Google Senior AI PM Open-Sources Always-On Memory Agent That Ditches Vector Databases for LLM-Driven SQLite Storage

A Google senior AI PM open-sources an always-on memory agent that replaces vector databases with LLM-driven SQLite storage, running continuously on Google Cloud with Gemini 3.1 Flash-Lite, but enterprise experts warn the system raises serious compliance and governance risks, calling it a compelling engineering template rather than a production-ready platform.

• Google senior AI PM Shubham Saboo open-sources an 'Always On Memory Agent' on Google Cloud Platform's GitHub, built with Google's Agent Development Kit and Gemini 3.1 Flash-Lite, offering a persistent memory system that ditches traditional vector databases in favor of LLM-driven structured memory stored in SQLite.
• The agent runs continuously, ingesting files and API inputs while performing scheduled memory consolidation every 30 minutes, with Gemini 3.1 Flash-Lite's low cost of $0.25 per million input tokens and high speed making the 'always on' approach economically viable for enterprise teams.
• Enterprise reactions are raising serious governance concerns, with experts warning that background memory consolidation without deterministic boundaries creates compliance risks, and critics noting that removing vector databases doesn't eliminate retrieval complexity but simply shifts it, making the project a compelling engineering template rather than a production-ready enterprise platform.