mem-llm 1.0.5

Memory-enabled AI assistant with local LLM support

🧠 mem-llm

Memory-enabled AI assistant that remembers conversations using local LLMs

---

🎯 What is it?

A lightweight Python library that adds persistent memory to local LLM chatbots. Each user gets their own conversation history that the AI remembers across sessions.

Perfect for:

• 💬 Customer service chatbots
• 🤖 Personal AI assistants
• 📝 Context-aware applications
• 🏢 Business automation

---

⚡ Quick Start

1. Install

pip install mem-llm

2. Setup Ollama (one-time)

# Install: https://ollama.ai/download
ollama serve

# Download model (only 2.5GB)
ollama pull granite4:tiny-h

3. Use

from mem_llm import MemAgent

# Create agent (one line!)
agent = MemAgent()

# Set user
agent.set_user("john")

# Chat - it remembers!
agent.chat("My name is John")
agent.chat("What's my name?")  # → "Your name is John"

---

💡 Features

Feature	Description
🧠 Memory	Remembers each user's conversation history
👥 Multi-user	Separate memory for each user
🔒 Privacy	100% local, no cloud/API needed
⚡ Fast	Lightweight SQLite/JSON storage
🎯 Simple	3 lines of code to get started

---

📖 Usage Examples

Basic Chat

from mem_llm import MemAgent

agent = MemAgent()
agent.set_user("alice")

# First conversation
agent.chat("I love pizza")

# Later...
agent.chat("What's my favorite food?")
# → "Your favorite food is pizza"

Customer Service Bot

agent = MemAgent()

# Customer 1
agent.set_user("customer_001")
agent.chat("My order #12345 is delayed")

# Customer 2 (different memory!)
agent.set_user("customer_002")
agent.chat("I want to return item #67890")

Check User Profile

# Get automatically extracted user info
profile = agent.get_user_profile()
# {'name': 'Alice', 'favorite_food': 'pizza', 'location': 'NYC'}

---

🔧 Configuration

JSON Memory (default - simple)

agent = MemAgent(
    model="granite4:tiny-h",
    use_sql=False,  # Use JSON files
    memory_dir="memories"
)

SQL Memory (advanced - faster)

agent = MemAgent(
    model="granite4:tiny-h",
    use_sql=True,  # Use SQLite
    memory_dir="memories.db"
)

Custom Settings

agent = MemAgent(
    model="llama2",  # Any Ollama model
    ollama_url="http://localhost:11434"
)

---

📚 API Reference

MemAgent

# Initialize
agent = MemAgent(model="granite4:tiny-h", use_sql=False)

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

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

# Get profile
profile = agent.get_user_profile(user_id: Optional[str] = None) -> Dict

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

---

🎨 Advanced: PDF/DOCX Config

Generate config from business documents:

from mem_llm import create_config_from_document

# Create config.yaml from PDF
create_config_from_document(
    doc_path="company_info.pdf",
    output_path="config.yaml",
    company_name="Acme Corp"
)

# Use config
agent = MemAgent(config_file="config.yaml")

---

🔥 Models

Works with any Ollama model:

Model	Size	Speed	Quality
granite4:tiny-h	2.5GB	⚡⚡⚡	⭐⭐
llama2	4GB	⚡⚡	⭐⭐⭐
mistral	4GB	⚡⚡	⭐⭐⭐⭐
llama3	5GB	⚡	⭐⭐⭐⭐⭐

ollama pull <model-name>

---

📦 Requirements

• Python 3.8+
• Ollama (for LLM)
• 4GB RAM minimum
• 5GB disk space

Dependencies (auto-installed):

• requests >= 2.31.0
• pyyaml >= 6.0.1

---

🐛 Troubleshooting

Ollama not running?

ollama serve

Model not found?

ollama pull granite4:tiny-h

Import error?

pip install mem-llm --upgrade

---

📄 License

MIT License - feel 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/

---

🌟 Star us on GitHub!

If you find this useful, give us a ⭐ on GitHub!

---

Made with ❤️ by C. Emre Karataş