langmem/guides/memory_tools/

[giscus[bot]]

langmem/guides/memory_tools/

Build language agents as graphs

https://langchain-ai.github.io/langmem/guides/memory_tools/

[farouk09]

Hello,

when using langgraph dev, do I need to omit the InMemoryStore entirely since LangGraph handles it automatically?
Or is there a recommended way to replace or extend the behavior of InMemoryStore (for example, to provide a custom embedding model for semantic indexing)?

In my case, I was using:

store = InMemoryStore(index={"embed": embed_model})

Because I’m building an email assistant that uses semantic memory via langmem, and this requires an embedding model to be set for vector search.
So I’m wondering what’s the right approach here when working with langgraph dev — should I just omit the store completely, or is there a way to customize the built-in store's behavior (e.g. embedding config)?

[hinthornw]

Great question! It's in the config (doc) (i.e., the langgraph.json file)

{
  "<...>",
  "store": {
    "index": {
      "embed": "openai:text-embedding-3-small",
      "dims": 1536,
      "fields": ["$"]
    }
  }
}