Lily Memory

Persistent memory plugin for OpenClaw agents. Gives your agent long-term memory that survives session resets, compaction, and restarts.

What It Does

Auto-recall: Injects relevant memories as context before each LLM turn
Auto-capture: Extracts facts from conversations and stores them automatically
Hybrid search: SQLite FTS5 keyword search + Ollama vector cosine similarity
Stuck detection: Detects topic repetition and nudges the agent to break loops
Memory consolidation: Deduplicates entries on startup
Dynamic entities: Config-driven allowlist + runtime tool to add entities
Graceful degradation: Works without Ollama (keyword-only mode)
Minimal dependencies: Uses better-sqlite3 for secure parameterized queries + native fetch

Requirements

Node.js 18+ (for native fetch)
better-sqlite3 npm package (installed via npm install)
Optional: Ollama with nomic-embed-text model for semantic search

Quick Start

Install the plugin to your extensions directory
Add to your openclaw.json:

{
  "plugins": {
    "slots": { "memory": "lily-memory" },
    "entries": {
      "lily-memory": {
        "enabled": true,
        "config": {
          "dbPath": "~/.openclaw/memory/decisions.db",
          "entities": ["config", "system"]
        }
      }
    }
  }
}

Restart the gateway: openclaw gateway restart

Tools

Tool	Description
`memory_search`	FTS5 keyword search across all facts
`memory_entity`	Look up all facts for a specific entity
`memory_store`	Save a fact to persistent memory
`memory_semantic_search`	Vector similarity search via Ollama
`memory_add_entity`	Register a new entity at runtime

Configuration

Option	Type	Default	Description
`dbPath`	string	`~/.openclaw/memory/decisions.db`	SQLite database path
`autoRecall`	boolean	`true`	Inject memories before each turn
`autoCapture`	boolean	`true`	Extract facts from responses
`maxRecallResults`	number	`10`	Max memories per turn
`maxCapturePerTurn`	number	`5`	Max facts per response
`stuckDetection`	boolean	`true`	Topic repetition detection
`vectorSearch`	boolean	`true`	Ollama semantic search
`ollamaUrl`	string	`http://localhost:11434`	Ollama endpoint
`embeddingModel`	string	`nomic-embed-text`	Embedding model
`consolidation`	boolean	`true`	Dedup on startup
`vectorSimilarityThreshold`	number	`0.5`	Min cosine similarity
`entities`	array	`[]`	Additional entity names

Architecture

Recall flow: Extract keywords from message -> FTS5 + vector search -> merge and deduplicate -> inject as context

Capture flow: Regex scan for entity: key = value patterns -> validate entity against allowlist -> store to SQLite -> async embed via Ollama

Stuck detection: Track top 5 content words per response -> Jaccard similarity -> if 3+ consecutive >60% overlap, inject Reflexion nudge

License

MIT