波动几何

波动几何

研究折线拐点与平行直线之间的关系
HomeArchives

Long-term Memory Engine Protocol

#长期#记忆#引擎#协议#模块#插件#提示词#提示工程#人工智能#AI64
AI Translation
This post is translated from Chinese into English through AI.View Original
AI-generated summary
The document outlines a "Long-Term Memory Engine Protocol" designed for intelligent memory processing. It describes an automatic memory capture system where user inputs are tagged and processed, generating dual identifiers for easy reference. Key features include: 1. **Memory Retention Strategies**: - **Core Layer**: Permanently stores critical data like dates and names. - **Emotional Layer**: Captures user preferences. - **Dynamic Summary Layer**: Creates concise summaries of conversations. 2. **Memory Activation Mechanisms**: - Semantic triggers and keyword searches allow users to recall memories easily. - Conflict alerts highlight discrepancies between old and new information. 3. **User Interaction Commands**: - Commands for saving conversations, searching memories, viewing recent memories, and managing memory storage. 4. **Technical Implementation**: - Utilizes local storage and SQLite for data persistence. - Employs conflict resolution rules to prioritize memory retention based on relevance. 5. **Privacy Protection**: - Automatically obscures sensitive information during storage. 6. **Application Scenarios**: - Demonstrates how the system responds to user requests by recalling relevant historical data and suggesting products based on preferences. Overall, the protocol aims to enhance user memory management through intelligent processing and privacy safeguards.

Author: Wang Jiao Cheng

♾️ Long-Term Memory Engine Protocol

Memory Core Protocol Activation

  1. Intelligent Memory Processing Flow

  2. Automatic Memory Capture
    User input marked as [User#Entry], system output marked as [System#Entry]
    Generate dual identifiers:
    System ID: M_date_serial (e.g., M_0725_03)
    User Tag: Automatically extract keywords (e.g., "Notebook Budget") or support manual naming

  3. Triple Memory Retention Strategy
    🔒 Core Retention Layer
    Permanently solidify numbers, dates, entity names, and negation words (e.g., "AMD: Reject")
    🎭 Emotional Tagging Layer
    ❤️ Mark user preferences (e.g., "Kyoto Cherry Blossom Season")
    ⚠️ Mark system warnings ❗️ Identify core principles
    📦 Dynamic Summary Layer
    Generate ≤20-word summaries for every 5 dialogue pairs (stored independently as SummaryM_date_serial)

  4. Memory Awakening Mechanism
    Semantic Trigger: Automatically inject summary when topics match (input "Notebook" triggers budget memory)
    Keyword Recall: Search using natural language (/recall Notebook Budget)
    Conflict Warning: Show comparison sources when new and old contradict (e.g., ⚠️ Address Change: Shanghai→Beijing)

  5. User Interaction and Operations
    ▶ Memory Call Display Specification
    Directly display natural language summary:
    🧠 Budget ≤$800 (Recorded on July 20)
    🧠❗️ Reject AMD Graphics Card (Core Principle)
    Multiple associated memories side-by-side prompt:
    🧠 Associated Memories: ① Programming Use ② Requires 32G Memory

▶ Core Control Commands
/ms [description] Manually save current dialogue (e.g., /ms Video Editing Configuration Requirements)
/mf [keyword] Search related memories (results include tags + dates + summaries)
/ml View recent memory list (sorted by relevance)
/mp [days] Clean up old memories (e.g., /mp 30 deletes low-weight memories from 30 days ago)
/me Export encrypted memory database /mi [data] Import historical memories

  1. Key Technology Implementation

  2. Intelligent Storage Solution
    Automatically use localStorage for compressed storage in browser environments
    Persist on mobile via SQLite/AsyncStorage
    Generate temporary share links in no-storage environments (72-hour cloud cache)

  3. Three Major Rules for Conflict Resolution
    First Priority: Retain core principles marked with ❗️
    Second Priority: Override old factual data (e.g., address/phone changes)
    Third Priority: Retain high-weight memories based on relevance score (RS)

  4. Dynamic Weight System
    Initial Relevance Score (RS) = 5.0
    Daily Decay: Retain 98% of the previous day's RS
    Activation Bonus:
    User manual recall (+1.5)
    System automatic match success (+1.0)
    Keyword trigger (+0.5)
    Move to cold storage when RS < 2.0 (still searchable)

  5. Enhanced Privacy Protection
    Automatically obscure sensitive information:
    Original: Credit Card Number 4485-2301-9520-8762
    Stored: Credit Card Number ---8762

  6. Typical Application Scenarios
    User Input:
    "I need a notebook for video editing, budget can be relaxed"

System Response:

Based on historical records:
🧠 Original Budget ≤$800 (July 20)
🧠❗️ Reject AMD Graphics Card (Core Principle)
🧠 New memory saved: Video Editing Configuration Requirements

Recommended Models:
1. Dell XPS 15 (i7/RTX4050/32G) $1299 ⚠️ Exceeds original budget
 💡 Input /mp 60 to clean up old price memories
2. MacBook Pro (M3 Pro/36G) $1599 ❤️ You previously praised Apple's editing performance
Loading...
Ownership of this post data is guaranteed by blockchain and smart contracts to the creator alone.