- Why are we building our memory system? we need to run on user devices in a private focus and limited environment we need temporal and context-aware memories about facts and life experiences we need a human-like system with the ability to forget, pay attention, and decide whats matters and whats is important -Why it is/will be much better than anything else out there (for our use case)? we see a limitation of classical knowledge graphs, and at the same time, technically, hyper and meta graphs have no tools and enough research. So we build a concept graph on event steroids that temporal aware we could model complex semantic facts and semantic schemas in the flexible way we could capture events and life experiences and build relations between them we could build smart recall and flexible forgetting
- Why do we do it this way? We believe in a hybrid neuro-symbolic approach that uses the best sides of classical ai and deep learning reasoning and logical systems on top of knowledge graphs empower LLM capabilities to reason and reduce hallucinations durable extraction and processing pipeline gives the agent the ability to think deeply
why we build our own memory system for ai agent
Loading comments...