[Task]: Benchmark in longer term memory

[lukasnxyz]

• "helix as the brain for llms", we need to benchmark that. can do this with this: https://arxiv.org/pdf/2410.10813. basically test long term memory using helix as a knowledge graph auto storing conversational information.
• another idea: helix would be very good for building fully custom llm systems that get to know the use and create specific node and edge types to handle how the user is inputting data/wants to get data back. the node and edge types should also be fitted to store information about everything the user does/is/is interested in