A high-performance memory system for AI agents using QST theory and matrix-based retrieval to bridge short-term context and long-term knowledge.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install qst-memory-system
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install qst-memory-system using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is QST Memory Skill?
The QST Memory Skill is a sophisticated memory management layer for Openclaw Skills designed to emulate human-like memory persistence. By leveraging the QST (Quantum Spacetime Topology) framework—including E8 geometric structures and Distributed Semantic Information (DSI) hierarchies—this skill allows agents to maintain coherent conversations over extended periods while efficiently offloading older information to long-term storage.
This skill is essential for developers building complex AI workflows that require more than just a simple message buffer. It provides a hybrid approach, combining traditional vector embeddings with advanced matrix projections to ensure millisecond-level retrieval of relevant context. By integrating this system, your Openclaw Skills can transition seamlessly between immediate working memory and persistent knowledge bases.
QST Memory Skill Use Cases
- Persistent agent personalities that remember user preferences and history across multiple sessions.
- Complex multi-step reasoning tasks where short-term context window limits often act as a bottleneck.
- Building a local, evolving knowledge base that grows through direct conversation history.
- High-performance retrieval-augmented generation (RAG) using the QST framework within Openclaw Skills.
- Intelligent context pruning for long-running autonomous agent loops.
How QST Memory Skill Works
- Capture conversation input and store it initially in the high-coherence Working Memory buffer for immediate access.
- Project the content into the QST Matrix using E8 geometry dimensions (defaulting to 16) for high-fidelity semantic mapping.
- Apply the ICT Collapse algorithm during retrieval to find the most relevant memory nodes based on the current user query.
- Periodically trigger the consolidation process to move high-coherence data from volatile short-term buffers to SQLite or JSON long-term storage.
- Use the decay function to automatically prune low-relevance or outdated memories, maintaining system efficiency over time.
QST Memory Skill Setup
To install the QST Memory Skill, navigate to your local workspace and clone the repository:
cd /root/.openclaw/workspace/skills
git clone https://github.com/ZhuangClaw/qst-memory-system.git qst-memory
Next, create a config.json file in the same directory as the SKILL.md to define your memory parameters:
{
"e8_dim": 16,
"top_k": 5,
"storage_type": "hybrid",
"embedding_type": "simple"
}
QST Memory Skill Data Schema & Taxonomy
The system organizes data based on a coherence coefficient (σ) and DSI layers. Data is structured as follows:
| Coherence (σ) | Memory Type | Persistence Strategy |
|---|---|---|
| 0.7 | Working Memory | 30 min - 24 hr Cache |
| 0.85 | Short Memory | Local Cache / JSON |
| 0.95 | Medium Memory | Indexed JSON |
| 1.1 | Long Memory | Permanent SQLite Database |
DSI layers are calculated using the Golden Ratio (φ ≈ 1.618) to determine the depth of storage across 36 potential hierarchical levels.
QST Memory Skill Advanced Features
- ICT Collapse Retrieval: Probabilistic memory retrieval based on quantum-inspired wavefunction collapse for more accurate context matching.
- Hybrid Embedding Support: Seamlessly switch between OpenAI Embeddings and local Sentence-Transformers for Openclaw Skills.
- Automatic Memory Decay: Configurable decay rates ensure that the agent remains focused on relevant, high-quality information while discarding noise.
- E8 Projection: Utilizes multi-dimensional geometric projection (E8 mathematics) for advanced semantic relationship mapping between disparate data points.
- Manual Consolidation: Provides an API to force the transfer of short-term dialogue into long-term structured knowledge.
SKILL.md
Loading
Related Openclaw Skills
Universal Memory Management (qst-memory)
A universal memory management system for Openclaw agents that optimizes token usage and provides intelligent state-aware context.
Q Memory: Universal Memory Management System
A sophisticated memory management system for AI agents that integrates state awareness, loop protection, and hybrid semantic search to optimize token usage.
Feishu Cache Guardian
An automated configuration guard that protects Feishu API quotas by maintaining optimized cache settings within Openclaw Skills.
XHS Research Daily
An automated pipeline for researching, synthesizing, and publishing daily Xiaohongshu content digests.
Universal Memory Management System (qst-memory)
A high-performance memory management framework for AI agents featuring hierarchical classification, hybrid search, and automated loop protection.
QST Universal Memory Management
A sophisticated multi-agent memory and state management system designed to optimize token usage and prevent infinite task loops for Openclaw agents.
Featured*
