novyx-llamaindex

Persistent memory + rollback + audit trail for LlamaIndex agents, powered by Novyx Core.

Install

pip install novyx-llamaindex

Quick Start

ChatStore (drop-in for SimpleChatStore)

from novyx_llamaindex import NovyxChatStore
from llama_index.core.memory import ChatMemoryBuffer

chat_store = NovyxChatStore(api_key="nram_your_key")
memory = ChatMemoryBuffer.from_defaults(
    chat_store=chat_store,
    chat_store_key="user-123",
    token_limit=3000,
)

BaseMemory (semantic recall)

from novyx_llamaindex import NovyxMemory

memory = NovyxMemory.from_defaults(
    api_key="nram_your_key",
    session_id="user-123",
    top_k=5,
)

# Use with any LlamaIndex chat engine
engine = SimpleChatEngine.from_defaults(memory=memory)
response = engine.chat("Hello!")

Retriever (RAG pipeline)

from novyx_llamaindex import NovyxRetriever

retriever = NovyxRetriever(api_key="nram_your_key", top_k=5)
nodes = retriever.retrieve("What are the user's preferences?")

Features

• Persistent: Memories survive across sessions, restarts, and deployments
• Semantic: recall() finds relevant memories by meaning, not just keywords
• Rollback: Undo memory mistakes with memory.rollback("2 hours ago")
• Audit Trail: Every operation is SHA-256 hashed and logged
• Cross-session: Agent A learns it, Agent B knows it (via Context Spaces)

API Reference

License

MIT