Skip to content

Mnemosyne OS — Combining GraphRAG-style retrieval with temporal memory distillation #2415

Description

@gymaira1990-jpg

Mnemosyne OS — AI agent memory system combining GraphRAG + temporal distillation

Hi GraphRAG team! We've built a memory OS that combines graph-based retrieval with temporal memory distillation — thought you might find the crossover interesting.

Our approach

Mnemosyne OS v5.2 uses Apache AGE (Cypher graph engine) for knowledge graph queries on agent memories, alongside pgvector for semantic search. The hybrid approach gives us:

  • Graph queries for entity relationships (Cypher MERGE/CREATE/MATCH)
  • Vector search for semantic similarity
  • 5-level temporal distillation (Fragment→Session→Daily→Weekly→Profile)
  • Smart RAG chunking with overlap windows

Why it's relevant

GraphRAG excels at extracting structured knowledge from unstructured text. We do the same but layer it with temporal memory hierarchy — knowledge doesn't just sit in a graph, it evolves through distillation levels.

Links

Curious about your thoughts on combining graph-based retrieval with temporal memory structures!

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions