Quantum Memory Graph

该 Skill 把对话按 500 字符切块，每块重叠 100 字符，用 gte-large 生成嵌入，再以会话内相似度最高的 3 个分块的平均余弦相似度打分；它可选进行 QAOA+CVaR 子图优化，并先查询个人图谱，相关性不足时回退到历史归档。它可以作为 Python 库调用，也可以作为 FastAPI 服务提供存储、召回与统计接口。性能数据由该 Skill 自述。

dustin-a11y

热度 62.41

最近更新 2026-05-30 09:51:04

分类：知识依赖：代码执行应用场景：笔记整理 / 研究 / 数据处理 / 推荐

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add quantum-memory

Metadata

name: quantum-memory-graph
title: Quantum Memory Graph
description: Quantum-enhanced long-term memory for AI agents — #1 on LongMemEval (98.6% R@5, 99.4% R@10, 0.9426 NDCG). Chunked gte-large retrieval with QAOA+CVaR subgraph optimization for agents.
tags: ["memory","retrieval","quantum","rag","embeddings","long-memory","AI-agents","agent-memory"]

Quantum Memory Graph (QMG)

Quantum-enhanced long-term memory for AI agents. Uses chunked gte-large embeddings for state-of-the-art semantic retrieval with QAOA+CVaR quantum subgraph optimization for graph-based reasoning.

Features

#1 on LongMemEval — 98.6% R@5, 99.4% R@10, 0.9426 NDCG
Chunked retrieval — 500-char blocks with 100-char overlap, mean-of-top-3 per session
QAOA+CVaR optimization — 12.8% edge over greedy on graph/PCE tasks
GPU accelerated — runs on NVIDIA GB10 (DGX Spark)
Cascade recall — personal graph → historical archive fallback
Per-agent isolation — each agent gets their own isolated memory graph

How It Works

Session chunking — Conversations split into overlapping 500-char chunks
Embedding — Chunks encoded with gte-large (1024-dim sentence transformer)
Scoring — Per-session score = mean of top-3 chunk cosine similarities
Refinement — Top-N candidates optionally refined via QAOA+CVaR subgraph optimizer
Cascade — Personal graph first, historical archive fallback if relevance < 0.4

Performance

Metric	Score
R@1	90.6%
R@5	98.6%
R@10	99.4%
NDCG@10	0.9426

Benchmark: LongMemEval-S (500 questions, 18,464 sessions), May 28 2026

Usage

from quantum_memory_graph import MemoryGraph

mg = MemoryGraph()
mg.store("Project Alpha uses React frontend with TypeScript.")
mg.store("Project Alpha backend is FastAPI with PostgreSQL.")

# Recall — chunked semantic retrieval + optional QAOA refinement
results = mg.retrieve("What is Project Alpha's tech stack?", top_k=5)

Or as a FastAPI server for agent integration:

pip install quantum-memory-graph
quantum-memory-graph serve  # serves /store, /recall, /stats endpoints

Requirements

Python 3.10+
sentence-transformers
numpy
(optional) qiskit for QAOA optimization on real hardware

License

MIT