Service Discovery

In the Griffin AI network, each Service Provider (SP) or Client Provider (CP) must start by creating a Griffin ID. This Root Identity is then utilized to register a new service through one of the Griffin Registration Smart Contracts. The network maintains specific registries tailored for various entities: a Service Registry for SPs and CPs operating full nodes, and an Agent Registry for AI Agents. This service registration ensures that each new SP, CP, or agent is discoverable in the network and their details are securely logged on the blockchain.

During registration, service providers and agents must provide comprehensive information about their offerings. This data includes:

  • Basic Service Information: Service name, owner / provider details, description.

  • Service Classification: Standardized tags that define the type of service, such as image processing, video processing, chatbots, or natural language processing (NLP).

  • Documentation: Link to IPFS-hosted documentation on Griffin distributed database providing further details such as marketing materials, examples of outputs, technical specifications of the model, and service level agreements (SLAs).

Changes to the services provided necessitate that providers update their documentation in the Griffin distributed database and submit a mutation log to the Griffin Registration Smart Contracts. This process ensures that the information remains accurate and current. Additionally, Guardian nodes oversee and verify updates to the registry, maintaining the integrity and reliability of the system.

Client Providers (CPs) enhance the discoverability of services by creating advanced indices of all service records. They may develop and manage Marketplace-like web applications with features for advanced searching, pinning, and tagging of services. These marketplaces are the main tool for users to browse the service and agent registry, allowing them to find, evaluate, and select services and agents that meet their requirements.

Type of Data
Data Location
Modification Procedure

Service Provider Identity

Service Registry (Smart Contract).

Once created, the Service Provider identity is not modifiable.

Basic service information

Griffin Distributed Database

Service information can be modified at will by holders of SP identification keys, utilizing Griffin node functionality.

Service Documentation

IPFS

IPFS storage is owned by its author but cannot be modified. Modification of IPFS data effectively creates a new data set with a new unique link.

Link to Service Documentation

Service Registry (Smart Contract)

Service Documentation Link can be modified at will by holders of SP identification keys by issuing a modification transaction in Griffin Service Registry smart contract.

Table 6: Service Data Management

  1. Reputation System

The Griffin AI network operates on a permissionless basis, allowing any entity to offer services, operate nodes, or deploy agents. To ensure accountability and build trust, the network assigns unique identities to participants, verifies ownership of these identities through private keys, and records their reliability in a centralized reputation system.

Guardian nodes are central to the management of the Griffin AI reputation system. These specialized nodes are responsible for recording and assessing the performance of service providers and AI agents, ensuring that all activities are transparent and accountable. To prevent conflicts of interest and promote impartiality, only Guardian nodes have the authority to write to the reputation system within the Griffin database.

Guardian nodes maintain and verify the accuracy of performance metrics, such as service availability and response times. They also aggregate reputation scores provided by network participants. In terms of dispute resolution, Guardian nodes are responsible for addressing and settling conflicts between service providers and users. They have the authority to modify reputation scores, retract negative feedback, or impose bans on Root Identities based on the outcome of their review.

While the permissionless nature of the network prevents the physical disconnection of non-compliant operators, these operators can be excluded from the Griffin reputation system. This exclusion acts as a deterrent, making them less visible to new users and signaling potential risks to existing users. That said, Client Providers (CPs) may establish alternative reputation systems with different standards and procedures, tailored to meet the specific needs of their services or community. These systems can vary in their emphasis on security and trust, potentially deviating from the standards set by the Griffin reputation system.

This approach to identity verification and the implementation of the Griffin reputation system create a secure and reliable environment for all users and agents on the Griffin AI network. Alongside the network's permissionless principles, this setup allows third parties, such as Client Providers (CPs), to operate their own reputation systems. This flexibility supports a censorship-resistant and open network, accommodating diverse needs and enhancing the overall ecosystem.

Last updated