Deployment

# Prerequisites: Go 1.23+, Docker, Ollama
docker compose up -d             # Start Memgraph
ollama pull nomic-embed-text     # Pull embedding model
go build -o bin/openclaw-cortex ./cmd/openclaw-cortex

export ANTHROPIC_API_KEY=sk-ant-...

# Index memory files
openclaw-cortex index --path ~/.openclaw/workspace/memory/

# Search
openclaw-cortex search "deployment best practices"

# Recall with context
openclaw-cortex recall "How do I handle errors?" --budget 2000

# Health check
openclaw-cortex health

docker build -t openclaw-cortex:latest .
docker run --rm \
  -e ANTHROPIC_API_KEY=sk-ant-... \
  -e OPENCLAW_CORTEX_MEMGRAPH_URI=bolt://host.docker.internal:7687 \
  openclaw-cortex:latest search "query"

docker compose up -d   # Starts Memgraph with persistent volume
docker compose down    # Stops Memgraph

services:
  memgraph:
    image: memgraph/memgraph:latest
    ports:
      - "7687:7687"   # Bolt (used by openclaw-cortex)
      - "7444:7444"   # HTTP Lab UI
    command: ["--storage-properties-on-edges=true"]
    volumes:
      - memgraph_data:/var/lib/memgraph
    restart: unless-stopped

volumes:
  memgraph_data:

apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: memgraph
  namespace: cortex
spec:
  selector:
    matchLabels:
      app: memgraph
  serviceName: memgraph
  replicas: 1
  template:
    metadata:
      labels:
        app: memgraph
    spec:
      containers:
      - name: memgraph
        image: memgraph/memgraph:latest
        args: ["--storage-properties-on-edges=true"]
        ports:
        - containerPort: 7687
          name: bolt
        - containerPort: 7444
          name: http
        volumeMounts:
        - name: data
          mountPath: /var/lib/memgraph
  volumeClaimTemplates:
  - metadata:
      name: data
    spec:
      accessModes: ["ReadWriteOnce"]
      resources:
        requests:
          storage: 10Gi
---
apiVersion: v1
kind: Service
metadata:
  name: memgraph
  namespace: cortex
spec:
  selector:
    app: memgraph
  ports:
  - name: bolt
    port: 7687
    targetPort: 7687
  - name: http
    port: 7444
    targetPort: 7444
  clusterIP: None

apiVersion: batch/v1
kind: CronJob
metadata:
  name: cortex-consolidate
  namespace: openclaw-cortex
spec:
  schedule: "0 */6 * * *"  # Every 6 hours
  jobTemplate:
    spec:
      template:
        spec:
          containers:
          - name: cortex
            image: openclaw-cortex:latest
            args: ["consolidate"]
            env:
            - name: OPENCLAW_CORTEX_MEMGRAPH_URI
              value: bolt://memgraph.cortex.svc.cluster.local:7687
            - name: ANTHROPIC_API_KEY
              valueFrom:
                secretKeyRef:
                  name: cortex-secrets
                  key: anthropic-api-key
          restartPolicy: OnFailure

memgraph:
  uri: bolt://localhost:7687
  username: ""
  password: ""

ollama:
  base_url: http://localhost:11434
  model: nomic-embed-text

memory:
  dedup_threshold: 0.92
  default_ttl_hours: 720

claude:
  # Option 1: direct Anthropic API
  api_key: sk-ant-...
  # Option 2: OpenClaw gateway (Max plan / subscription users)
  # gateway_url: http://127.0.0.1:18789/v1/chat/completions
  # gateway_token: <your-gateway-token>

# Full health check (Memgraph + Ollama + LLM)
openclaw-cortex health

# Verify Ollama model is available
ollama list | grep nomic-embed-text

# Verify Memgraph connectivity via stats
openclaw-cortex stats

# Copy the snapshot directory from the running container
docker cp memgraph:/var/lib/memgraph/snapshots ./memgraph-backup

# Or with Docker volume backup
docker run --rm \
  -v memgraph_data:/data \
  -v $(pwd):/backup \
  alpine tar czf /backup/memgraph-backup.tar.gz /data

# Restore from volume backup
docker run --rm \
  -v memgraph_data:/data \
  -v $(pwd):/backup \
  alpine tar xzf /backup/memgraph-backup.tar.gz -C /

# Then restart Memgraph
docker compose up -d

# Export all memories
openclaw-cortex export --format json -o memories-backup.json

# Import on a new instance
openclaw-cortex import --file memories-backup.json