mem-llm 1.0.10

Memory-enabled AI assistant with local LLM support

🧠 mem-llm# 🧠 mem-llm

Memory-enabled AI assistant that remembers conversations using local LLMs****Memory-enabled AI assistant that remembers conversations using local LLMs

---

🎯 What is mem-llm?## 📚 İçindekiler

mem-llm is a lightweight Python library that adds persistent memory to your local LLM chatbots. Each user gets their own conversation history that persists across sessions.- 🎯 mem-llm nedir?

• ⚡ Hızlı başlangıç

Use Cases:- 🧑‍🏫 Tutorial

• 💬 Customer service bots- 💡 Özellikler

• 🤖 Personal assistants- 📖 Kullanım örnekleri

• 📝 Context-aware applications- 🔧 Yapılandırma seçenekleri

• 🏢 Business automation solutions- 🗂 Bilgi tabanı ve dokümanlardan yapılandırma

• 🔥 Desteklenen modeller

---- 📦 Gereksinimler

• 🐛 Sık karşılaşılan problemler

⚡ Quick Start

---

1. Install the package

🎯 mem-llm nedir?

pip install mem-llm`mem-llm`, yerel bir LLM ile çalışan sohbet botlarınıza **kalıcı hafıza** kazandıran hafif bir Python kütüphanesidir. Her kullanıcı için ayrı bir konuşma geçmişi tutulur ve yapay zeka bu geçmişi bir sonraki oturumda otomatik olarak kullanır.

Nerelerde kullanılabilir?

2. Start Ollama and download a model (one-time setup)- 💬 Müşteri hizmetleri botları

• 🤖 Kişisel asistanlar

# Start Ollama service- 🏢 İş süreçlerini otomatikleştiren çözümler

ollama serve

---

# Download lightweight model (~2.5GB)

ollama pull granite4:tiny-h## ⚡ Hızlı başlangıç

0. Gereksinimleri kontrol edin

💡 Keep ollama serve running in one terminal, run your Python code in another.

• Python 3.8 veya üzeri

3. Create your first agent- Ollama kurulu ve çalışır durumda

• En az 4GB RAM ve 5GB disk alanı

from mem_llm import MemAgent### 1. Paketi yükleyin



# Create agent in one line```bash

agent = MemAgent()pip install mem-llm==1.0.7

Set user (each user gets separate memory)

agent.set_user("john")### 2. Ollama'yı başlatın ve modeli indirin (tek seferlik)

Chat with memory!```bash

response = agent.chat("My name is John")# Ollama servisini başlatın

print(response)ollama serve

response = agent.chat("What's my name?")# Yaklaşık 2.5GB'lık hafif modeli indirin

print(response) # Output: "Your name is John"ollama pull granite4:tiny-h

### 4. Verify your setup (optional)> 💡 Ollama `serve` komutu terminalde açık kalmalıdır. Yeni bir terminal sekmesinde Python kodunu çalıştırabilirsiniz.



```bash### 3. İlk ajanınızı çalıştırın

# Using CLI

mem-llm check```python

from mem_llm import MemAgent

# Or in Python

agent.check_setup()# Tek satırda ajan oluşturun

```agent = MemAgent()



---# Kullanıcıyı belirleyin (her kullanıcı için ayrı hafıza tutulur)

agent.set_user("john")

## 💡 Features

# Sohbet edin - hafıza devrede!

| Feature | Description |agent.chat("My name is John")

|---------|-------------|agent.chat("What's my name?")  # → "Your name is John"

| 🧠 **Memory** | Remembers each user's conversation history |```

| 👥 **Multi-user** | Separate memory for each user |

| 🔒 **Privacy** | 100% local, no cloud/API needed |### 4. Kurulumunuzu doğrulayın (isteğe bağlı)

| ⚡ **Fast** | Lightweight SQLite/JSON storage |

| 🎯 **Simple** | 3 lines of code to get started |```python

| 📚 **Knowledge Base** | Load information from documents |agent.check_setup()

| 🌍 **Multi-language** | Works with any language (Turkish, English, etc.) |# {'ollama': 'running', 'model': 'granite4:tiny-h', 'memory_backend': 'sql', ...}

