DOC: Unclear relation between Mem0 and Open Memory MCP / Imports

[janus-reith]

Issue with current documentation:

The examples for Mem0 showcase adding whole conversations to Mem0. With Open Memory MCP, a tool is provided that an external LLM uses to add individual pieces of information to memory, determining on its own which details are relevant.

Now with long dumps of existing conversations, I'm wondering what the best approach is to get the most information out of it.

Batch-process them with an LLM that extracts relevant memories and uses the MCP (That seems a bit redundant since Mem0 would utilize an LLM to create relevant data aswell.

Or directly ingest all information using Mem0, skipping the MCP in the process?

[murtaza-motorwala]

In my understanding, it would be a part of your workflow/prompt. After you have instructed your agent with the task, you can probably add a small line towards the end saying - Summarise and store this in OpenMemory. And that should avoid OpenMemory making redundant calls to summarise it further. My 2 cents.