Skip to content

This repository introduces the Letta framework, empowering developers to build LLM-based agents with long-term, persistent memory and advanced reasoning capabilities. It leverages concepts from MemGPT to optimize context usage and enable multi-agent collaboration for real-world applications like research, HR, and task management.

Notifications You must be signed in to change notification settings

ksm26/LLMs-as-Operating-Systems-Agent-Memory

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 

Repository files navigation

Welcome to the "LLMs as Operating Systems: Agent Memory" course! 🧠 Learn how to build agents with long-term, persistent memory using Letta, an open-source framework for memory-enhanced LLM agents. This course is taught by Charles Packer and Sarah Wooders, co-founders of Letta, and is based on the innovative ideas presented in the MemGPT research paper.

📘 Course Summary

This course equips you with the skills to create AI agents that manage and edit memory autonomously, optimizing context usage for real-world applications like research and HR. Learn how to leverage Letta to add persistent, long-term memory to your LLM agents, enabling advanced reasoning and adaptability.

What You’ll Learn

  1. 🔄 Agent Memory Management: Build agents with self-editing memory, utilizing tool-calling and multi-step reasoning.
  2. 🛠️ Using Letta Framework: Explore Letta’s features for adding memory capabilities to LLMs, including core and archival memory.
  3. 🧩 MemGPT Concepts: Understand the key ideas behind MemGPT, including two-tier memory systems and how agent states are converted into prompts.
  4. 🤝 Multi-Agent Collaboration: Learn to implement collaborative agents by sharing memory blocks and exchanging messages.

Practical Applications

  • 🔍 Conversation Memory Control: Manage expanding conversations by summarizing and moving less relevant information to a searchable database, ensuring smooth context flow.
  • 📂 Persistent Fact Storage: Save and edit details like names, dates, and preferences for future interactions.
  • 📑 Task-Specific Memory: Develop agents capable of swapping context-relevant information in real-time from a database for tasks like research.

🔑 Key Points

  • 🧠 Enhanced Memory Management: Use Letta to create agents with long-term, persistent memory and advanced reasoning capabilities.
  • 📋 Efficient Context Optimization: Optimize LLM context window usage to reduce costs and improve processing speed.
  • 🤖 Collaboration Between Agents: Enable multi-agent systems that share memory and collaborate seamlessly.

👨‍🏫 About the Instructors

  • Charles Packer: Co-Founder of Letta and co-author of the MemGPT paper.
  • Sarah Wooders: Co-Founder of Letta and a leading expert in building memory-enhanced LLM applications.

🔗 To enroll in the course or for more details, visit 📚 deeplearning.ai.

About

This repository introduces the Letta framework, empowering developers to build LLM-based agents with long-term, persistent memory and advanced reasoning capabilities. It leverages concepts from MemGPT to optimize context usage and enable multi-agent collaboration for real-world applications like research, HR, and task management.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published