> For the complete documentation index, see [llms.txt](https://griffinai.gitbook.io/griffinai-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://griffinai.gitbook.io/griffinai-docs/scope-of-ai-services.md).

# Scope of AI Services

Griffin AI introduces a decentralized network architecture designed to host a broad spectrum of AI services and middleware components. This network provides the infrastructure for efficient and secure AI interactions within a decentralized framework.

At the heart of the network are the AI models, surrounded by a suite of essential services required to build robust end-user applications — middleware components, datasets and libraries, and computation and storage solutions. An overview of high-level service categories is provided below.

| Category                | Description                                                                                                                                                                                                                                                                                                                  |
| ----------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| AI Models               | Includes a diverse array of AI offerings such as open-source and proprietary Large Language Models (LLMs), specialized ML models, and models for image and video generation, text-to-voice conversion, etc. Also provides tailored solutions designed for decentralized applications, catering to blockchain-specific needs. |
| Middleware Services     | Provides support for seamless AI operations: advanced data retrieval (Vector Search, Web Search), context and prompt management, arithmetic computation capabilities, and verification tools that operate both on-chain and off-chain. This layer connects raw AI functionalities with practical, application-focused uses.  |
| Data Sets and Libraries | Comprehensive datasets and libraries crucial for blockchain-centric AI applications. These provide standardized, blockchain-focused datasets to ensure models are well-trained and finely tuned, and they can also be leveraged for retrieval augmentation.                                                                  |
| Computation and Storage | Flexible computation and storage options, including centralized cloud providers and decentralized blockchain-based networks. This enables optimization for cost, speed, and degree of decentralization appropriate for each application.                                                                                     |

Table 1: AI Service Categories

While initial prototypes for AI services are developed by Griffin AI, a broad ecosystem of external service providers is being established. Providers include large-scale enterprises, established projects, decentralized autonomous organizations (DAOs), and individual engineers. This diversity ensures a rich pool of services and capabilities.

Providers may offer specialized services or comprehensive suites of interconnected functionalities, enabling users to tailor AI solutions. For example, users can combine blockchain-specific data and processing services from Griffin AI with their own LLMs or models from external providers to create customized solutions.

The ecosystem supports both consumption and contribution: entities within Griffin AI can offer services as well as consume them, fostering collaboration and continuous improvement.

Looking ahead, Griffin AI envisions an expansive marketplace where third-party DApps and AI Agents are built on and interact with this foundational network. These entities will be able to communicate and collaborate directly, enabling dynamic multi-actor workflows and collective goal achievement.

Last updated 1 year ago