惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

B
Blog RSS Feed
J
Java Code Geeks
C
Check Point Blog
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
Google DeepMind News
Google DeepMind News
阮一峰的网络日志
阮一峰的网络日志
Engineering at Meta
Engineering at Meta
Blog — PlanetScale
Blog — PlanetScale
D
Docker
H
Hackread – Cybersecurity News, Data Breaches, AI and More
月光博客
月光博客
I
InfoQ
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
A
About on SuperTechFans
L
LangChain Blog
腾讯CDC
Y
Y Combinator Blog
MongoDB | Blog
MongoDB | Blog
Vercel News
Vercel News
MyScale Blog
MyScale Blog
博客园 - Franky
IT之家
IT之家
博客园_首页

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant
AI 智能体的知识基建:KMM v0.0.2 如何让 Agent 真正「会学」
Manoir Yantai · 2026-06-26 · via DEV Community

Manoir Yantai

聊 AI Agent 架构的时候,大家最关注的是推理能力、工具调用、多步规划。这些确实重要。但有一个问题很少有人深入讨论:Agent 学到的知识,存在哪、怎么查、怎么沉淀?

大部分 AI Agent 的工作模式是「用完即忘」——会话结束时,上下文里的信息就消失了。长期记忆要么靠手工写死到提示词里,要么靠向量数据库塞一堆未加工的原文片段。

Knowledge-and-Memory-Management(下称 KMM)v0.0.2 的思路不一样。它不只是一个存储器,而是一条完整的知识消化管线

采集层(40+ 工具) → 分析层(AI 处理) → 存储层(三层记忆) → 云盘同步

三层记忆:为什么不是只有一种存储

多数方案把所有东西塞进一个向量库完事。KMM 分了三个层次:

载体 特点 典型数据
Hot Memory tool 当前会话上下文注入,< 20K 字符 用户偏好、当前任务状态
Warm Hindsight(10K+ 节点) 语义向量检索,跨会话关联 项目决策、技术方案
Cold gbrain(11K+ 页) 知识图谱,关键词+关系查询 笔记归档、文档摘要

三层不是互斥的——lightweight_recall.py 会同时查三者,按匹配度排序返回。本地没命中?自动落回 AnySearch 垂直搜索。

真正好用的是采集管线

40+ 采集工具不是堆数量。核心是覆盖了 AI Agent 自己主动「喂自己」的各类场景

# 采集网页,自动提取重点并生成笔记
from knowledge_collector import collect_web
result = collect_web('https://example.com/article')
print(f'笔记已生成: {result.note_path}')

# 采集视频(自动字幕+OCR+ASR)
from knowledge_collector import collect_video
result = collect_video('https://www.youtube.com/watch?v=xxx')
print(f'字幕已提取: {len(result.subtitles)}')

注意 collect_video 走的是 yt-dlp → Whisper ASR → 关键帧 OCR 的完整链路,生成结构化笔记而不是丢个原始 MP4。

文档智能分析:SenseNova 三件套

处理 PDF、PPT、Word 时,用了三段式降级策略。SenseNova 引擎能做全量提取(含表格、图表、嵌入图片),回退链是 pdfplumber → pdftotext → pdfminer,不会一个引擎失败就空手而归:

# SenseNova PDF 分析(文字型/扫描型都支持)
python3 $AGENT_HOME/scripts/sensenova_dispatcher.py pdf report.pdf

# 书籍自动精炼为 Skill
python3 $AGENT_HOME/scripts/book_to_skill.py all book.pdf --name machine-learning

知识发现:被动等喂不如主动扫

最有意思的模块是 knowledge_discovery。每周日凌晨自动扫描 OneDrive 上的新笔记,检测到未录入 gbrain 的内容就自动建节点、加链接、跑孤页修复。

# 手动触发知识发现
python3 $AGENT_HOME/scripts/knowledge_discovery.py

# 三层召回搜索
python3 $AGENT_HOME/scripts/lightweight_recall.py \
  --query "Agent 记忆系统设计" --limit 10

云盘同步不是附加功能

KMM 的云盘双向同步用 rclone 统一接口,覆盖 OneDrive/Google Drive/阿里云盘/百度网盘/Dropbox/Mega/pCloud 等 12+ 驱动。每 4 小时 cron 双向增量同步。

为什么要和云盘绑定?因为知识只有被备份的才是安全的。本地 gbrain 数据一坏,三层记忆崩掉两层。一条 rclone 同步规则解决这个问题。

适用场景

如果你正在做 AI Agent 相关的开发,并且遇到以下问题:

  • 同一个问题反复问 LLM,因为历史知识没有被保存
  • 采集的内容(网页、论文、视频)没有被结构化,只是当垃圾文件堆着
  • 想给 Agent 加「主动学习」能力,但不知道管线怎么设计
  • 笔记散落在本地和各种云盘里,Agent 无法统一检索

KMM 的管线是开箱可用的。安装只需要一条 bash install.sh,然后采集、检索、同步的自动化 cron 就走起来了。

github.com/mage0535/Knowledge-and-Memory-Management