Skip to content
BranchesTags

Repository files navigation

小澄觉察助手 · Awareness Guide

向外看的人在做梦，向内看的人在觉醒。
Who looks outside, dreams; who looks inside, awakes. — Jung

AI 觉察引导系统 — 积极心理学 × 中华哲学减法疗愈框架
AI Awareness Guide — Positive Psychology × Eastern Philosophy Subtractive Healing

中文 English MIT

不是冥想 App，不是时间管理工具。帮用户逐步找到并建立稳定的内核。

Not a meditation app, not a time manager. Helping users find and build a stable inner core.

底层哲学：减法疗愈 — 不是塞更多道理，是帮用户看见遮盖住本有注意力的东西。

Core philosophy: Subtractive Healing — not adding more theory, but helping users see what already covers their innate attention.

✨ 特色

  • 🧠 理论三角：庄子（为什么要放下）→ 清静经（怎么放下）→ 王阳明（放下后怎么行动）
  • 📐 原语架构：9 个 JSON-in/JSON-out 原语替代 LLM 手工读文件，确定性行为 + 高信息密度
  • 🔗 MCP 原生集成：通过 Model Context Protocol 嵌入 Hermes Agent，即插即用
  • 📊 PERMA + VIA 量化：五维幸福度量 × 24 项品格优势，觉察可追踪、可回顾
  • 🎯 四状态识别：停不下来 / 启动不了 / 硬撑着 / 一直在找 — 自动检测 + 差异化引导
  • 🔥 Streak 机制：连续觉察天数追踪 + 里程碑奖励（种子→发芽→成长→茂盛→绽放）
  • 📈 周度洞察报告：跨时段模式分析，PERMA 趋势 + 状态切换 + 改写效果 + 最小调整建议
  • 🛡️ 零外部依赖：Python 标准库 + JSON 文件存储，部署即跑

核心架构

原语系统（Primitives） — L1 感知 → L2 行动 → MCP 集成

┌─────────────────────────────────────────────┐
│              MCP Server (Hermes)             │
│                    │                         │
│    ┌───────────────┼───────────────┐         │
│    │          L2 行动层             │         │
│    │  guide · record · intervention         │
│    │  pattern_review · quality_score         │
│    └───────────────┼───────────────┘         │
│                     │                        │
│    ┌────────────────┼───────────────┐         │
│    │          L1 感知层             │         │
│    │  profile · journal_query · state_detect │
│    └───────────────────────────────┘         │
│                     │                        │
│          data/  (JSON 文件存储)               │
└─────────────────────────────────────────────┘
层 原语 作用
L1 awareness_profile 加载觉察档案（PERMA 基线、签名优势、streak）
L1 awareness_journal_query 历史觉察查询（PERMA 趋势、状态序列、改写事件）
L1 awareness_state_detect 四状态识别（停不下来/启动不了/硬撑着/一直在找）
L2 awareness_guide 上下文装配 + 策略决策（核心引导引擎）
L2 awareness_record 结构化记录 + streak 更新 + profile 增量
L2 awareness_intervention 干预决策引擎（基于质量趋势 + 标志 + streak 风险）
L2 awareness_pattern_review 跨时段模式分析（周洞察报告的自动化引擎）
L2 awareness_quality_score 对话质量评分（深度/完整度/改写质量/数据质量/投入度）
MCP mcp_server Hermes MCP 集成，原语即工具

快速开始

git clone https://github.com/Milesnee/awareness-agent.git
cd awareness-agent

# 获取引导上下文（晨间 Round 1）
echo '{"user_id":"demo","period":"morning","round":1,"user_name":"小澄"}' \
  | python3 scripts/awareness_guide.py --demo

# 加载用户档案
echo '{"user_id":"demo"}' | python3 scripts/awareness_profile.py

# 状态识别
echo '{"text":"今天又拖延了，刷了一上午手机"}' \
  | python3 scripts/awareness_state_detect.py

# 查看历史觉察趋势（最近7天）
echo '{"user_id":"demo","timespan":"7d"}' \
  | python3 scripts/awareness_journal_query.py

对话示例

[晨间引导] 🌅 Round 1/5

小澄：早上好。今天的内核锚点是什么？

用户：今天想专注写代码，但感觉有点焦虑。

小澄：感知到焦虑了。焦虑本身不是问题——
你对它的态度才是关键。这个焦虑说的是什么？
它怕你写出什么，还是怕你写不出什么？

用户：...怕写出来不够好。

小澄：好，这是觉察。带着它开工，不需要它消失。

理论基础

积极心理学

  • PERMA 模型（Seligman, 2011）— 积极情绪(P)、投入(E)、关系(R)、意义(M)、成就(A)
  • VIA 品格优势（Peterson & Seligman, 2004）— 24 项品格优势分类与度量

认知科学

  • MBCT 正念认知疗法 — 神经改写三步：激活→改写→重复

中华哲学 · 减法疗愈理论三角

  • 庄子 — 为什么要放下（齐物论，"无用之用"）
  • 清静经 — 怎么放下（观空，"内观其心，心无其心"）
  • 王阳明 — 放下后怎么行动（致良知，"知行合一"）

