EverMemoryArchive

Appearance

ema / @internals/db

@internals/db ​

Database interfaces for EverMemoryArchive.

  • We associate roles with prompt and assets.
  • We can clone roles to create actor entities. Each actor entity has a unique state, such as memory buffer.
  • We can create users to access multiple actor entities.
  • We can create conversations with actors. The conversations are not the same as the conversations in the system, but the messages array passed to the llm agent when calling openai APIs (chat.completions.create).
  • We can create conversation messages, associated with a conversation.
  • We can create short term memories for actors. The short term memories are associated with conversation messages ( for debugging purpose).
  • We can create long term memories for actors. The long term memories are associated with conversation messages ( for debugging purpose).
  • We can search long term memories. We can have multiple implementations, such as text-based searcher, vector- based searcher.

All of the above interfaces, except the searcher, are implemented in mongo db. The searcher can be implemented by backends other than mongo, like elasticsearch.

Classes ​

Interfaces ​

Type Aliases ​

Functions ​