Overview
What is Satori?
Satori can remember facts about your data and conversations indefinitely. In other, more technical terms, Satori is memory architecture to store semantic data long term.
Any memories that are added are stored in a layer, and you can create unlimited layers. While newer memories are prioritized by our algorithm, older memories will last forever.
Satori uses a knowledge graphs alongside vectors, which identifies entities (people, places, things), and relationships (WORKS_FOR, FRIENDS_WITH, TRAVELED_TO). This gives meaning to your data, and allows you to ask questions that provide both relationships, temporal awareness (the understanding of time), and similarity.
Satori can semantically search data and relationships. There are two benefits:
- Reduces context, which increases recall accuracy over 30%
- Because the context window is smaller, it saves on tokens
You can add memory easily to any AI application using the CLI and our SDK in a couple lines of code.
See CLI Overview for more CLI commands.
What makes Satori unique?
Satori is the only memory provider that allows you to compose memory layers. This means you can take any number of memory layers you've built and query each individually or together.
This allows you to build applications, such as saving patient data and adding diagnosis layers to identify illnesses. Or, you might want to add your stock portfolio layer alongside a layer containing new financial analysis to see how it might perform.
Other providers give you a single memory layer.