Google Cloud's Always-On Memory Agent Replaces RAG and Embeddings With Continuous LLM Consolidation on Gemini 3.1 Flash-Lite

Google Cloud released a sample Always-On Memory Agent built with Google ADK and Gemini 3.1 Flash-Lite. The reference implementation runs continuously as a background process, using an LLM to read, think, and write structured memory into SQLite without a vector database or embeddings. Three sub-agents handle ingestion, consolidation, and queries. The ConsolidateAgent links related memories every 30 minutes, surfacing new insights while idle. It accepts 27 file types dropped into an inbox folder.