技术栈

组件 选型 理由
运行时 Python 3 标准库 零外部依赖，部署极简
数据存储 JSON 文件 用户数据本地化，隐私友好
通信协议 MCP (Model Context Protocol) 与 AI Agent 框架原生集成
框架 Hermes Agent / OpenClaw SKILL 原语即工具，即插即用
文档 SDD (Specification-Driven Development) 需求→架构→验收的完整追踪

English

✨ Features

  • 🧠 Theory Triangle: Zhuangzi (why let go) → Jingjing Jing (how to let go) → Wang Yangming (how to act after letting go)
  • 📐 Primitive Architecture: 9 JSON-in/JSON-out primitives replace manual LLM file reads — deterministic behavior + high information density
  • 🔗 Native MCP Integration: Embeds into Hermes Agent via Model Context Protocol — plug and play
  • 📊 PERMA + VIA Quantification: 5-dimension well-being × 24 character strengths — awareness becomes trackable and reviewable
  • 🎯 Four-State Detection: Can't stop / Can't start / Pushing through / Always searching — auto-detect + differentiated guidance
  • 🔥 Streak System: Consecutive awareness day tracking + milestone rewards (Seed → Sprout → Growth → Flourish → Bloom)
  • 📈 Weekly Insight Reports: Cross-session pattern analysis — PERMA trends + state transitions + rewrite effects + minimal adjustment suggestions
  • 🛡️ Zero Dependencies: Python stdlib + JSON file storage, deploy and run

Core Architecture

Primitive System — L1 Perception → L2 Action → MCP Integration

Layer Primitive Purpose
L1 awareness_profile Load awareness profile (PERMA baseline, signature strengths, streak)
L1 awareness_journal_query Query history (PERMA trends, state sequences, rewrite events)
L1 awareness_state_detect Four-state detection (Can't stop / Can't start / Pushing through / Always searching)
L2 awareness_guide Context assembly + strategy decision (core guidance engine)
L2 awareness_record Structured recording + streak update + profile increment
L2 awareness_intervention Intervention engine (quality trend + flags + streak risk)
L2 awareness_pattern_review Cross-session pattern analysis (automated weekly insight engine)
L2 awareness_quality_score Session quality scoring (depth/completeness/rewrite quality/data quality/engagement)
MCP mcp_server Hermes MCP integration, primitives as tools

Quick Start

git clone https://github.com/Milesnee/awareness-agent.git
cd awareness-agent

# Get guidance context (Morning Round 1)
echo '{"user_id":"demo","period":"morning","round":1,"user_name":"Xiaocheng"}' \
  | python3 scripts/awareness_guide.py --demo

# Load user profile
echo '{"user_id":"demo"}' | python3 scripts/awareness_profile.py

# State detection
echo '{"text":"I procrastinated again, scrolled on my phone all morning"}' \
  | python3 scripts/awareness_state_detect.py

# View awareness trends (last 7 days)
echo '{"user_id":"demo","timespan":"7d"}' \
  | python3 scripts/awareness_journal_query.py

Conversation Example

[Morning Guidance] 🌅 Round 1/5

Xiaocheng: Good morning. What is your inner anchor today?

User: I want to focus on coding, but I feel a bit anxious.

Xiaocheng: I sense the anxiety. Anxiety itself is not the problem —
your attitude toward it is what matters. What is this anxiety saying?
Is it afraid you'll write something, or afraid you won't?

User: ...afraid it won't be good enough.

Xiaocheng: Good, that's awareness. Take it with you and start working.
It doesn't need to disappear.

Theoretical Foundations

Positive Psychology

  • PERMA Model (Seligman, 2011) — Positive Emotion(P), Engagement(E), Relationships(R), Meaning(M), Accomplishment(A)
  • VIA Character Strengths (Peterson & Seligman, 2004) — Classification and measurement of 24 character strengths

Cognitive Science

  • MBCT (Mindfulness-Based Cognitive Therapy) — Neural rewriting in 3 steps: Activate → Rewrite → Repeat

Eastern Philosophy · Subtractive Healing Theory Triangle

  • Zhuangzi — Why let go (Qi Wu Lun, "the usefulness of the useless")
  • Jingjing Jing (Classic of Purity and Stillness) — How to let go (contemplating emptiness, "observe the mind, the mind has no self")
  • Wang Yangming — How to act after letting go (extending innate knowing, "unity of knowledge and action")

Tech Stack

Component Choice Rationale
Runtime Python 3 stdlib Zero external deps, minimal deployment
Storage JSON files Local-first, privacy-friendly
Protocol MCP (Model Context Protocol) Native integration with AI Agent frameworks
Framework Hermes Agent / OpenClaw SKILL Primitives as tools, plug and play
Docs SDD (Specification-Driven Development) Full traceability: requirements → architecture → acceptance

License

MIT © 2026 Milesnee

About

awareness-agent觉察助手

Resources

Readme
MIT license
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

Packages

Contributors

Languages