Architecture Overview

OMRP is designed as a robust and scalable architecture that facilitates efficient AI model discovery. This section provides a high-level overview of the system components and their interactions.

High-level System Components

  • Discovery Service: The core component responsible for matching user prompts with appropriate AI models.

  • Model Database (ModelDB): A repository of AI model information, including their capabilities and embeddings.

  • Embedding Service: Generates vector representations (embeddings) for both user queries and AI models.

  • Discovery Contract: A smart contract that implements the promptEval() function and the generative algorithm for model selection.

Interaction Flow

  1. User submits an encrypted prompt through their Smart Account.

  2. The OMRP receives the prompt and associated metadata.

  3. The Prompt Categorization Module classifies the prompt.

  4. The Discovery Engine, using the Generative Prompt Search Model, queries the ModelDB for suitable models.

  5. The Discovery Contract is called to finalize model selection.

  6. The user is presented with options, including free and paid models.

  7. Upon user selection, the chosen model processes the prompt.

  8. The response is stored in Decentralized Storage (IPFS) and returned to the user.

  9. Users can claim ownership of successful AIGC outputs.

  10. Claimed AIGC can be monetized through the Marketplace Contracts.

OMRP Workflow Overview

  1. An external system sends a user's prompt to the Discovery Service.

  2. The Discovery Service interacts with the Embedding Service to generate query embeddings.

  3. The Discovery Service retrieves relevant model embeddings from the ModelDB.

  4. Using the Discovery Protocol (OMRP), the system performs a k-NN search with cosine similarity to find the best matching models.

  5. The Discovery Contract is called to evaluate and finalize the model selection.

  6. The identifiers of the selected model(s) are returned to the external system.

Integration Points

OMRP is designed to integrate with external systems that handle:

  • User interaction and account management

  • Execution of selected AI models

  • Content generation and delivery

  • Proof creation and verification (if applicable)

  • Content ownership and rights management (if applicable)

This architecture ensures OMRP focuses solely on its core function of optimal model discovery, while allowing seamless integration with broader AI interaction ecosystems.

In the following sections, we will delve deeper into each of these components, explaining their roles, implementations, and interactions in greater detail, specifically within the context of OMRP's model discovery function.

Last updated