A zero-dependency persistent memory plugin for OpenClaw agents that combines keyword and semantic search for long-term context retention.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install lily-memory-5-0-0
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install lily-memory-5-0-0 using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Lily Memory?
Lily Memory is a sophisticated persistent memory plugin designed to empower Openclaw Skills with the ability to retain and recall information across sessions, resets, and restarts. By bridging the gap between short-term context windows and long-term knowledge, it allows agents to maintain a consistent state and remember user preferences, project decisions, and historical facts. This tool is essential for developers building complex autonomous agents that require a reliable way to store and retrieve data without manual intervention.
The plugin utilizes a hybrid search architecture, combining the speed of SQLite FTS5 keyword indexing with the deep understanding of Ollama-powered vector semantic search. This ensures that whether an agent needs a specific keyword or a conceptually related piece of information, Lily Memory can provide the most relevant context. It operates with zero npm dependencies, relying on native system tools to maintain a lightweight and high-performance footprint within the Openclaw Skills ecosystem.
Lily Memory Use Cases
- Persistent storage of user configurations and architectural decisions across multiple agent restarts.
- Building autonomous researchers that can recall previously discovered facts from past sessions.
- Preventing agent loops and repetitive behavior through integrated stuck detection and reflexion nudges.
- Categorizing and retrieving entity-specific data like system logs, config files, or user profiles automatically.
How Lily Memory Works
- Recall Phase: Upon receiving a message, the skill extracts relevant keywords and generates vector embeddings to query the persistent database.
- Hybrid Search: It executes a parallel search using SQLite FTS5 for exact matches and Ollama cosine similarity for semantic matches.
- Context Injection: The most relevant memories are merged, deduplicated, and injected into the LLM context window before the turn begins.
- Capture Phase: The plugin monitors agent responses for predefined entity patterns to automatically extract and store new facts.
- Maintenance: On startup, the system performs memory consolidation to deduplicate entries and optimize the search index.
Lily Memory Setup
To integrate Lily Memory into your Openclaw Skills environment, ensure you have Node.js 18+ and SQLite 3.33+ installed. Follow these steps:
- Install the plugin into your extensions directory.
- Configure your
openclaw.jsonto enable the plugin and define the database path:
{
"plugins": {
"slots": { "memory": "lily-memory" },
"entries": {
"lily-memory": {
"enabled": true,
"config": {
"dbPath": "~/.openclaw/memory/decisions.db",
"entities": ["config", "system"]
}
}
}
}
}
- Apply the changes by restarting your gateway:
openclaw gateway restart
Lily Memory Data Schema & Taxonomy
Lily Memory organizes data within a localized SQLite database, managing both keyword indices and vector embeddings. The schema focuses on the following primary components:
| Component | Description |
|---|---|
| Facts Table | Stores the raw text, entity association, and timestamp of captured information. |
| FTS5 Index | A virtual table enabling high-speed full-text search across all stored facts. |
| Vector Store | Optional storage for embeddings generated via Ollama for semantic similarity lookup. |
| Entity Allowlist | A configuration-driven list of valid categories (e.g., config, user, system) used for data validation. |
Lily Memory Advanced Features
- Stuck Detection: Uses Jaccard similarity to track word repetition and trigger a reflexion nudge if the agent becomes repetitive.
- Graceful Degradation: Automatically switches to keyword-only mode if the Ollama service is unavailable, ensuring the skill remains functional.
- Dynamic Entity Management: Provides tools to register and track new data entities at runtime without manual configuration updates.
- Zero-Dependency Architecture: Built using native fetch and the sqlite3 CLI to avoid the bloat and security risks of external npm packages.
SKILL.md
Loading
Related Openclaw Skills
Warden Studio Agent Publisher
A specialized browser automation skill designed to register and publish Community Agents to the Warden Agent Hub via the Warden Studio interface.
Warden Agent Builder
A comprehensive toolkit for developing, testing, and deploying original LangGraph-based AI agents within the Warden Protocol ecosystem.
Dify AI Application Development Skill
Dify is an open-source LLM application development platform designed to orchestrate AI workflows, agents, and RAG-based knowledge bases.
Sogni Image & Video Generation
A powerful CLI tool for generating decentralized AI images and videos through the Openclaw Skills framework.
Trade Signal
An advanced financial intelligence skill providing real-time trade signals, price targets, and institutional-grade analysis for global stocks and ETFs.
Trade Signal
A professional-grade financial intelligence skill providing real-time trade signals, technical analysis, and fundamental research for global stocks and ETFs.
Featured*
