Google Cloud has released a sample agent that removes the need for vector databases and embeddings. The Always-On Memory Agent runs as a continuous background process using Google ADK and Gemini 3.1 Flash-Lite. It stores structured memory directly in SQLite instead of relying on external vector stores.
The architecture
An orchestrator routes requests to three specialist sub-agents. Each agent has its own tools for reading or writing the memory store.
The IngestAgent handles incoming content. It uses Gemini’s multimodal capabilities to extract a summary, entities, topics, and an importance score. The structured record then lands in the memories table.
The ConsolidateAgent runs on a timer, every 30 minutes by default. It reviews unconsolidated memories and finds connections between them. The agent writes a synthesized summary, one key insight, and those connections to the database. This builds new understanding while idle, with no prompt required.
The QueryAgent answers questions. It reads all memories and consolidation insights, then synthesizes a response. It cites the memory IDs it used as sources.
What it means
Developers no longer need to manage complex vector search infrastructure. A single lightweight process handles ingestion, consolidation, and retrieval. The system reduces costs and latency by avoiding embeddings. It offers a simpler path to persistent memory for background AI tasks.
