langchain-superlocalmemory

LangChain integration for SuperLocalMemory V3 — local-first AI memory with mathematical foundations.

Features

• SuperLocalMemoryChatHistory — Drop-in BaseChatMessageHistory for conversation persistence across sessions
• SuperLocalMemoryRetriever — BaseRetriever for RAG-style memory augmentation with 4-channel retrieval
• Direct Python API — No subprocess calls, no CLI dependency at runtime
• Privacy-first — All data stays on your device (Mode A: zero cloud)

Installation

pip install langchain-superlocalmemory

Requires SuperLocalMemory V3 to be installed:

pip install "superlocalmemory[search]"
# or
npm install -g superlocalmemory && slm setup

Usage

Chat Message History

from langchain_superlocalmemory import SuperLocalMemoryChatHistory
from langchain_core.runnables.history import RunnableWithMessageHistory

# Create history for a session
history = SuperLocalMemoryChatHistory(session_id="my-project")

# Use with RunnableWithMessageHistory
chain_with_memory = RunnableWithMessageHistory(
    chain,
    lambda session_id: SuperLocalMemoryChatHistory(session_id=session_id),
    input_messages_key="input",
    history_messages_key="history",
)

Memory Retriever (RAG)

from langchain_superlocalmemory import SuperLocalMemoryRetriever

retriever = SuperLocalMemoryRetriever(k=5, score_threshold=0.3)
docs = retriever.invoke("authentication middleware patterns")

for doc in docs:
    print(f"{doc.page_content} (score: {doc.metadata['score']:.2f})")

Modes

# EU AI Act compliant (default)
history = SuperLocalMemoryChatHistory(session_id="s1", mode="a")

# Full power
history = SuperLocalMemoryChatHistory(session_id="s1", mode="c")

Links

• SuperLocalMemory
• Paper: arXiv:2603.14588
• GitHub

License

MIT

Part of Qualixar | Author: Varun Pratap Bhardwaj