{"version":"2.1.51","anchor":"llm-based-memory-retrieval","canonical_anchor":"llm-based-memory-retrieval","heading":"LLM-Based Memory Retrieval","tier":null,"area":null,"scope":null,"heads_up":null,"url":"https:\/\/changelogs.core-directive.com\/v\/2.1.51\/e\/llm-based-memory-retrieval","release_url":"https:\/\/changelogs.core-directive.com\/v\/2.1.51","markdown":"### LLM-Based Memory Retrieval\n\nThe memory system has been redesigned to use LLM-based retrieval instead of grep-based keyword search. Previously, memories were found by extracting search terms and grep-matching against markdown files. Now:\n\n- Memories are listed with filenames and descriptions\n- An LLM selects the most relevant memories (up to 5) based on the user's query\n- Memories already read via the `Read` tool in the current conversation are filtered out to avoid duplication\n- Memory files support a `description` frontmatter field for better retrieval\n\n**Evidence**\n\nSelection prompt `\"You are selecting relevant memories for a user's query...\"` and function `x8q` for deduplication (search for `\"You are selecting relevant memories\"`)"}