Required prerequisites

• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

Motivation

We have supported GraphRAG. However, knowledge graphs are not used in AgentMemory yet. We many want to introduce GraphDBBlock to LongtermAgentMemory and add GraphDBMemory.

https://github.com/camel-ai/camel/blob/master/camel/memories/agent_memories.py

An example on how to use Neo4jGraph for GraphRAG here:

https://colab.research.google.com/drive/1meBf9w8KzZvQdQU2I1bCyOg9ehoGDK1u#scrollTo=l4VANToobUzu

A couple things to be discussed:

• What is the default structure of the LongtermAgentMemory? Should we always have both VectorDBBlock and GraphDBBlock or allow different combinations?
• Should we allow user to load a pre-built knowledge graph?
• Should we allow agent to automatically convert chat messages into knowledge graphs?
• How to score, retrieve and decide whether to use the retrieved results?
• How to combine to retrieved KGs with vector_db_retrieve and chat_history?

Solution

No response

Alternatives

No response

Additional context

No response