Qmd Memory

What This Skill Does

The QMD Memory skill by As Above Technologies is a high-performance local hybrid search engine for OpenClaw. It shifts memory indexing, embedding, and retrieval from costly cloud APIs to your local machine. By leveraging GGUF local models and a custom hybrid search architecture, it enables you to manage your agent's long-term memory at zero operational cost, saving users significant monthly expenditure. The skill automatically indexes your workspace, daily logs, and documentation, providing semantic, keyword, and hybrid search capabilities that keep your agent's knowledge base organized and instantly queryable.

Installation

Getting started is streamlined via the ClawHub ecosystem. Simply execute the command clawhub install asabove/qmd-memory within your OpenClaw terminal. Once installed, run openclaw skill run qmd-memory setup to initialize your local indexing collections and configure the required directory watchdogs. The installation automatically sets up necessary cron jobs to ensure your memory remains refreshed and indexed without manual intervention.

Use Cases

This skill is perfect for power users and developers managing extensive project documentation, daily research logs, or complex agent identity configurations. If you are building multiple agents, the multi-agent MCP server functionality allows a single 'source of truth' memory store to be shared across multiple agents, preventing knowledge silos. It is also an essential tool for users looking to optimize cloud infrastructure costs while maintaining high-performance retrieval speeds that local execution guarantees.

Example Prompts

1. "QMD, search the workspace collection for all decisions made regarding the pricing strategy last month and summarize them for me."
2. "Perform a semantic search in my intelligence folder for notes regarding 'customer churn' and list the top 3 relevant findings."
3. "Run a hybrid query to retrieve the API key documentation from my project notes and display the matching code blocks."

Tips & Limitations

• Tip: Use the 'hybrid' search mode for the most accurate results, as it intelligently combines semantic understanding with exact keyword matching.
• Tip: Configure your cron jobs during the setup process to ensure your index is updated automatically at 3 AM to avoid performance hits during peak usage.
• Limitation: As this runs locally, ensure your system has sufficient RAM to handle the GGUF embedding models. Performance may vary on hardware with less than 16GB of system memory.