Aister Vector Memory

What This Skill Does

Aister Vector Memory is a sophisticated knowledge management tool designed for the OpenClaw AI agent. Unlike traditional keyword-based search mechanisms that rely on exact string matching (grep), this skill leverages semantic understanding to retrieve information based on meaning. By utilizing PostgreSQL, the pgvector extension, and the robust e5-large-v2 embedding model, it transforms textual data into high-dimensional vectors. This allows Aister to recall context, preferences, and historical data with human-like conceptual association. Whether the user is working in English or Russian, the system accurately maps queries to relevant memory chunks stored within the local workspace.

Installation

To integrate this skill into your environment, use the OpenClaw CLI tool. Run the following command in your terminal:

clawhub install openclaw/skills/skills/alekhm/aister-vector-memory

Post-installation, ensure you have a running PostgreSQL 16 instance with the pgvector extension enabled. You must configure the VECTOR_MEMORY_DB_PASSWORD environment variable to grant the agent access to the database. You may also adjust optional parameters such as VECTOR_MEMORY_DIR or EMBEDDING_SERVICE_URL if you are hosting your embedding service on a custom port or remote server.

Use Cases

This skill is ideal for users who manage complex projects or maintain detailed identity files. It is particularly useful for:

• Storing and retrieving long-form project notes or documentation.
• Maintaining user identity profiles where the AI needs to remember personal preferences or past communication styles.
• Quickly indexing large repositories of text files (like MEMORY.md or USER.md) for instant query-based retrieval.
• Enhancing agent autonomy by allowing it to maintain 'long-term memory' across multiple sessions.

Example Prompts

1. /search_memory how do I format my daily standup logs?
2. /search_memory what was the decision we made regarding the Moltbook UI design?
3. /search_memory my communication style preferences

Tips & Limitations

To maintain optimal performance, remember to trigger /reindex_memory whenever you perform significant updates to your local memory files. The system uses a default similarity threshold of 0.5; if you find the search results are too broad or too restrictive, you can tune this value via the VECTOR_MEMORY_THRESHOLD variable. Keep in mind that this tool requires an active embedding service (defaulting to port 8765) to function, so ensure your local backend is running before attempting to query your memories.