Titan Memory
Createdhenryhawke
Empowers LLMs to maintain memory state across interactions using a neural memory system that learns and predicts sequences.
About
Titan Memory is a neural memory system designed for LLMs, enabling them to learn and predict sequences while maintaining state through a memory vector. Compatible with tools like Cursor, it provides a transformer-based memory architecture with efficient tensor operations and automatic memory cleanup. It is designed to allow LLMs to maintain a 'brain' independent of LLM version, by saving and loading memory states between sessions.
Key Features
- Text encoding for converting text inputs to tensor representations
- Compatible with MCP clients like Cursor
- Transformer-based neural memory architecture
- Memory persistence for saving and loading memory states
- Efficient tensor operations with automatic memory cleanup
- 43 GitHub stars
Use Cases
- Creating AI agents that can learn and adapt over time
- Maintaining context across multiple interactions with an LLM
- Improving the consistency and coherence of LLM responses