Memoripy is a Python library for AI applications that require sophisticated memory management, offering short-term and long-term storage with features like semantic clustering and memory decay. It's designed for developers building context-aware AI agents and chatbots, enabling more natural and persistent conversational experiences.

How It Works

Memoripy manages memory through a MemoryManager class, leveraging separate short-term and long-term storage mechanisms. It uses embedding models (like OpenAI or Ollama) to represent interactions semantically. Retrieval is enhanced by hierarchical clustering and graph-based associations, allowing for concept extraction and spreading activation to find relevant memories. Memory decay and reinforcement mechanisms dynamically adjust the relevance of past interactions over time.

Quick Start & Requirements

• Install via pip: pip install memoripy
• Prerequisites: openai, faiss-cpu, numpy, networkx, scikit-learn, langchain, ollama.
• Requires API keys for OpenAI or Ollama.
• Example usage and detailed class descriptions are available in the README.

Highlighted Details

• Supports multiple LLM providers: OpenAI, Azure OpenAI, OpenRouter, and Ollama.
• Implements memory decay and reinforcement for dynamic relevance.
• Utilizes graph-based associations and spreading activation for retrieval.
• Offers hierarchical clustering for semantic grouping of memories.

Maintenance & Community

• Open to contributions via issues and pull requests.
• No specific community channels (Discord/Slack) or notable contributors are listed in the README.

Licensing & Compatibility

• Licensed under the Apache 2.0 License.
• Permissive license suitable for commercial use and integration into closed-source applications.

Limitations & Caveats

The provided example requires manual API key setup. While faiss-cpu is listed, performance with large memory stores might benefit from a GPU-enabled FAISS installation, which is not explicitly detailed. The project appears to be actively maintained with contributions welcome, but specific community support channels are not highlighted.