caspianmoon/memoripy

Implements graph-based spreading activation and hierarchical clustering for semantic memory retrieval across OpenAI, Azure OpenAI, OpenRouter, and Ollama integrations. The architecture uses embeddings and concept extraction to build a concept graph, enabling decay/reinforcement mechanisms that age unused memories while strengthening frequently accessed ones. Provides pluggable storage backends (JSON or in-memory) and combines cosine similarity with graph traversal for multi-dimensional context retrieval.

682 stars and 110 monthly downloads. No commits in the last 6 months. Available on PyPI.