| 🛠️ **CLI Tool** | Built-in command-line interface |```



---<<<<<<< HEAD

| Feature | Description |

## 📖 Usage Examples|---------|-------------|

| 🧠 **Memory** | Remembers each user's conversation history |

### Example 1: Basic Conversation with Memory| 👥 **Multi-user** | Separate memory for each user |

| 🔒 **Privacy** | 100% local, no cloud/API needed |

```python| ⚡ **Fast** | Lightweight SQLite/JSON storage |

from mem_llm import MemAgent| 🎯 **Simple** | 3 lines of code to get started |

| 📚 **Knowledge Base** | Config-free document integration |

# Create agent| 🌍 **Multi-language** | Works with any language |

print("🤖 Creating AI agent...")| 🛠️ **CLI Tool** | Built-in command-line interface |

agent = MemAgent()

---

# Set user

print("👤 Setting user: alice\n")## 🔄 Memory Backend Comparison

agent.set_user("alice")

Choose the right backend for your needs:

# First conversation

print("💬 User: I love pizza")| Feature | JSON Mode | SQL Mode |

response1 = agent.chat("I love pizza")|---------|-----------|----------|

print(f"🤖 Bot: {response1}\n")| **Setup** | ✅ Zero config | ⚙️ Minimal config |

| **Conversation Memory** | ✅ Yes | ✅ Yes |

# Memory test - bot remembers!| **User Profiles** | ✅ Yes | ✅ Yes |

print("💬 User: What's my favorite food?")| **Knowledge Base** | ❌ No | ✅ Yes |

response2 = agent.chat("What's my favorite food?")| **Advanced Search** | ❌ No | ✅ Yes |

print(f"🤖 Bot: {response2}")| **Multi-user Performance** | ⭐⭐ Good | ⭐⭐⭐ Excellent |

```| **Data Queries** | ❌ Limited | ✅ Full SQL |

| **Best For** | 🏠 Personal use | 🏢 Business use |

**Output:**

```**Recommendation:**

🤖 Creating AI agent...- **JSON Mode**: Perfect for personal assistants and quick prototypes

👤 Setting user: alice- **SQL Mode**: Ideal for customer service, multi-user apps, and production

=======

💬 User: I love pizzaKurulum sırasında sorun yaşarsanız [🐛 Sık karşılaşılan problemler](#-sık-karşılaşılan-problemler) bölümüne göz atın.

🤖 Bot: That's great! Pizza is a popular choice...>>>>>>> f002396c8c531e4cde33d19ac6a755494b1b30cd



💬 User: What's my favorite food?---

🤖 Bot: Based on our conversation, your favorite food is pizza!

```## 💡 Özellikler



---<<<<<<< HEAD

### Command Line Interface (CLI)

### Example 2: Multi-User Support

The easiest way to get started:

```python

from mem_llm import MemAgent```bash

# Install with CLI support

agent = MemAgent()pip install mem-llm



# Customer 1# Start interactive chat

print("=" * 60)mem-llm chat --user john

print("👤 Customer 1: John")

print("=" * 60)# Check system status

agent.set_user("customer_john")mem-llm check



print("💬 John: My order #12345 is delayed")# View statistics

response = agent.chat("My order #12345 is delayed")mem-llm stats

print(f"🤖 Bot: {response}\n")

# Export user data

# Customer 2 - SEPARATE MEMORY!mem-llm export john --format json --output data.json

print("=" * 60)

print("👤 Customer 2: Sarah")# Get help

print("=" * 60)mem-llm --help

agent.set_user("customer_sarah")```



print("💬 Sarah: I want to return item #67890")**Available CLI Commands:**

response = agent.chat("I want to return item #67890")

print(f"🤖 Bot: {response}\n")| Command | Description | Example |

|---------|-------------|---------|

# Back to Customer 1 - remembers previous conversation!| `chat` | Interactive chat session | `mem-llm chat --user alice` |

print("=" * 60)| `check` | Verify system setup | `mem-llm check` |

print("👤 Back to Customer 1: John")| `stats` | Show statistics | `mem-llm stats --user john` |

print("=" * 60)| `export` | Export user data | `mem-llm export john` |

agent.set_user("customer_john")| `clear` | Delete user data | `mem-llm clear john` |



print("💬 John: What was my order number?")### Basic Chat

response = agent.chat("What was my order number?")=======

print(f"🤖 Bot: {response}")| Özellik | Açıklama |

```|---------|----------|

| 🧠 **Kalıcı hafıza** | Her kullanıcının sohbet geçmişi saklanır |

**Output:**| 👥 **Çoklu kullanıcı** | Her kullanıcı için ayrı hafıza yönetimi |

```| 🔒 **Gizlilik** | Tamamen yerel çalışır, buluta veri göndermez |

============================================================| ⚡ **Hızlı** | Hafif SQLite veya JSON depolama seçenekleri |

👤 Customer 1: John| 🎯 **Kolay kullanım** | Üç satırda çalışan örnek |

============================================================| 📚 **Bilgi tabanı** | Ek yapılandırma olmadan dökümanlardan bilgi yükleme |

💬 John: My order #12345 is delayed| 🌍 **Türkçe desteği** | Türkçe diyaloglarda doğal sonuçlar |

🤖 Bot: I'll help you check your order status...| 🛠️ **Araç entegrasyonu** | Gelişmiş araç sistemi ile genişletilebilir |



============================================================---

👤 Customer 2: Sarah

============================================================## 🧑‍🏫 Tutorial

💬 Sarah: I want to return item #67890

🤖 Bot: I can help you with the return process...Tamamlanmış örnekleri adım adım incelemek için [examples](examples) klasöründeki rehberleri izleyebilirsiniz. Bu dizinde hem temel kullanım senaryoları hem de ileri seviye entegrasyonlar yer alır. Öne çıkan içerikler:



============================================================- [Basic usage walkthrough](examples/basic_usage.py) – ilk hafızalı ajanın nasıl oluşturulacağını gösterir.

👤 Back to Customer 1: John- [Customer support workflow](examples/customer_support.py) – çok kullanıcılı müşteri destek senaryosu.

============================================================- [Knowledge base ingestion](examples/knowledge_base.py) – dokümanlardan bilgi yükleme.

💬 John: What was my order number?

🤖 Bot: Your order number is #12345, which you mentioned was delayed.Her dosyada kodun yanında açıklamalar bulunur; komutları kopyalayıp çalıştırarak sonuçları deneyimleyebilirsiniz.

```

## 📖 Kullanım örnekleri

---

### Basic conversation

### Example 3: Turkish Language Support>>>>>>> f002396c8c531e4cde33d19ac6a755494b1b30cd



```python```python

from mem_llm import MemAgentfrom mem_llm import MemAgent



agent = MemAgent()agent = MemAgent()

agent.set_user("alice")

print("🇹🇷 Türkçe Konuşma Örneği")

print("=" * 60)# İlk konuşma

agent.chat("I love pizza")

agent.set_user("ahmet")

# Later on...

print("💬 Kullanıcı: Benim adım Ahmet ve İstanbul'da yaşıyorum")agent.chat("What's my favorite food?")

response = agent.chat("Benim adım Ahmet ve İstanbul'da yaşıyorum")# → "Your favorite food is pizza"

print(f"🤖 Bot: {response}\n")```



print("💬 Kullanıcı: Nerede yaşıyorum?")<<<<<<< HEAD

response = agent.chat("Nerede yaşıyorum?")### Multi-language Support

print(f"🤖 Bot: {response}\n")

```python

print("💬 Kullanıcı: Adımı hatırlıyor musun?")# Works with any language

response = agent.chat("Adımı hatırlıyor musun?")=======

print(f"🤖 Bot: {response}")### Turkish language support

```

```python

**Output:**# Handles Turkish dialogue naturally

```>>>>>>> f002396c8c531e4cde33d19ac6a755494b1b30cd

🇹🇷 Türkçe Konuşma Örneğiagent.set_user("ahmet")

============================================================agent.chat("Benim adım Ahmet ve pizza seviyorum")

💬 Kullanıcı: Benim adım Ahmet ve İstanbul'da yaşıyorumagent.chat("Adımı hatırlıyor musun?")

🤖 Bot: Memnun oldum Ahmet! İstanbul güzel bir şehir...# → "Evet, adınız Ahmet!"

```

💬 Kullanıcı: Nerede yaşıyorum?

