Skip to content

A framework for extending memory capabilities of large language models like OpenAI's GPT-4 through efficient encoding, external storage, indexing, recall, and decoding mechanisms.

License

Notifications You must be signed in to change notification settings

tang-ji/LLM-Extended-Memory

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

LLM Extended Memory Framework

The LLM Extended Memory Framework is an open-source project designed to enhance the memory capabilities of large language models like OpenAI's GPT-series. By implementing an efficient encoding, storage, indexing, recall, and decoding system, this project aims to overcome the memory limitations associated with typical input+context character limits.

Feature Implementation Progress Tracker

  • LLM API Calling Tools : Tools for calling LLM APIs to get and decode responses for user inputs
  • Memory Encoder : Compact semantic representations of conversation and context
  • External Memory Storage: Support for storing data on various storage media (local files, databases, etc.)
  • Indexing System: Efficient retrieval of required information based on keyword, topic, or semantic searches
  • Memory Recall: Seamless retrieval and reintegration of essential data into ongoing tasks or conversations
  • Memory Decoder: Reconstruction of original data from their encoded representations for LLM model compatibility
  • Integration with LLM: Working in conjunction with LLM models for improved performance in complex tasks and long conversations

Running Tests

To run tests, execute the following command from the project directory:

python llm_extended_memory/run_tests.py

About

A framework for extending memory capabilities of large language models like OpenAI's GPT-4 through efficient encoding, external storage, indexing, recall, and decoding mechanisms.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages