novyx-crewai

Persistent memory storage for CrewAI agents using Novyx Core.

Drop-in storage backend that gives your CrewAI agents persistent, semantic memory powered by Novyx RAM.

Installation

pip install novyx-crewai

Quick Start

1. Get an API Key

Sign up at novyxlabs.com to get your API key.

2. Use with CrewAI

import os
from crewai import Agent, Crew, Task
from novyx_crewai import NovyxStorage

# Create persistent storage backed by Novyx
storage = NovyxStorage(
    api_key=os.getenv("NOVYX_API_KEY"),
    memory_type="short_term",
)

# Create an agent
researcher = Agent(
    role="Researcher",
    goal="Find and summarize information",
    backstory="You are a research assistant.",
    verbose=True,
)

task = Task(
    description="Research the latest trends in AI agents.",
    expected_output="A short summary of trends.",
    agent=researcher,
)

# Build a crew with Novyx-backed memory
crew = Crew(
    agents=[researcher],
    tasks=[task],
    memory=True,
    short_term_memory=storage,
)

result = crew.kickoff()
print(result)

3. Standalone Usage

You can also use NovyxStorage directly without CrewAI:

from novyx_crewai import NovyxStorage

storage = NovyxStorage(api_key="nram_xxx")

# Save a memory
storage.save("The user prefers concise answers.", agent="researcher")

# Search memories
results = storage.search("user preferences", limit=5)
for r in results:
    print(f"[{r['score']:.2f}] {r['context']}")

# Clear all memories for this storage type
storage.reset()

API

NovyxStorage(api_key, memory_type="short_term", api_url=None)

Methods

• save(value, metadata=None, agent=None) -- Store a memory with optional metadata and agent tag.
• search(query, limit=10, score_threshold=0.5) -- Search memories. Returns list[dict] with "context" and "score" keys.
• reset() -- Delete all memories tagged with this storage's memory type.

License

MIT License - see LICENSE for details.

Links

• Documentation
• Novyx Core
• Pricing