TAPE

Introduction

This repository contains the code used in our paper: TAPE: A Temporal Graph-based Memory System for Personal LLM Agents. TAPE is a temporal graph-based memory system designed to serve as the memory backbone for personal LLM agents.

The demo link for TAPE is https://youtu.be/ndmagVTMzDw

Running Environment

a 64-bit Linux-based OS; a MacOS

Usage

How to run

• Step 1: Push the demo-backend part code to the linux server

• Step 2: Move to directory /back-end and run the python file, through which the data interaction between the front-end interface and the back-end TAPE is realized

  python app.py

• Step 3: Open the front-end interface /front-end/templates/index.html at MacOS or Windows to use TAPE

Requirements

• cmake
• g++
• OpenMP
• Boost
• flask
• openai
• tiktoken
• torch==2.5.1+cu121
• faiss-cpu==1.10.0
• pyg_lib