Skip to content

README_he.md

Latest commit

 

History

History
88 lines (61 loc) · 2.79 KB

README_he.md

File metadata and controls

88 lines (61 loc) · 2.79 KB
Raw

CausaMem - 永久性记忆，因果记忆系统

Causal Memory System for AI Agents

让 AI Agent 拥有一生的记忆 | Permanent Memory for Life | 一记记一生

Current reference for Cognitive Anchoring and Session Memory Refiner: English / 中文.

Design Philosophy

The goal of this project: Make AI agents remember 2 million characters.

For example, a human's lifetime memory is roughly 1 million characters. 2 million characters is sufficient for an agent's lifetime. Memory is not about accumulation—it's about how you remember. The real question is: How do you make memory work like a human's—connecting dots into lines, lines into planes, forming a traceable, reasoning-enabled causal network?

CausaMem's answer: Four-Layer Structured Memory + 13-Dimensional Causal Reasoning

Core promise: 记忆永久性存在，一记记一生 (Permanent memory, remember for a lifetime)

13-Dimensional Causal Reasoning

# Question Description
1 Why? Cause tracing
2 What then? Effect prediction
3 Full chain? Causal chains
4 Where from/to? Bidirectional reasoning
5 Causal or correlated? Relation classification
6 Necessary or probabilistic? Strong/weak cause
7 What if? Intervention thinking
8 Still valid? Temporal decay
9 Who was main cause? Multi-cause attribution
10 Which link broke? Chain breakage
11 Unexpected? Anomaly discovery
12 What did they want? Intent inference
13 What was situation? Context reconstruction

Core Features (8 Total)

  1. 13-Dimensional Causal Reasoning - Auto cause/effect inference
  2. AI Structured Compression - decided/learned/completed/next_steps/concepts/cause/effect
  3. Dual-Engine Search - Vector + FTS5 + Causal chain
  4. Wiki Four-Layer (Human-Readable) - events/timeline/relations/_dream
  5. Dream Abstraction (Cron) - Daily 02:30 + Weekly Thu 03:00
  6. Type Tag System - DECISION/INSIGHT/BUG/FEATURE/CHANGE/DAILY
  7. Multi-Agent Sharing - Filesystem based
  8. AI Emotion Module - Track AI state as causal chain node (happy/sad/hungry/sated/tired/energetic/anxious/focused/satisfied/empty) - Filesystem based

Quick Start

git clone https://github.com/MaiHHConnect/MHH-Causality-Memory.git
cd MHH-Causality-Memory
pip install requests
cd scripts/gbrain
python gbrain.py init
python gbrain.py put-structured my-event "Discussed system design"
python gbrain.py causal "system design"

Cron Setup

30 2 * * * python scripts/dream.py small
0 3 * * 4 python scripts/dream.py big

License

CC BY-NC 4.0 - Commercial use requires permission.

Contact: 3871169@qq.com

Authors

  • Vinson
  • 牛马2号 (AI Agent)