🤖 Bot: İstanbul'da yaşıyorsunuz.### Customer service scenario



💬 Kullanıcı: Adımı hatırlıyor musun?```python

🤖 Bot: Evet, adınız Ahmet!agent = MemAgent()

```

# Müşteri 1

---agent.set_user("customer_001")

agent.chat("My order #12345 is delayed")

### Example 4: User Profile Extraction

# Customer 2 (separate memory!)

```pythonagent.set_user("customer_002")

from mem_llm import MemAgentagent.chat("I want to return item #67890")

```

agent = MemAgent()

agent.set_user("alice")### Inspecting the user profile



print("📝 Building user profile...")```python

print("=" * 60)# Retrieve automatically extracted user information

profile = agent.get_user_profile()

# Have natural conversations# {'name': 'Alice', 'favorite_food': 'pizza', 'location': 'NYC'}

conversations = [```

    "My name is Alice and I'm 28 years old",

    "I live in New York City",---

    "I work as a software engineer",

    "My favorite food is pizza"## 🔧 Yapılandırma seçenekleri

]

### JSON hafıza (varsayılan ve basit)

for msg in conversations:

    print(f"💬 User: {msg}")```python

    response = agent.chat(msg)agent = MemAgent(

    print(f"🤖 Bot: {response}\n")    model="granite4:tiny-h",

    use_sql=False,  # JSON dosyaları ile hafıza

# Extract profile automatically    memory_dir="memories"

print("=" * 60))

print("📊 Extracted User Profile:")```

print("=" * 60)

profile = agent.get_user_profile()### SQL hafıza (gelişmiş ve hızlı)



for key, value in profile.items():```python

    print(f"   {key}: {value}")agent = MemAgent(

```    model="granite4:tiny-h",

    use_sql=True,  # SQLite tabanlı hafıza

**Output:**    memory_dir="memories.db"

```)

📝 Building user profile...```

============================================================

💬 User: My name is Alice and I'm 28 years old### Diğer özelleştirmeler

🤖 Bot: Nice to meet you, Alice!...

```python

💬 User: I live in New York Cityagent = MemAgent(

🤖 Bot: New York City is a vibrant place...    model="llama2",  # Herhangi bir Ollama modeli

    ollama_url="http://localhost:11434"

💬 User: I work as a software engineer)

🤖 Bot: That's an interesting career...```



💬 User: My favorite food is pizza---

🤖 Bot: Pizza is delicious!...

## 📚 API referansı

============================================================

📊 Extracted User Profile:### `MemAgent`

============================================================

   name: Alice```python

   age: 28# Initialize

   location: New York Cityagent = MemAgent(model="granite4:tiny-h", use_sql=False)

   occupation: Software Engineer

   favorite_food: Pizza# Set active user

```agent.set_user(user_id: str, name: Optional[str] = None)



---# Chat

response = agent.chat(message: str, metadata: Optional[Dict] = None) -> str

### Example 5: Complete Customer Service Workflow

# Get profile

```pythonprofile = agent.get_user_profile(user_id: Optional[str] = None) -> Dict

from mem_llm import MemAgent

# System check

# Initialize customer service agentstatus = agent.check_setup() -> Dict

print("🏢 Customer Service Bot Initializing...")```

agent = MemAgent(use_sql=True)  # SQL for better performance

---

# Simulate customer support session

def handle_customer(customer_id, customer_name):## 🗂 Bilgi tabanı ve dokümanlardan yapılandırma

    print("\n" + "=" * 70)

    print(f"📞 New Customer Session: {customer_name} (ID: {customer_id})")Kurumsal dokümanlarınızdan otomatik `config.yaml` üretin:

    print("=" * 70)

    ```python

    agent.set_user(customer_id, name=customer_name)from mem_llm import create_config_from_document

    

    # Customer introduces issue# PDF'den config.yaml üretin

    print(f"\n💬 {customer_name}: Hi, my order hasn't arrived yet")create_config_from_document(

    response = agent.chat("Hi, my order hasn't arrived yet")    doc_path="company_info.pdf",

    print(f"🤖 Support: {response}")    output_path="config.yaml",

        company_name="Acme Corp"

    # Ask for details)

    print(f"\n💬 {customer_name}: My order number is #45678")

    response = agent.chat("My order number is #45678")# Oluşan yapılandırmayı kullanın

    print(f"🤖 Support: {response}")agent = MemAgent(config_file="config.yaml")

