Feature / Open Source AI Memory Database

AI Memory Pruning and Forgetting

Memorose includes memory maintenance flows so long-lived agents can update, consolidate, or forget memory deliberately instead of accumulating stale context forever.

Read the docs Star on GitHub

The problem this solves

Persistent memory becomes harmful when obsolete facts, duplicate summaries, and low-value traces keep influencing future retrieval.

How Memorose handles it

Memorose separates preview from execution, giving operators a way to inspect semantic updates or forgetting behavior before changing durable memory.

runtime example 
curl -X POST "$BASE_URL/v1/users/dylan/memory/forget/preview" \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"reason":"remove stale deployment assumptions"}'
Next paths
AI agent memory Hybrid retrieval Benchmarks Documentation GitHub repository