Interaction Flow Overview
1.User Initiates Model Call Request:
The user selects the desired AI model (such as identity verification, credit scoring, image recognition, etc.) from a supported platform or DApp.
The input data is encrypted and fragmented locally before being submitted.
2.Authorization and Task Submission:
The user signs the authorization and submits the "task invocation transaction + data pointer" to the Slor smart contract.
Service fees are paid (either in SLOR tokens or USDT).
3.Model Availability Verification:
The contract verifies the model's availability and the user's permission before accepting the task.
4.Task Scheduling and Node Distribution:
The Slor network dispatches the task across multiple nodes via the on-chain task scheduling contract.
Nodes are selected based on their reputation scores, response times, and geographical latency.
5.MPC Parallel Collaborative Inference:
The MPC nodes perform parallel inference and invoke other large models as necessary for scalability.
Each node independently processes its shard of the data, ensuring that no node has access to the raw data.
Nodes do not share intermediate variables, only performing secure local computations.
Final results are aggregated using cryptographic protocols.
6.Result On-Chain Record:
The system generates a hash of the aggregated result and the invocation proof.
A complete execution record (including node signatures, steps, and CID) is written to the on-chain smart contract for transparency and auditing.
7.User Receives Results and Node Revenue Settlement:
The user views the returned result via their wallet or invocation record.
The system automatically distributes the fees to:
MPC nodes that participated in the execution.
Model developers.
Network maintainers (such as foundation fees).
Last updated