AB
AB
AiBoss

All Categories

project

MemoryScope - A long-term memory system for LLM chatbots

MemoryScope is a long-term memory system designed for chatbots with large language models (LLMs). By building a framework, it enables the chatbot to remember basic user information, habits, and preferences, providing a personalized interactive experience...

What is MemoryScope?

MemoryScope is a long-term memory system designed for chatbots with large language models (LLMs). By building a framework, it enables the chatbot to remember basic user information, habits, and preferences, providing a personalized interactive experience. MemoryScope features a memory database, a core worker library, and a core operation library, supporting functions such as memory retrieval, memory consolidation, reflection, and reconsolidation. MemoryScope stores memory fragments through a vector database and handles memory-related tasks based on multiple workers. MemoryScope supports time awareness, providing relevant memories based on temporal context. This enhances the chatbot's coherence and "understanding" capabilities, allowing users to experience a sense of "chemistry" during use.

MemoryScope's main functions

  • Memory database: Store memory fragments based on vector databases (such as ElasticSearch).
  • Core Worker LibraryLong-term memory capacity is broken down into multiple independent workers, each responsible for different tasks, such as information retrieval and filtering, observation and extraction, and insight updates.
  • Core Operation LibraryBased on worker pipeline construction, it realizes core capabilities such as memory retrieval and memory consolidation.
  • Memory retrievalReturns semantically relevant memory fragments based on user input. If the input includes time information, it will return memory fragments for the corresponding time.
  • Memory consolidationProcess user input, extract important information, and store it as memory fragments in the form of observations.
  • Reflection and ReconsolidationRegularly reflect on newly recorded observations, form and update memory fragments in the form of insights, and perform memory reconsolidation, addressing contradictions and repetitions between memory fragments.

MemoryScope's technical principles

  • Vector Database: Store memory fragments using a vector database and perform efficient retrieval based on vector similarity.
  • Atomized WorkerThe system breaks down long-term memory capacity into multiple independent workers, each responsible for specific tasks such as information filtering and observation retrieval, thereby improving the system's flexibility and scalability.
  • Pipeline ArchitectureIt organizes workers based on a pipeline architecture to implement operations such as memory retrieval and memory consolidation, thereby improving processing efficiency.
  • Semantic relevanceIt uses natural language processing technology to analyze the semantics of user input and return relevant memory fragments.
  • Time context processingThe system can understand the temporal context and retrieve memory fragments based on time information.
  • Hierarchical structure of memory fragmentsMemory fragments are stored in a hierarchical structure, and insights, as high-level information, are aggregated from observations on similar topics.

MemoryScope project address

Application scenarios of MemoryScope

  • Personal AssistantAs a personal assistant, MemoryScope helps the robot remember users' schedules, preferences, and frequently used contacts, providing more personalized services.
  • Emotional companionshipIn emotional companion robots, MemoryScope records users' emotional changes and communication history, enabling the robot to better understand and respond to users' emotional needs.
  • Customer ServiceIn the field of customer service, MemoryScope helps robots remember customers' historical interaction records and preferences, providing more personalized services.
  • Educational guidanceIn educational tutoring robots, MemoryScope records students' learning progress and preferences, providing customized learning suggestions and tutoring.
  • Health ConsultationIn the health consultation robot, MemoryScope records the user's health status and lifestyle habits, and provides personalized health advice.
Back