In-Memory Store

In-Memory Store is a long-term memory store for an agent: a place to keep data that goes beyond a single conversation, such as a customer’s preferences. It differs from a checkpointer, which keeps the thread of a conversation so it can be resumed.

With this component, the data is kept in NappAI’s temporary memory, not in a database: it is lost when the run ends or if NappAI restarts.

Component in development

This component only appears in installations with development components turned on, and for now no component in the catalogue accepts its output. To make an agent remember each user’s data across conversations, use Redis Memory or Postgres Memory with Enable Store turned on.

When to use it

• Testing long-term memory where it does not matter that the data is lost when the run ends.

Configuration

Connection	What to connect
Embedding (required)	An embeddings component, for example OpenAI Embeddings. Embeddings turn text into numbers so it can be searched by meaning.

Output

Checkpointer is the memory store, in temporary memory. It sits at the top of the component. Although the output is called Checkpointer, what it delivers is the long-term data store.

Good practices

• Don’t use it when the data must be kept: everything stored is lost when the run ends.

Related components

• In-Memory Checkpointer
• Redis Memory
• Postgres Memory
• OpenAI Embeddings