# OpenClaw Memory System

A structured memory system for OpenClaw agents with semantic search, automatic evolution, and multi-container access.

## Features

- 🧠 **Structured Storage** - PostgreSQL-backed with rich metadata
- 🔍 **Semantic Search** - Full-text search with flexible filtering
- 📊 **Analytics** - Memory statistics and usage patterns
- 🔄 **Evolution** - Automatic merging, archiving, and forgetting
- 🌐 **Multi-Container Access** - REST API for all OpenClaw containers
- 📝 **Audit Trail** - Complete mutation and access logging

## Quick Start

### Docker Compose (Recommended)

```bash
# Create .env file
cp .env.example .env

# Edit .env with your database credentials
# nano .env

# Start the service
docker-compose up -d

# Check health
curl http://localhost:3000/health
```

### Manual Docker Build

```bash
# Build image
docker build -t openclaw-memory:latest .

# Run container
docker run -d \
  --name openclaw-memory \
  -p 3000:3000 \
  -e DB_HOST=postgres \
  -e DB_PORT=5432 \
  -e DB_NAME=openclaw \
  -e DB_USER=postgres \
  -e DB_PASSWORD=your_password \
  openclaw-memory:latest
```

### Local Development

```bash
# Install dependencies
npm install

# Run database migration
npm run migrate

# Start development server
npm run dev
```

## Database Setup

The service requires PostgreSQL with the `pgvector` extension. If using an existing database:

```sql
CREATE EXTENSION IF NOT EXISTS vector;
```

Then run the migration:

```bash
npm run migrate
```

## API Endpoints

### Create Memory

```bash
curl -X POST http://localhost:3000/api/memories \
  -H "Content-Type: application/json" \
  -d '{
    "content": "The user prefers concise responses without filler words",
    "type": "preference",
    "category": "personal",
    "priority": 2,
    "tags": ["communication", "style"],
    "source_session": "session-123"
  }'
```

### Get Memory

```bash
curl http://localhost:3000/api/memories/{id}
```

### Update Memory

```bash
curl -X PUT http://localhost:3000/api/memories/{id} \
  -H "Content-Type: application/json" \
  -d '{
    "priority": 1,
    "tags": ["communication", "style", "important"]
  }'
```

### Search Memories

```bash
curl "http://localhost:3000/api/memories/search?query=preference&type=preference&limit=10"
```

### Get Context for Session

```bash
curl "http://localhost:3000/api/memories/context?session=session-123&limit=5"
```

### Get Related Memories

```bash
curl http://localhost:3000/api/memories/{id}/related?limit=10
```

### Merge Memories

```bash
curl -X POST http://localhost:3000/api/memories/{id}/merge \
  -H "Content-Type: application/json" \
  -d '{
    "target_id": "target-memory-id",
    "reason": "Similar content about user preferences"
  }'
```

### Archive Memory

```bash
curl -X POST http://localhost:3000/api/memories/{id}/archive \
  -H "Content-Type: application/json" \
  -d '{
    "reason": "Outdated information"
  }'
```

### Get Analytics

```bash
curl http://localhost:3000/api/memories/analytics
```

## Memory Types

| Type | Description |
|------|-------------|
| `event` | Something that happened |
| `insight` | Learning or discovery |
| `pattern` | Recurring behavior or theme |
| `preference` | User preference |
| `decision` | Decision made |

## Memory Categories

| Category | Description |
|----------|-------------|
| `work` | Work-related |
| `personal` | Personal |
| `technical` | Technical |
| `social` | Social |

## Priority Levels

| Priority | Description |
|----------|-------------|
| 1 | Critical (highest) |
| 2 | High |
| 3 | Medium (default) |
| 4 | Low |
| 5 | Trivial (lowest) |

## Integration with OpenClaw

### Environment Variables

Set these in your OpenClaw containers:

```bash
MEMORY_API_URL=http://openclaw-memory:3000
MEMORY_API_ENABLED=true
```

### Example: Create Memory from Agent

```typescript
// In your agent code
const response = await fetch(`${process.env.MEMORY_API_URL}/api/memories`, {
    method: 'POST',
    headers: { 'Content-Type': 'application/json' },
    body: JSON.stringify({
        content: "User corrected my response about X",
        type: "insight",
        category: "work",
        priority: 2,
        source_session: session.id
    })
});
```

### Example: Get Context Before Session

```typescript
// Load context memories before starting a new session
const response = await fetch(
    `${process.env.MEMORY_API_URL}/api/memories/context?session=${session.id}&limit=5`
);
const { data } = await response.json();

// Inject into system prompt
const systemPrompt = `
## Relevant Memories
${data.map(m => `- ${m.summary || m.content}`).join('\n')}
`;
```

## Schema

See `src/db/schema.sql` for the complete database schema including:

- `memories` - Main memories table
- `memory_accesses` - Access log
- `memory_mutations` - Mutation log

## Migration from Files

The memory system can import data from:

- `MEMORY.md` - Long-term memory
- `memory/YYYY-MM-DD.md` - Daily memory files
- `.learnings/*.md` - Learning records

A migration script will be provided in future versions.

## Development

```bash
# Install dependencies
npm install

# Run in development mode
npm run dev

# Build for production
npm run build

# Run database migration
npm run migrate
```

## License

MIT

## Author

daotong