# Awesome Agent Memory — Long-Term Context Index

> Awesome Agent Memory curates systems, benchmarks, and papers on long-term context for LLMs/MLLMs—use it to compare approaches and pick tools to try.

## Install

Copy the content below into your project:

## Quick Use

```bash
git clone https://github.com/TeleAI-UAGI/Awesome-Agent-Memory.git
cd Awesome-Agent-Memory
```

Open the README, then pick:
1) one benchmark, 2) one production system, 3) one open-source implementation to prototype.

## Intro

Awesome Agent Memory curates systems, benchmarks, and papers on long-term context for LLMs/MLLMs—use it to compare approaches and pick tools to try.

- **Best for:** engineers doing memory design/selection for coding agents and long-running assistants
- **Works with:** GitHub reading + your preferred papers/tools stack; use it as an index, not a framework
- **Setup time:** 5–15 minutes

## Practical Notes

- Organized into products, tutorials, surveys, benchmarks, and paper sections (see README table of contents).
- Use one benchmark to define your acceptance bar (latency, recall, token budget), then pick an approach.
- Keep a “memory regression set”: 20–50 queries that used to work, to catch drift when you change memory policy.

## Main

A selection workflow that actually works:

1. **Define what “memory” means for your agent**: project facts, user preferences, tool state, or long transcripts.
2. Decide your constraint triangle: **latency**, **privacy**, **token budget**.
3. Pick a baseline approach (summaries + retrieval, vector store, graph/wiki, or hybrid).
4. Evaluate on one benchmark + your own domain tasks, then iterate.

The key is avoiding “infinite context”. Good memory systems are selective: they store *high-signal* facts and can justify why a memory was retrieved.

### FAQ

**Q: Is vector search enough?**
A: Sometimes. For coding agents, you often need hybrid memory: durable facts + searchable artifacts + updated summaries.

**Q: What’s the first metric to watch?**
A: Retrieval precision: how often retrieved items actually help the answer. Low precision is the fastest way to waste tokens.

**Q: How do I prevent stale memory?**
A: Attach timestamps and sources; re-validate critical facts periodically and prune memories that don’t get used.

## Source & Thanks

> Source: https://github.com/TeleAI-UAGI/Awesome-Agent-Memory
> License: Apache-2.0
> GitHub stars: 407 · forks: 28

---

<!-- ZH -->

## 快速使用

```bash
git clone https://github.com/TeleAI-UAGI/Awesome-Agent-Memory.git
cd Awesome-Agent-Memory
```

打开 README 后建议先各选一个：1）benchmark，2）可落地系统，3）开源实现做原型。

## 简介

Awesome Agent Memory 汇总 agent 长期记忆的系统、基准与论文（含可复现开源实现标注与排序），用来快速了解记忆机制、检索策略与评测指标，并跟踪最新实践与新闻，适合做选型清单。

- **适合谁：** 在做 coding agent / 长会话助手“记忆选型”的工程团队
- **可搭配：** GitHub 阅读 + 你的论文/工具栈；把它当索引而不是框架本身
- **准备时间：** 5–15 分钟

## 实战建议

- 按 products/tutorials/surveys/benchmarks/papers 分区（见 README 目录）。
- 先用一个 benchmark 定义门槛（延迟、召回、token 预算），再做方案选择。
- 维护一份“记忆回归集”：20–50 条历史问题，用于检测记忆策略改动后的漂移。

## 主要内容

一个更“工程化”的选型流程：

1. **先定义你的记忆对象**：项目事实、用户偏好、工具状态，还是长转写。
2. 明确三角约束：**延迟**、**隐私**、**token 预算**。
3. 选一个 baseline（摘要+检索、向量库、图谱/wiki、或混合）。
4. 用一个 benchmark + 你的业务任务评测，再迭代。

重点是避免“无限上下文”。好的记忆系统是选择性的：只存高信号事实，并能解释为什么这条记忆会被取回。

### FAQ

**只有向量检索够吗？**
答：不一定。对 coding agent 通常需要混合：可继承的事实 + 可搜索的产物 + 可更新的摘要。

**最先看哪个指标？**
答：召回精度：取回的内容有多大比例真的有用。精度低会最快烧 token。

**怎么避免记忆过期？**
答：给记忆加时间戳与来源；定期复核关键事实，并清理长期不被使用的条目。

## 来源与感谢

> Source: https://github.com/TeleAI-UAGI/Awesome-Agent-Memory
> License: Apache-2.0
> GitHub stars: 407 · forks: 28


---
Source: https://tokrepo.com/en/workflows/awesome-agent-memory-long-term-context-index
Author: AI Open Source