LlamaIndex Chat Store — SuperLocalMemory V2

A LlamaIndex BaseChatStore implementation backed by SuperLocalMemory V2. All chat history stays 100% local on your machine — zero cloud calls, zero telemetry, zero API keys.

Prerequisites

• Python 3.9+
• SuperLocalMemory V2 installed (~/.claude-memory/ must exist)

# Install SuperLocalMemory V2 (one-time)
curl -fsSL https://raw.githubusercontent.com/varun369/SuperLocalMemoryV2/main/install.sh | bash

Installation

pip install llama-index-storage-chat-store-superlocalmemory

Quick Start

from llama_index.core.memory import ChatMemoryBuffer
from llama_index.core.base.llms.types import ChatMessage, MessageRole
from llama_index.storage.chat_store.superlocalmemory import SuperLocalMemoryChatStore

# Create the chat store (uses default SLM database)
chat_store = SuperLocalMemoryChatStore()

# Use with ChatMemoryBuffer for automatic conversation management
memory = ChatMemoryBuffer.from_defaults(
    chat_store=chat_store,
    chat_store_key="user-123",
    token_limit=3000,
)

# Or use directly for manual message management
chat_store.add_message("session-1", ChatMessage(role=MessageRole.USER, content="Hello!"))
chat_store.add_message("session-1", ChatMessage(role=MessageRole.ASSISTANT, content="Hi there!"))

messages = chat_store.get_messages("session-1")
print(messages)  # [ChatMessage(role=user, content="Hello!"), ChatMessage(role=assistant, content="Hi there!")]

# List all session keys
keys = chat_store.get_keys()

# Delete a session
chat_store.delete_messages("session-1")

Features

• 100% Local — All data stored in SQLite at ~/.claude-memory/memory.db
• Zero Cloud — No API keys, no subscriptions, no data leaves your machine
• Shared Memory — Chat history is accessible from Claude, Cursor, Windsurf, and 16+ other AI tools via SuperLocalMemory
• Session Isolation — Each chat key is cleanly isolated using SLM tags
• Persistent — Survives process restarts (SQLite-backed, not in-memory)
• Full BaseChatStore API — set_messages, get_messages, add_message, delete_messages, delete_message, delete_last_message, get_keys
• Async Support — Async methods inherited from BaseChatStore (delegates to sync via asyncio.to_thread)

How It Works

Each chat message is stored as a separate memory entry in SuperLocalMemory V2:

• Content: JSON-serialized {role, content, additional_kwargs}
• Tag: llamaindex:chat:<session_key> for session isolation
• Project: llamaindex for easy identification
• Importance: 3 (low, since chat messages are transient)

Custom Database Path

# Use a custom database file
chat_store = SuperLocalMemoryChatStore(db_path="/path/to/custom/memory.db")

Links

• SuperLocalMemory V2
• LlamaIndex Documentation
• LlamaIndex Chat Stores Guide