Auto-Recall

Overview

Semantic retrieval from the perpetual memory vector store at query time. Parses query intent, searches the perpetual_memory LanceDB table, ranks results by relevance and recency, and returns structured context for injection into agent prompts.

Core principle: Every agent should be able to recall any past interaction instantly by meaning, not by filename or keyword. Auto-recall is the read side of the perpetual memory architecture.

When to Use