A local hybrid search engine that eliminates API costs by running embeddings and semantic queries directly on your machine.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install qmd-memory-1-0-0
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 qmd-memory-1-0-0 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 QMD Memory Skill?
The QMD Memory Skill is a powerful local-first search infrastructure designed to supercharge your AI agents while eliminating monthly cloud API fees. By moving embedding, re-ranking, and query expansion tasks to your local hardware, it provides a high-performance alternative to expensive vector search providers. This skill is a core addition for users of Openclaw Skills who require private, fast, and cost-effective context retrieval for large knowledge bases.
It works by automatically organizing your workspace into specialized collections such as daily logs, core agent identity files, and project documentation. Through the use of efficient GGUF models, the skill ensures that your agent has access to its entire history and intelligence without the latency or privacy concerns associated with cloud-based search operations.
QMD Memory Skill Use Cases
- Reducing monthly API overhead for high-frequency memory search and context retrieval tasks.
- Implementing private, local-first semantic search for sensitive internal documentation and research.
- Maintaining a persistent, searchable history of agent decisions and daily work logs.
- Creating shared knowledge bases across multiple agents using the built-in MCP server integration.
How QMD Memory Skill Works
- The skill analyzes your local workspace to identify core directories and automatically configures optimized collections.
- It downloads lightweight GGUF models for vector embeddings, re-ranking, and query expansion during the initial setup.
- A local SQLite index is generated to manage both keyword and semantic vector data for fast retrieval.
- Search queries are automatically routed through a hybrid pipeline that combines exact keyword matching with conceptual semantic analysis.
- Automated cron jobs keep the local index synchronized with file changes, ensuring the agent's memory is always up to date.
QMD Memory Skill Setup
# Install the skill via ClawHub
clawhub install asabove/qmd-memory
# Execute the setup routine to install QMD and download GGUF models
openclaw skill run qmd-memory setup
# Optional: Run the savings calculator to see your projected ROI
openclaw skill run qmd-memory calculate-savings
QMD Memory Skill Data Schema & Taxonomy
The skill organizes data within a dedicated cache directory (~/.cache/qmd/) to ensure persistence and speed.
| Component | Format | Description |
|---|---|---|
| index.sqlite | SQLite | The primary search index containing text chunks and vector embeddings. |
| models/ | GGUF | Local models for embedding (gemma), re-ranking (qwen3), and query expansion. |
| collections | URI Schema | Logical groupings (e.g., qmd://daily-logs) that provide context to the agent. |
| mcp.pid | PID File | Management file for the multi-agent MCP server process. |
QMD Memory Skill Advanced Features
- Multi-Agent MCP Server: Start a shared memory server that allows different Openclaw Skills agents to access a collective knowledge base.
- Workspace Templates: Pre-configured collection sets for specialized roles like developers, traders, and content creators.
- Local Hybrid Search: Combines keyword, semantic (vsearch), and expanded query modes for maximum retrieval accuracy.
- Automated Indexing: Built-in cron job support to refresh your memory index nightly or on a custom schedule.
- Local Reranking: Uses a 0.6b parameter model to re-sort search results for better relevance before passing them to the LLM.
SKILL.md
Loading
Related Openclaw Skills
NEON-SOUL
NEON-SOUL provides an automated framework for synthesizing an AI agent's identity and core principles directly from its interaction history and memory files.
Memory System v2.0
A high-performance, file-based semantic memory system enabling AI agents to persist and recall learnings and decisions with sub-20ms search speeds.
mem0 Memory Integration
The mem0 skill provides an intelligent, adaptive memory layer that learns and recalls user preferences, patterns, and context across conversations.
gog: Google Workspace CLI
A versatile command-line tool for managing Gmail, Calendar, Drive, Sheets, and Docs via simple terminal commands.
Save Money
An intelligent behavioral skill that automatically routes tasks between high-performance and low-cost AI models to slash API expenses by over 50%.
Self-Improving Agent (With Self-Reflection)
An autonomous learning framework that enables AI agents to evaluate their work, incorporate user corrections, and build a permanent knowledge base through self-reflection.
Featured*
