Triple Memory System with Baidu Embedding

A comprehensive memory architecture combining three complementary systems for maximum context retention across sessions, with full privacy protection using Baidu Embedding technology.

📋 Original Source & Modifications

Original Source: Triple Memory (by Clawdbot Team) Modified By: [Your Clawdbot Instance] Modifications: Replaced LanceDB with Baidu Embedding DB for enhanced privacy and Chinese language support

Original Triple Memory SKILL.md was adapted to create this version that: - Replaces OpenAI-dependent LanceDB with Baidu Embedding DB - Maintains the same three-tier architecture - Preserves Git-Notes integration - Adds privacy-focused local storage

🏗️ Architecture Overview

User Message
     ↓
[Baidu Embedding auto-recall] → injects relevant conversation memories
     ↓
Agent responds (using all 3 systems)
     ↓
[Baidu Embedding auto-capture] → stores preferences/decisions automatically
     ↓
[Git-Notes] → structured decisions with entity extraction
     ↓
[File updates] → persistent workspace docs

The Three Systems

1. Baidu Embedding (Conversation Memory)

• Auto-recall: Relevant memories injected before each response using Baidu Embedding-V1 (requires API credentials)
• Auto-capture: Preferences/decisions/facts stored automatically with local vector storage (requires API credentials)
• Privacy Focused: All embeddings processed via Baidu API with local storage
• Chinese Optimized: Better understanding of Chinese language semantics
• Tools: baidu_memory_recall, baidu_memory_store, baidu_memory_forget (require API credentials)
• Triggers: "remember", "prefer", "my X is", "I like/hate/want"
• Note: When API credentials are not provided, this layer is unavailable and the system operates in degraded mode.

2. Git-Notes Memory (Structured, Local)

• Branch-aware: Memories isolated per git branch
• Entity extraction: Auto-extracts topics, names, concepts
• Importance levels: critical, high, normal, low
• No external API calls

3. File Search (Workspace)

• Searches: MEMORY.md, memory/*.md, any workspace file
• Script: scripts/file-search.sh

🛠️ Setup

Install Dependencies

clawdhub install git-notes-memory
clawdhub install memory-baidu-embedding-db

Configure Baidu API

Set environment variables:

export BAIDU_API_STRING='your_bce_v3_api_string'
export BAIDU_SECRET_KEY='your_secret_key'

Create File Search Script

Copy scripts/file-search.sh to your workspace.

📖 Usage

Session Start (Always)

python3 skills/git-notes-memory/memory.py -p $WORKSPACE sync --start

Store Important Decisions

python3 skills/git-notes-memory/memory.py -p $WORKSPACE remember 
  '{"decision": "Use PostgreSQL", "reason": "Team expertise"}' 
  -t architecture,database -i h

Search Workspace Files

./scripts/file-search.sh "database config" 5

Baidu Embedding Memory (Automatic)

Baidu Embedding handles this automatically when API credentials are available. Manual tools: - baidu_memory_recall "query" - search conversation memory using Baidu vectors (requires API credentials) - baidu_memory_store "text" - manually store something with Baidu embedding (requires API credentials) - baidu_memory_forget - delete memories (GDPR, requires API credentials)

In Degraded Mode (without API credentials): - System operates using only Git-Notes and File System layers - Manual tools are unavailable - Auto-recall and auto-capture are disabled

🎯 Importance Levels

Flag	Level	When to Use
-i c	Critical	"always remember", explicit preferences
-i h	High	Decisions, corrections, preferences
-i n	Normal	General information
-i l	Low	Temporary notes

📋 When to Use Each System

System	Use For
Baidu Embedding	Conversation context, auto-retrieval with privacy
Git-Notes	Structured decisions, searchable by entity/tag
File Search	Workspace docs, daily logs, MEMORY.md

📁 File Structure

workspace/
├── MEMORY.md              # Long-term curated memory
├── memory/
│   ├── active-context.md  # Current session state
│   └── YYYY-MM-DD.md      # Daily logs
├── scripts/
│   └── file-search.sh     # Workspace search
└── skills/
    ├── triple-memory-baidu-embedding/  # Enhanced memory system
    ├── git-notes-memory/  # Structured memory
    └── memory-baidu-embedding-db/  # Vector storage

🔧 Hook Integration

This skill can be integrated with Clawdbot's startup hooks for automatic initialization:

Memory Boot Hook Integration

To ensure the Triple Memory Baidu system starts automatically when the gateway starts:

1. The memory-boot-loader hook can be configured to run /root/clawd/session-init-triple-baidu.sh
2. This ensures all three memory layers are initialized on startup
3. Provides immediate access to Git-Notes, Baidu Embedding, and file-based memory

Configuration

The system can be integrated with the gateway startup sequence to ensure: - Git-Notes synchronization on startup - Baidu Embedding connection check - File-based memory initialization - Ready-to-use memory system from the start

🤫 Silent Operation

Never announce memory operations to users. Just do it: - ❌ "I'll remember this" - ❌ "Saving to memory" - ✅ (silently store and continue)

🔄 Migration from Original Triple Memory

1. Install this skill in your skills/ directory
2. Configure your Baidu API credentials
3. Disable original triple-memory if desired
4. Update your bot configuration to use this memory system
5. Verify data integrity and performance

📈 Performance Benefits

• Enhanced Privacy: All vector storage local with Baidu API
• Better Chinese Support: Baidu Embedding optimized for Chinese
• Reduced Costs: Potentially lower API costs compared to OpenAI
• Same Architecture: Maintains proven three-tier design

🤝 Contributing

Based on original Triple Memory system by Clawdbot Team. Contributions welcome to enhance the Baidu Embedding integration.

📄 License

Original license applies with modifications noted above. Credit given to original authors.