v0.2.0-beta.1

Kive v0.2.0-beta.1 Release Notes

Release Date: December 13, 2025

---

Overview

Kive v0.2.0 is a major milestone that establishes Kive as the universal adapter for AI memory systems. This release focuses on providing a unified, production-ready API that seamlessly bridges cloud and local memory providers, making it easier than ever to build memory-enabled AI applications.

---

What's New

Core Features

Unified Memory API

• One API, Multiple Backends - Switch between 11 memory providers without changing a single line of code
• Cloud & Local Support - Use the same interface for managed cloud services or self-hosted local deployments
• Consistent CRUD Operations - Add, search, get, update, and delete memories with identical methods across all providers

Multi-Tenancy Built-in

• 5-Level Context Isolation - Support for tenant, app, AI agent, user, and session-level isolation
• Out-of-the-box Multi-tenancy - Built for B2B SaaS, multi-product platforms, and multi-agent systems
• Flexible Context Parameters - Use any combination of tenant_id, app_id, ai_id, user_id, and session_id

---

Supported Providers

Cloud Providers (7)

Managed services - just provide API key:

• Mem0 Cloud - Fast vector search with real-time queries and auto-extraction
• Zep Cloud - Conversational AI with session management and fact extraction
• SuperMemory Cloud - Document memory with PDF/Web ingestion and semantic search
• Cognee Cloud - Knowledge graphs with deep reasoning and entity linking
• Memobase Cloud - Personal memory with user profiles and context building
• Memos Cloud - Factual memory with fact extraction and preference tracking
• MemU Cloud - Category memory with auto-categorization and summaries

Local Providers (4)

Self-hosted - full control over data:

• Mem0 Local - ChromaDB + Kuzu for local-first vector search
• Cognee Local - Local graph database with batch processing
• Graphiti Local - Temporal graphs with time-aware facts and episodic memory
• Memos Local - Local fact extraction and preference tracking

---

Key Improvements

Developer Experience

Simplified Initialization

from kive import Memory

# Cloud provider - just API key
memory = Memory("cloud/mem0", api_key="m0-xxx")

# Local provider - full configuration
memory = Memory(
    "local/mem0",
    llm_provider="openai",
    llm_model="gpt-4",
    llm_api_key="YOUR_KEY",
    embedding_provider="openai",
    embedding_model="text-embedding-3-small",
    embedding_api_key="YOUR_KEY",
    vector_db_provider="chroma",
    graph_db_provider="kuzu",
)

Flexible Content Input

• Text content - Simple string input for quick memory storage
• Conversation messages - Native support for chat message arrays
• Rich metadata - Attach custom metadata for advanced filtering

Comprehensive Examples

• 7 cloud provider examples in examples/cloud_memory/
• 4 local provider examples in examples/local_memory/
• Complete server example with server_example.py

---

Technical Details

Installation Options

# Basic installation
pip install kive

# Specific provider
pip install kive[mem0-cloud]
pip install kive[graphiti-local]

# All cloud providers
pip install kive[cloud]

# All local providers
pip install kive[local]

# Everything
pip install kive[all]

Dependencies

• Core: pydantic>=2.0.0 for robust data validation
• Python Support: Python 3.8, 3.10, 3.11, 3.12
• Provider-specific: Modular dependencies loaded only when needed

---

Use Cases

Kive v0.2.0 is ideal for:

• Multi-tenant AI Applications - Build B2B SaaS with organization-level isolation
• Conversational AI Systems - Maintain context across sessions and users
• Knowledge Management - Store and retrieve factual information with semantic search
• Multi-agent Systems - Separate memory spaces for different AI agents
• Hybrid Deployments - Start with cloud, migrate to local, or use both

---

Breaking Changes

This is the initial production release (v0.2.0), so there are no breaking changes from previous versions.

---

Documentation

• README: Comprehensive guide with examples
• Examples Directory: 17+ working examples for all providers
• Provider Comparison: Detailed feature comparison table
• API Reference: Complete parameter documentation for cloud and local providers

---

What's Next

Future roadmap includes:

• Additional memory provider integrations
• Advanced query capabilities (filtering, aggregation)
• Batch operations support
• Memory analytics and insights
• Performance optimizations

---

Acknowledgments

Special thanks to all the amazing memory provider teams whose work makes Kive possible:

• Mem0, Zep, SuperMemory, Cognee, Memobase, Memos, MemU, Graphiti

---

Get Started

pip install kive[mem0-cloud]

import asyncio
from kive import Memory

async def main():
    memory = Memory("cloud/mem0", api_key="YOUR_API_KEY")
    
    result = await memory.add(
        content="Python is a programming language",
        user_id="user_123"
    )
    print(f"Added: {result.id}")
    
    results = await memory.search(
        query="what is Python?",
        user_id="user_123"
    )
    
    for memo in results.results:
        print(f"- {memo.content}")

asyncio.run(main())

---

Links

• GitHub: https://github.com/zhixiangxue/kive-ai
• PyPI: https://pypi.org/project/kive/
• Documentation: https://github.com/zhixiangxue/kive-ai#readme
• Issues: https://github.com/zhixiangxue/kive-ai/issues

---

init release

init release