momex 0.2.2

Memory system for AI agents with Structured RAG, built on TypeAgent

Fork Changes

I've been keeping an eye on this repo for a while. I'm also personally interested in personal assistant agents, and I've been trying to find best practices for memory. Structured RAG is a great design, but since this is an experimental project, the feature set isn't complete yet. So I forked the original repo and added more features, aiming to make it work for more general use cases and projects.

This fork adds Momex - a high-level memory API for AI agents, built on TypeAgent's Structured RAG.

What's New

• Hybrid search - Structured RAG + embedding similarity in parallel for best recall
• Structured memory - Entity/action/topic extraction via LLM, term-based indexing
• Embedding-only mode - search_by_embedding() for fast similarity search without LLM
• Multi-tenant support with hierarchical collections (user:xiaoyuzhang)
• PostgreSQL backend with pgvector for production deployment

Installation

pip install momex

Quick Start

import asyncio
from momex import Memory, MomexConfig, LLMConfig

async def main():
    config = MomexConfig(
        llm=LLMConfig(provider="openai", model="gpt-4o", api_key="sk-xxx"),
    )

    # Create a memory collection
    memory = Memory(collection="user:xiaoyuzhang", config=config)

    # Add memories — automatically extracts entities, actions, topics
    await memory.add("My name is Xiaoyu, I love Python programming")

    # Search — returns structured results you can feed to your own agent
    results = await memory.search("Python")
    for item in results:
        print(f"[{item.type}] {item.text}")

asyncio.run(main())

See docs/momex.md for full documentation.

---

Python package 'typeagent'

This is an experimental prototype

Working toward a shared understanding of the MVP for structured RAG.

This is sample code

This is an in-progress project aiming at a Pythonic translation of TypeAgent KnowPro and a few related packages from TypeScript to Python.

Warning

This library will send its input to an LLM hosted by a third party. Don't use it to index confidential information.

Documentation

• Found in the docs directory
• Quick install: pip install typeagent
• Download the PyBay '25 PowerPoint slides
• Download the PyBay '25 slides as PDF
• Watch the PyBay '25 video

Trademarks

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.