Pidan Memory

What This Skill Does

Pidan Memory is a sophisticated, locally-hosted long-term memory solution for OpenClaw AI agents. By leveraging LanceDB for vector storage and Ollama for embedding generation (using the nomic-embed-text model), this skill enables your assistant to store, recall, and manage user-specific information persistently. Unlike volatile context windows, Pidan Memory creates a permanent knowledge base that grows with every interaction, supporting both private, isolated user data and shared memory environments. The system includes automated hooks that evaluate and store significant conversation details, ensuring your AI retains important preferences and context across sessions.

Installation

To integrate Pidan Memory into your OpenClaw environment, ensure you have Ollama installed and running. Follow these steps:

1. Copy the necessary hooks by running: mkdir -p ~/.openclaw/hooks/pidan-memory && cp HOOK.md handler.ts ~/.openclaw/hooks/pidan-memory/.
2. Deploy the memory backend: cp auto_memory.py ~/.openclaw/workspace/memory/.
3. Enable the hooks via the CLI: openclaw hooks enable pidan-memory.
4. Apply changes by restarting the gateway: openclaw gateway restart.

Use Cases

• Personalization: Remembering user food preferences, birthdays, or technical skill levels to customize future interactions.
• Project Context: Keeping track of recurring project requirements or specific task history across long-term workflows.
• Multi-User Collaboration: Utilizing the 'shared' mode to maintain a common knowledge base in a team setting where specific data needs to be accessible to authorized team members.
• Knowledge Retrieval: Quickly searching for past notes or decisions that would otherwise be lost in chat logs.

Example Prompts

1. "Remember that my preferred programming language is Rust and I prefer clean, concise documentation style."
2. "Search your memory for any notes I saved regarding the quarterly project budget."
3. "Who has access to my current memory base, and can you list the last five things we discussed?"

Tips & Limitations

• Permissions: Memory security is managed via user IDs. Ensure you are using proper user context, as sensitive information is persistent locally.
• Maintenance: While the system automatically triggers deduplication every 20 entries, you can manually trigger it to optimize database size.
• Hardware: Because this uses Ollama, ensure your machine has sufficient RAM and CPU/GPU capacity to handle local vector embedding tasks smoothly.
• Deletion: Deletion is a two-step process requiring confirmation to prevent accidental loss of important data.