openclaw-memory-graphlisted

# openclaw-memory-graph **Neo4j-backed graph memory for Openclaw providers.** Replaces the flat-file `memory/` system with a property graph that supports typed memory nodes, semantic edges, vector-similarity recall, and automatic migration of existing flat-file memories. --- ## Prerequisites ### 1. Start or Discover Neo4j **Option A — Docker Compose (local dev):** ```bash docker compose up -d ``` Starts Neo4j 5.x on `bolt://localhost:7687` with credentials `neo4j / openclaw-dev`. **Option B — Neo4j AuraDB (hosted):** Create a free instance and note the connection URI and credentials. **Option C — Existing Neo4j Instance (existing):** Use an existing instance and use the connection URI and credentials. ### 2. Set environment variables | Variable | Required | Default | Description | |----------|----------|---------|-------------| | `NEO4J_URI` | Yes | — | e.g. `bolt://localhost:7687` | | `NEO4J_USER` | Yes | — | Neo4j username | | `NEO4J_PASSWORD` | Yes | — | Neo4j password | | `OPENCLAW_EMBEDDING_PROVIDER` | No | `sentence-transformers` | `openai`, `ollama`, or `sentence-transformers` | | `OPENCLAW_EMBEDDING_MODEL` | No | `all-MiniLM-L6-v2` | Model name or local path | | `OPENAI_API_KEY` | If using OpenAI | — | OpenAI API key | | `OLLAMA_BASE_URL` | If using Ollama | `http://localhost:11434` | Local Ollama URL | ### 3. Install dependencies ```bash pip install -r requirements.txt ``` **Sentence Transformers provider** (optional — local, no API key needed): ```bash