cleanup

/memora:cleanup — Memory Maintenance

Find duplicates, merge similar memories, and prune stale items.

Workflow

1. Find duplicates:

   memory_find_duplicates(min_similarity=0.7, max_similarity=0.95, limit=10, use_llm=false)
   

   Set use_llm=true only if the user wants AI-powered comparison.

2. Present duplicate pairs with:

   • Both memory previews
   • Similarity score
   • Suggested action (merge/keep_both/review)

3. For each pair the user approves, merge:

   memory_merge(source_id=<lower_quality>, target_id=<higher_quality>, merge_strategy="append")
   

4. Find stale items:

   memory_insights(period="30d", include_llm_analysis=false)
   

   Report open TODOs and issues that haven't been touched in 14+ days.

5. Rebuild indexes if needed:

   • After significant merges: memory_rebuild_crossrefs()
   • After embedding model change: memory_rebuild_embeddings()

6. Report summary: duplicates merged, stale items flagged, total memories before/after.

Arguments

Argument	Action
(none)	Find duplicates + stale items
duplicates	Only find duplicates
stale	Only find stale items
rebuild	Rebuild embeddings and crossrefs
deep	Use LLM for duplicate comparison

Safety

• Always present duplicate pairs for user approval before merging
• Never auto-delete memories without confirmation
• Show before/after counts