Skip to main content

Key Concepts

This section outlines the foundational elements and technologies that constitute the Shinkai ecosystem. Understanding these concepts is crucial for interacting with and leveraging the Shinkai network effectively.

Core Components​

  • Shinkai Node: The central unit of the Shinkai Network, handling tasks such as agent management, job processing, and secure communications.
  • Shinkai Identity: A unique identifier for users or nodes on the Shinkai Network, established on-chain as an NFT through Shinkai tokens.
    • Sub-identities: Underneath a single Shinkai identity, an infinite number of sub-identities can be created for all devices, computers, and AI agents that the user owns, enhancing the flexibility and scalability of the Shinkai Network.
  • Shinkai Message: The primary method of secure communication within the Shinkai ecosystem, ensuring data integrity and confidentiality.
  • Vector File System (VectorFS): A specialized file system for efficient management and querying of AI embeddings.
    • Merkle Trees: VectorFS implements full merkelization, where every folder and item stored has its own merkle root, enabling data integrity verification and efficient syncing and distribution optimizations in the decentralized network.
  • Decentralized Network: The peer-to-peer infrastructure enabling real-time AI data sharing and agent communication.
    • Router/Proxy Nodes: By specifying one or more router/proxy nodes, the network supports further scalability, enhanced message delivery assurance, and privacy.

Advanced Technologies​

  • Vector Resource: A component of VectorFS that stores data embeddings and their source documents.
  • Retrieval-Augmented Generation (RAG): A technique for enhancing AI responses by dynamically integrating up-to-date data.
  • Zero-Knowledge Multi-Party Computation (zk-MPC): A protocol ensuring data embedding authenticity while preserving privacy.
  • Containerized Tooling System: A framework for developing and deploying modular toolkits to extend LLM Agents' capabilities, consisting of:
    • Toolkit-Builder: A scaffolding tool for kickstarting toolkit development.
    • Toolkit-Executor: Facilitates the execution of toolkits in various modes.
    • Toolkit-Lib: Provides core functionalities for toolkit development.
    • Toolkit-Registry: A centralized repository for discovering and integrating toolkits.
  • Planning and Execution Runtime: An environment within Shinkai nodes for sophisticated AI planning and task execution.

Security and Encryption​

Keypairs​

Shinkai utilizes several types of keypairs for security and encryption:

  • Global Identity Keypairs: For user identity verification on-chain.
  • Device Keypairs: Unique to each device for secure device registration and communication.
  • E2E Keypairs: Shared across all devices for end-to-end encrypted messaging.

Global Identity Keypairs​

Upon staking Shinkai tokens, users generate these keypairs to register their global identity. The public keys are submitted on-chain, while the private keys are securely stored and used by the Shinkai node.

Device Keypairs​

Generated during device registration, these keypairs are unique to each device and facilitate secure device-to-node communication.

E2E Keypairs​

Used for encrypting messages between devices, ensuring privacy even from the node itself. These keypairs are shared among all devices under a single profile.

Messaging and Data Management​

Inboxes​

Inboxes serve as the primary method for message and data exchange within the Shinkai network, supporting both direct messages and group chats. They can be E2E encrypted for privacy and are accessible to devices, agents, and global identities with appropriate permissions.

Vector Resources and VectorFS​

VectorFS and its Vector Resources enable the storage and retrieval of AI embeddings, crucial for the network's AI-driven operations. This system supports the efficient handling of large-scale vector data.

Decentralized Network and Protocols​

The Shinkai ecosystem is built on a decentralized network, facilitating secure and real-time data sharing. Protocols like zk-MPC ensure data authenticity without compromising privacy, while the containerized tooling system and planning and execution runtime enable the development and execution of complex AI tasks.

Understanding these key concepts is essential for navigating and leveraging the Shinkai ecosystem to its full potential.