Good news!
Notice Google uses the SQLite database. Me too!
"An open-sourced agent that ditches vector databases and embeddings entirely. Instead it runs continuous LLM consolidation on Gemini 3.1 Flash-Lite — ingesting, connecting, and compacting context on the fly (1M input / 64K output) so the model just remembers rather than retrieving chunks.
So what: If a cheap enough model can consolidate memory in-context, the entire vector-DB + embedding pipeline becomes optional plumbing. It's a direct bet that falling inference cost eats RAG — and it's coming from Google itself."
"Most AI agents have amnesia. They process information when asked, then forget everything. This project gives agents a persistent, evolving memory that runs 24/7 as a lightweight background process, continuously processing, consolidating, and connecting information.
No vector database. No embeddings. Just an LLM that reads, thinks, and writes structured memory.
The Problem
Current approaches to LLM memory fall short:
Approach Limitation
Vector DB + RAG Passive. Embeds once, retrieves later. No active processing.
Conversation summary Loses detail over time. No cross-reference.
Knowledge graphs Expensive to build and maintain.
The gap: No system actively consolidates information like a human brain does. Humans don't just store memories. During sleep, the brain replays, connects, and compresses information. This agent does the same thing. ..."
No comments:
Post a Comment