# mem0 NixOS Module

Mem0 REST API server module for AI memory management.

## Overview

This module provides a systemd service for the Mem0 REST API server, enabling AI agents to maintain persistent memory across conversations.

## Quick Start

```nix
{config, ...}: {
  imports = [m3ta-nixpkgs.nixosModules.mem0];

  m3ta.mem0 = {
    enable = true;
    port = 8000;
    llm = {
      provider = "openai";
      apiKeyFile = "/run/secrets/openai-api-key";
      model = "gpt-4o-mini";
    };
    vectorStore = {
      provider = "qdrant";
      config = {
        host = "localhost";
        port = 6333;
      };
    };
  };
}
```

## Module Options

### `m3ta.mem0.enable`

Enable the Mem0 REST API server.

- Type: `boolean`
- Default: `false`

### `m3ta.mem0.package`

The mem0 package to use.

- Type: `package`
- Default: `pkgs.mem0`

### `m3ta.mem0.host`

Host address to bind the server to.

- Type: `string`
- Default: `"127.0.0.1"`

### `m3ta.mem0.port`

Port to run the REST API server on.

- Type: `port`
- Default: `8000`

### `m3ta.mem0.workers`

Number of worker processes.

- Type: `integer`
- Default: `1`

### `m3ta.mem0.logLevel`

Logging level for the server.

- Type: `enum` ["critical" "error" "warning" "info" "debug" "trace"]
- Default: `"info"`

### `m3ta.mem0.stateDir`

Directory to store mem0 data and state.

- Type: `path`
- Default: `"/var/lib/mem0"`

### `m3ta.mem0.user`

User account under which mem0 runs.

- Type: `string`
- Default: `"mem0"`

### `m3ta.mem0.group`

Group under which mem0 runs.

- Type: `string`
- Default: `"mem0"`

### `m3ta.mem0.environmentFile`

Environment file containing additional configuration.

- Type: `nullOr path`
- Default: `null`

## LLM Configuration

### `m3ta.mem0.llm.provider`

LLM provider to use.

- Type: `enum` ["openai" "anthropic" "azure" "groq" "together" "ollama" "litellm"]
- Default: `"openai"`

### `m3ta.mem0.llm.model`

Model name to use.

- Type: `string`
- Default: `"gpt-4o-mini"`

### `m3ta.mem0.llm.apiKeyFile`

Path to file containing the API key.

- Type: `nullOr path`
- Default: `null`
- Example: `"/run/secrets/openai-api-key"`

### `m3ta.mem0.llm.temperature`

Temperature parameter for LLM generation.

- Type: `nullOr float`
- Default: `null`

### `m3ta.mem0.llm.maxTokens`

Maximum tokens for LLM generation.

- Type: `nullOr int`
- Default: `null`

### `m3ta.mem0.llm.extraConfig`

Additional LLM configuration options.

- Type: `attrs`
- Default: `{}`

## Vector Store Configuration

### `m3ta.mem0.vectorStore.provider`

Vector database provider.

- Type: `enum` ["qdrant" "chroma" "pinecone" "weaviate" "faiss" "pgvector" "redis" "elasticsearch" "milvus"]
- Default: `"qdrant"`

### `m3ta.mem0.vectorStore.config`

Configuration for the vector store.

- Type: `attrs`
- Default: `{}`

Example for Qdrant:

```nix
vectorStore.config = {
  host = "localhost";
  port = 6333;
  collection_name = "mem0_memories";
};
```

Example for pgvector:

```nix
vectorStore.config = {
  host = "localhost";
  port = 5432;
  dbname = "postgres";
  user = "postgres";
  password = "postgres";
};
```

## Embedder Configuration

### `m3ta.mem0.embedder.provider`

Embedding model provider.

- Type: `nullOr (enum` ["openai" "huggingface" "ollama" "vertexai"])
- Default: `null`

### `m3ta.mem0.embedder.model`

Embedding model name.

- Type: `nullOr string`
- Default: `null`

### `m3ta.mem0.embedder.config`

Configuration for the embedder.

