Add mmr reranking

[prasmussen15]

[!IMPORTANT] Add MMR reranking to search, normalize embeddings, and update community building with group ID support.

• Search Enhancements:
  • Add MMR reranking method to search.py for edges, nodes, and communities.
  • Update search_config.py and search_config_recipes.py to include MMR reranking.
  • Add maximal_marginal_relevance() function to search_utils.py.
• Embedding Normalization:
  • Use normalize_l2() in openai.py and voyage.py to normalize embeddings.
• Community Building:
  • Update build_communities() in graphiti.py and community_operations.py to accept group_ids.
• Testing:
  • Update test_graphiti_int.py to test new search configurations and community building.

^{This description was created by for 6f69a613efee88e3e2fc379933c2958ac88d8cf8. It will automatically update as commits are pushed.}

[danielchalef]

@prasmussen15 it looks like we're normalizing vectors. Do we need to run a migration on existing installs? Would it now make sense to use dot product rather then cosine as distance metric?

[prasmussen15]

@prasmussen15 it looks like we're normalizing vectors. Do we need to run a migration on existing installs? Would it now make sense to use dot product rather then cosine as distance metric?

Neo4j only offers cosine similarity and not dot product so a migration on existing installs isn't necessary for the search. We could run a migration for MMR since I use dot product there.