lily-memory-plugin
by kevinodell
Coding Agents & IDEsBrief
Persistent memory plugin for OpenClaw agents.
Description
What This Skill Does
Persistent memory plugin for OpenClaw agents that stores and retrieves facts across session resets, compaction, and restarts. Uses hybrid search combining SQLite FTS5 keyword lookup with Ollama vector embeddings for semantic similarity. Auto-captures facts from conversations and injects relevant memories as context before each LLM turn.
Combining keyword and semantic search in a single zero-dependency plugin removes the need to configure separate embedding infrastructure for most use cases.
When to Use It
- Remembering user preferences set in a previous chat session
- Recalling a project decision made weeks ago without re-explaining context
- Finding past facts about a specific entity using keyword or semantic search
- Preventing an agent from looping on the same topic repeatedly
- Maintaining continuity after an agent gateway restart
Example Workflow
User asks: 'What database did we decide to use for the auth service?'
- 1
Extract keywords from the message and run FTS5 search against SQLite
- 2
Run vector similarity search via Ollama using nomic-embed-text if available
- 3
Merge and deduplicate results, inject top matches as context before the LLM turn
- 4
LLM responds using injected memory context
- 5
Auto-capture any new facts from the response and store them to SQLite
Agent answers with the previously stored decision, e.g. 'You decided on PostgreSQL for the auth service on March 3rd.'
Requirements
- Ollama with nomic-embed-text model (optional, required only for semantic vector search)
Installation
clawhub install kevinodell/lily-memory-pluginRun the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.