    ```

    # Follow up later in conversation

    print(f"\n💬 {customer_name}: Can you remind me what we were discussing?")---

    response = agent.chat("Can you remind me what we were discussing?")

    print(f"🤖 Support: {response}")## 🔥 Desteklenen modeller



# Handle multiple customers[Ollama](https://ollama.ai/) üzerindeki tüm modellerle çalışır. Tavsiye edilen modeller:

handle_customer("cust_001", "Emma")

handle_customer("cust_002", "Michael")| Model | Size | Speed | Quality |

|-------|------|-------|---------|

# Return to first customer - memory persists!| `granite4:tiny-h` | 2.5GB | ⚡⚡⚡ | ⭐⭐ |

print("\n" + "=" * 70)| `llama2` | 4GB | ⚡⚡ | ⭐⭐⭐ |

print("📞 Returning Customer: Emma (ID: cust_001)")| `mistral` | 4GB | ⚡⚡ | ⭐⭐⭐⭐ |

print("=" * 70)| `llama3` | 5GB | ⚡ | ⭐⭐⭐⭐⭐ |

agent.set_user("cust_001")

```bash

print("\n💬 Emma: What was my order number again?")ollama pull <model-name>

response = agent.chat("What was my order number again?")```

print(f"🤖 Support: {response}")

# Output: "Your order number is #45678"---

```

## 📦 Gereksinimler

**Output:**

```- Python 3.8+

🏢 Customer Service Bot Initializing...- Ollama (LLM için)

- Minimum 4GB RAM

======================================================================- 5GB disk alanı

📞 New Customer Session: Emma (ID: cust_001)

======================================================================**Kurulum ile gelen bağımlılıklar:**

- `requests >= 2.31.0`

💬 Emma: Hi, my order hasn't arrived yet- `pyyaml >= 6.0.1`

🤖 Support: I'm sorry to hear that. I'll help you track your order...- `sqlite3` (Python ile birlikte gelir)



💬 Emma: My order number is #45678---

🤖 Support: Thank you for providing order #45678. Let me check...

## 🐛 Sık karşılaşılan problemler

💬 Emma: Can you remind me what we were discussing?

🤖 Support: We're discussing your order #45678 that hasn't arrived yet...### Ollama çalışmıyor mu?



======================================================================```bash

📞 New Customer Session: Michael (ID: cust_002)ollama serve

======================================================================```



💬 Michael: Hi, my order hasn't arrived yet### Model bulunamadı hatası mı alıyorsunuz?

🤖 Support: I'm sorry to hear that. I'll help you track your order...

```bash

💬 Michael: My order number is #78901ollama pull granite4:tiny-h

🤖 Support: Thank you for providing order #78901...```



======================================================================### ImportError veya bağlantı hatası mı var?

📞 Returning Customer: Emma (ID: cust_001)

======================================================================```bash

pip install --upgrade mem-llm

💬 Emma: What was my order number again?```

🤖 Support: Your order number is #45678.

```> Hâlâ sorun yaşıyorsanız `agent.check_setup()` çıktısını ve hata mesajını issue açarken paylaşın.



------



## 🔧 Configuration Options## 📄 Lisans



### JSON Memory (Simple, Default)MIT Lisansı — kişisel veya ticari projelerinizde özgürce kullanabilirsiniz.



```python---

agent = MemAgent(

    model="granite4:tiny-h",## 🔗 Faydalı bağlantılar

    use_sql=False,  # JSON file-based memory

    memory_dir="memories"- **PyPI:** https://pypi.org/project/mem-llm/

)- **GitHub:** https://github.com/emredeveloper/Mem-LLM

```- **Ollama:** https://ollama.ai/



### SQL Memory (Advanced, Recommended for Production)---



```python## 🌟 Bize destek olun

agent = MemAgent(

    model="granite4:tiny-h",Proje işinize yaradıysa [GitHub](https://github.com/emredeveloper/Mem-LLM) üzerinden ⭐ vermeyi unutmayın!

    use_sql=True,  # SQLite-based memory

    memory_dir="memories.db"---

)

```<div align="center">

Sevgiyle geliştirildi — <a href="https://github.com/emredeveloper">C. Emre Karataş</a>

### Custom Configuration</div>