- Type: `attrs`
- Default: `{}`

## Usage Examples

### Minimal Configuration

```nix
{config, ...}: {
  m3ta.mem0 = {
    enable = true;
  };
}
```

### With OpenAI

```nix
{config, ...}: {
  m3ta.mem0 = {
    enable = true;
    llm = {
      provider = "openai";
      apiKeyFile = "/run/secrets/openai-api-key";
      model = "gpt-4";
    };
  };
}
```

### With Local LLM (Ollama)

```nix
{config, ...}: {
  m3ta.mem0 = {
    enable = true;
    llm = {
      provider = "ollama";
      model = "llama2";
    };
  };
}
```

### With Port Management

```nix
{config, ...}: {
  m3ta.ports = {
    enable = true;
    definitions = {
      mem0 = 8000;
    };
    currentHost = config.networking.hostName;
  };

  m3ta.mem0 = {
    enable = true;
    port = config.m3ta.ports.get "mem0";
  };
}
```

### With Qdrant

```nix
{config, ...}: {
  m3ta.mem0 = {
    enable = true;
    vectorStore = {
      provider = "qdrant";
      config = {
        host = "localhost";
        port = 6333;
      };
    };
  };
}

services.qdrant = {
  enable = true;
  port = 6333;
};
```

### With Secrets (agenix)

```nix
{config, ...}: {
  age.secrets.openai-api-key = {
    file = ./secrets/openai-api-key.age;
  };

  m3ta.mem0 = {
    enable = true;
    llm = {
      apiKeyFile = config.age.secrets.openai-api-key.path;
    };
  };
}
```

## Service Management

### Start/Stop/Restart

```bash
# Start service
sudo systemctl start mem0

# Stop service
sudo systemctl stop mem0

# Restart service
sudo systemctl restart mem0

# Check status
sudo systemctl status mem0
```

### View Logs

```bash
# View logs
sudo journalctl -u mem0 -f

# View last 100 lines
sudo journalctl -u mem0 -n 100
```

### Service File

The module creates a systemd service at `/etc/systemd/system/mem0.service` with:

- Security hardening enabled
- Automatic restart on failure
- Proper user/group setup

## API Usage

### Add Memory

```bash
curl -X POST http://localhost:8000/v1/memories \
  -H "Content-Type: application/json" \
  -d '{
    "content": "User prefers coffee over tea",
    "metadata": {"user_id": "123"}
  }'
```

### Search Memories

```bash
curl http://localhost:8000/v1/memories/search?q=coffee
```

### Update Memory

```bash
curl -X PATCH http://localhost:8000/v1/memories/memory_id \
  -H "Content-Type: application/json" \
  -d '{
    "content": "User prefers coffee over tea, but also likes chai"
  }'
```

### Delete Memory

```bash
curl -X DELETE http://localhost:8000/v1/memories/memory_id
```

## Dependencies

### Required Services

Depending on your configuration, you may need:

- **qdrant** service (if using qdrant vector store)
- **postgresql** with pgvector (if using pgvector)
- **chroma** service (if using chroma)
- **ollama** (if using local LLMs)

### Example: Qdrant

```nix
services.qdrant = {
  enable = true;
  port = 6333;
};
```

### Example: PostgreSQL

```nix
services.postgresql = {
  enable = true;
  enableTCPIP = true;
  package = pkgs.postgresql_15;

  extensions = ["pgvector"];

  settings = {
    port = 5432;
  };
};
```

## Firewall

The module automatically opens the firewall port if binding to non-localhost addresses:

```nix
# Opens port if host is not "127.0.0.1" or "localhost"
m3ta.mem0 = {
  enable = true;
  host = "0.0.0.0";  # Binds to all interfaces
  port = 8000;
};
# Firewall automatically opens port 8000
```

## Security

### User/Group

Creates dedicated user and group:

- User: `mem0`
- Group: `mem0`
- Home: `/var/lib/mem0`

### Hardening

Systemd service includes security hardening:

- `NoNewPrivileges`
- `PrivateTmp`
- `ProtectSystem=strict`
- `ProtectHome=true`
- `RestrictRealtime=true`
- `RestrictNamespaces=true`
- `LockPersonality=true`

### Secrets

Use `apiKeyFile` for API keys instead of plain text:

```nix
# Good
llm.apiKeyFile = "/run/secrets/openai-api-key";

# Bad (insecure)
llm.apiKey = "sk-xxx";
```

## Troubleshooting

### Service Won't Start

Check logs:

```bash
sudo journalctl -u mem0 -n 50
```

Common issues:

1. **API key missing**: Ensure `apiKeyFile` exists and is readable
2. **Vector store unavailable**: Ensure qdrant/other store is running
3. **Port in use**: Check if port is available

### API Not Responding

Check service status:

```bash
sudo systemctl status mem0

# Check if port is open
ss -tuln | grep 8000
```

### Memory Issues

Increase memory limit in systemd override:

```bash
sudo systemctl edit mem0

[Service]
MemoryMax=2G
```

## Related

- [mem0 Package](../../packages/mem0.md) - Package documentation
- [Port Management Guide](../../guides/port-management.md) - Using with port management
- [Using Modules Guide](../../guides/using-modules.md) - Module usage patterns