mem

What This Skill Does

The mem skill provides OpenClaw with a powerful local-first memory retrieval system, allowing the agent to query a local vector index of your personal files, notes, and documentation. By bypassing the need for cloud-based external search tools, mem ensures that your private data stays on your machine, providing fast and private context for your AI agent. It is specifically optimized for /mem commands within the Telegram interface, enabling instant access to your historical data.

Installation

To install the memory retrieval skill, use the clawhub command-line tool. Open your terminal and run:

clawhub install openclaw/skills/skills/trumppo/mem

Once installed, ensure you have initialized your database. Navigate to the source repository folder and run the indexing script to parse your documents into a searchable format:

python3 scripts/index-memory.py

This will generate the vector embeddings required for the search queries to function correctly. You may need to rerun this script whenever you add significant new information to your local directory.

Use Cases

This skill is ideal for users who manage large amounts of local text-based data and need quick access to specific information without searching through directories manually. It is particularly useful for:

• Retrieving specific meeting notes or project requirements from local Markdown files.
• Finding historical technical configurations or command snippets that you have previously saved.
• Querying knowledge bases or personal wiki files stored on your local disk.
• Providing the AI with context from your local documentation during a chat session.

Example Prompts

1. "/mem What was the protocol for the Q3 project migration we discussed last month?"
2. "/mem Search my notes for the API configuration settings for the production server."
3. "/mem Find the code snippet for the authentication helper I wrote last week."

Tips & Limitations

To get the best results, ensure that your index-memory.py script is scheduled or triggered periodically to reflect changes in your data. The accuracy of the search is highly dependent on the quality and formatting of your source documents; keeping notes structured with clear headings helps the indexing process. Please note that this skill is currently limited to local search; it does not crawl live web pages or cloud-synced storage unless they are mounted locally. Large file indices may increase the latency of the initial search response, but the top-5 relevance threshold is designed to keep retrieval times efficient for real-time conversation.