```python
agent = MemAgent(
    model="llama2",  # Any Ollama model
    ollama_url="http://localhost:11434",
    check_connection=True  # Verify setup on startup
)
```

---

## 🛠️ Command Line Interface

```bash
# Start interactive chat
mem-llm chat --user john

# Check system status
mem-llm check

# View statistics
mem-llm stats

# Export user data
mem-llm export john --format json

# Clear user data
mem-llm clear john

# Get help
mem-llm --help
```

---

## 🔄 Memory Backend Comparison

| Feature | JSON Mode | SQL Mode |
|---------|-----------|----------|
| **Setup** | ✅ Zero config | ⚙️ Minimal config |
| **Conversation Memory** | ✅ Yes | ✅ Yes |
| **User Profiles** | ✅ Yes | ✅ Yes |
| **Knowledge Base** | ❌ No | ✅ Yes |
| **Advanced Search** | ❌ No | ✅ Yes |
| **Multi-user Performance** | ⭐⭐ Good | ⭐⭐⭐ Excellent |
| **Best For** | 🏠 Personal use | 🏢 Business use |

**Recommendation:**
- **JSON Mode**: Perfect for personal assistants and quick prototypes
- **SQL Mode**: Ideal for customer service, multi-user apps, and production

---

## 📚 API Reference

### MemAgent Class

```python
# Initialize
agent = MemAgent(
    model="granite4:tiny-h",
    use_sql=True,
    memory_dir=None,
    ollama_url="http://localhost:11434",
    check_connection=False
)

# Set active user
agent.set_user(user_id: str, name: Optional[str] = None)

# Chat (returns response string)
response = agent.chat(message: str, metadata: Optional[Dict] = None) -> str

# Get user profile (auto-extracted from conversations)
profile = agent.get_user_profile(user_id: Optional[str] = None) -> Dict

# System check
status = agent.check_setup() -> Dict
```

---

## 🔥 Supported Models

Works with any [Ollama](https://ollama.ai/) model. Recommended models:

| Model | Size | Speed | Quality | Best For |
|-------|------|-------|---------|----------|
| `granite4:tiny-h` | 2.5GB | ⚡⚡⚡ | ⭐⭐ | Quick testing |
| `llama2` | 4GB | ⚡⚡ | ⭐⭐⭐ | General use |
| `mistral` | 4GB | ⚡⚡ | ⭐⭐⭐⭐ | Balanced |
| `llama3` | 5GB | ⚡ | ⭐⭐⭐⭐⭐ | Best quality |

```bash
# Download a model
ollama pull <model-name>

# List installed models
ollama list
```

---

## 📦 Requirements

- Python 3.8+
- [Ollama](https://ollama.ai/) (for LLM)
- Minimum 4GB RAM
- 5GB disk space

**Python Dependencies (auto-installed):**
- `requests >= 2.31.0`
- `pyyaml >= 6.0.1`
- `click >= 8.1.0`

---

## 🐛 Troubleshooting

### Ollama not running?

```bash
ollama serve
```

### Model not found error?

```bash
# Download the model
ollama pull granite4:tiny-h

# Check installed models
ollama list
```

### Connection error?

```bash
# Check if Ollama is running
curl http://localhost:11434

# Restart Ollama
ollama serve
```

### Import error?

```bash
# Upgrade to latest version
pip install --upgrade mem-llm
```

> If issues persist, run `mem-llm check` or `agent.check_setup()` and share the output when opening an issue.

---

## 📄 License

MIT License - Free to use in personal and commercial projects.

---

## 🔗 Links

- **PyPI:** https://pypi.org/project/mem-llm/
- **GitHub:** https://github.com/emredeveloper/Mem-LLM
- **Ollama:** https://ollama.ai/
- **Documentation:** [GitHub Wiki](https://github.com/emredeveloper/Mem-LLM/wiki)

---

## 🌟 Support Us

If you find this project useful, please ⭐ [star it on GitHub](https://github.com/emredeveloper/Mem-LLM)!

---

## 🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (`git checkout -b feature/AmazingFeature`)
3. Commit your changes (`git commit -m 'Add some AmazingFeature'`)
4. Push to the branch (`git push origin feature/AmazingFeature`)
5. Open a Pull Request

---

<div align="center">
Made with ❤️ by <a href="https://github.com/emredeveloper">C. Emre Karataş</a>
</div>