Fabric Memory API

Using Fabric Memory API, you can upgrade your LLM assistants and agents with long-term memory capabilities. This enables them to remember past interactions, user preferences, and important information across multiple sessions or tasks.

Memory organization

Fabric Memory automatically self-organizes into a tree-like structure of concepts and memories.

• Concepts: High-level generated categories or topics that group related memories together. Some examples of concepts could be: "Work", "Personal", "Projects", etc.

• Memories: Individual pieces of information or experiences associated with a concept. By default, memories are created by inferring them from input data that you provide. You may also choose to define verbatim memories by providing exact snippets of text. New memories will be automatically associated with the most relevant concept.

By default memories are managed automatically. This means they can be modified, deleted, consolidated, and recategorized by the system as time goes on. You can opt to make memories read-only by setting the immutable flag during creation. This will prevent them from being modified or deleted.