for token-efficient recall Select, move, and delete memories in batches yourself