Maasv Memory

What This Skill Does

Maasv Memory replaces OpenClaw's default memory backend with a high-performance cognition layer designed for long-term state management. By leveraging the maasv architecture, this skill introduces a 3-signal retrieval system that combines semantic search, keyword matching, and a dynamic knowledge graph. Unlike standard vector stores that treat memories as flat data points, maasv performs automated entity extraction to build a web of relationships between people, projects, and concepts. This allows your agent to not only recall facts but understand the context behind them. Furthermore, it implements temporal versioning, ensuring that your agent's understanding evolves as information changes over time, all while maintaining absolute privacy on your local machine using SQLite.

Installation

To integrate Maasv Memory into your workflow, follow these steps:

1. Server Setup: Install the server components using pip install "maasv[server,anthropic,voyage]". Create a .env file from the example provided to input your necessary API keys (Anthropic, OpenAI, or Ollama).
2. Run Server: Launch the daemon using the maasv-server command. Ensure the port (default 18790) is accessible by OpenClaw.
3. Plugin Installation: Execute openclaw plugins install @maasv/openclaw-memory in your terminal.
4. Activation: Update your ~/.openclaw/openclaw.json configuration file to point the memory slot to "memory-maasv" and enable the desired flags like autoRecall, autoCapture, and enableGraph to suit your agent's needs.

Use Cases

• Project Management: Agents can track complex project dependencies and timelines across multiple sessions, maintaining coherence over weeks or months of work.
• Personal CRM: Build a sophisticated database of professional or personal contacts, linking people to the specific projects or topics they were associated with in previous conversations.
• Learning & Research: Use the knowledge graph to map concepts together, allowing the agent to provide summary insights based on cross-references from your research history.

Example Prompts

1. "Based on my notes from last month, what were the main concerns I had regarding the Q3 architectural changes?"
2. "Who are the developers I discussed with about the database migration project last week, and what was their final recommendation?"
3. "Summarize all the entities and concepts related to 'Maasv' that I've interacted with since I installed this plugin."

Tips & Limitations

• Optimization: For maximum performance, utilize the Qwen3-Embedding-8B model via Ollama to keep processing local. This minimizes latency and removes dependency on external cloud providers.
• Graph Complexity: If you enable the knowledge graph, monitor your database size periodically. Large, interconnected webs of data can increase memory retrieval time, though the SQLite backend remains highly efficient.
• Privacy: Remember that while data is stored locally, the entity extraction process may still send text to an LLM provider if you choose a cloud-based model like Anthropic or OpenAI. Use a local LLM if strict data sovereignty is required.