What is Memoripy
Memoripy is a Python library that provides context-aware memory management for AI applications. Memoripy supports short-term and long-term memory storage and is compatible with the OpenAI and Ollama APIs. Core features include memory retrieval, concept extraction, graph-based association, hierarchical clustering, and memory decay and reinforcement. Based on these features, Memoripy helps AI systems better understand and respond to user interactions, improving the coherence and personalized experience of conversations.
Main features of Memoripy
- Short-term and long-term memory management: Memoripy divides memories into short-term and long-term based on usage frequency and relevance, optimizing information storage and retrieval.
- Context retrieval: By analyzing embedding vectors, concepts, and historical interactions, Memoripy retrieves the past memories most relevant to the current interaction.
- Concept extraction and embedding generation: Based on the OpenAI and Ollama models, it extracts key concepts and generates corresponding embedding vectors to facilitate memory comparison and retrieval.
- Graph-based association: It builds a concept graph and implements relevance-based memory retrieval through a spreading activation mechanism.
- Hierarchical clustering: It clusters similar memories based on semantic similarity to facilitate context-related retrieval.
- Memory decay and reinforcement: Memoripy implements dynamic memory management, allowing infrequently accessed memories to gradually decay while frequently accessed memories are reinforced.
Technical principles of Memoripy
- Memory storage: Memoripy uses two storage methods, in-memory storage and JSON file storage, to save interaction data.
- Embedding technology: Using natural language processing technology, Memoripy converts text into embedding vectors. The vectors capture the semantic information of the text, making comparison and retrieval easier.
- Similarity retrieval: Using algorithms such as cosine similarity, it finds the historical memories most similar to the current interaction based on embedding vectors.
- Concept graph: It builds a graph in which nodes represent concepts and edges represent relationships between concepts, enabling retrieval of complex relationships based on graph algorithms.
- Clustering algorithms: It uses clustering techniques to group similar memories, improving retrieval efficiency and relevance.
Project address of Memoripy
- GitHub repository：https://github.com/caspianmoon/memoripy
Application scenarios of Memoripy
- Chatbots: In customer service and online chat applications, it helps chatbots remember users' historical interactions and provide a more personalized and coherent conversation experience.
- Virtual assistants: In personal assistant applications, it stores users' preferences and habits, allowing virtual assistants to provide more thoughtful services based on users' historical behavior.
- Educational software: In educational applications, it tracks students' learning progress and preferences and provides customized learning content and suggestions.
- Recommendation systems: In e-commerce or content recommendation platforms, it analyzes users' historical interactions and preferences to provide more precise personalized recommendations.
- Health monitoring: In health and fitness applications, it records users' activity and health data, helping users track progress and adjust health plans.