Mem-LLM

Mem-LLM is a privacy-first Python framework for building memory-enabled AI assistants that run locally.

What's New in v2.4.6

• Fixed critical memory, tool parsing, and backend compatibility issues.
• Improved SQL ordering and thread-safety behavior.
• Added missing runtime dependencies (psutil, networkx).
• Updated backend defaults:
  • Ollama: granite4:3b
  • LM Studio: google/gemma-3-12b

Quick Start

Install

pip install mem-llm

Ollama

from mem_llm import MemAgent

agent = MemAgent(backend="ollama", model="granite4:3b")
agent.set_user("alice")
print(agent.chat("My name is Alice."))
print(agent.chat("What is my name?"))

LM Studio

from mem_llm import MemAgent

agent = MemAgent(backend="lmstudio", model="google/gemma-3-12b")
agent.set_user("alice")
print(agent.chat("Summarize Python in one sentence."))

Core Features

• Persistent memory per user (JSON or SQLite)
• Multi-backend support (Ollama, LM Studio)
• Tool calling system (@tool, built-in tools, validation)
• Streaming responses
• Knowledge base integration
• Conversation analytics
• REST API and Web UI

Repository Layout

• Memory LLM/ - main package source and release files
• quickstart/ - step-by-step usage examples & tutorials

Links

• PyPI: https://pypi.org/project/mem-llm/
• Documentation: Memory LLM/README.md
• Changelog: Memory LLM/CHANGELOG.md
• Issues: https://github.com/emredeveloper/Mem-LLM/issues

License

Mem-LLM is released under the MIT License.