Core Functions

$DGAI binds the DGrid.AI ecosystem together, fostering a self-sustaining cycle ofparticipation: Users pay $DGAI for inference, node operators earn $DGAI for contributions, and staking/governance ensure long-term network health. By embedding $DGAI into every interaction—from task submission to result verification—we create a decentralized AI economy where value flows directly between contributors and users, aligned with our core philosophy: "AI should serve the open network."

$DGAI powers every critical interaction in the DGrid.AI ecosystem, from inferencepayments to node accountability:

Inference Task Payments

Users pay $DGAI to access LLM inference services via DGrid Adaptor Nodes. Feesare dynamically calculated based on:

• Compute Units (CU): A standardized measure of inference complexity (tied tomodel size, input tokens, and execution time).
• Latency: Adjustments for task speed (rewards for faster execution, minor penalties for delays).

Payments are processed automatically via the Bill Contract, a blockchain-deployed smart contract that deducts $DGAI from users and distributes funds to node operators—ensuring transparency and immutability.

Node Operator Rewards

$DGAI incentivizes community participation in network operations:

• DGrid Nodes: Earn $DGAI based on inference task contributions (e.g., CU processed, low latency, and uptime). Rewards are proportional to performance, with higher payouts for hosting large models (13B+ parameters) or serving high-priority requests.
• DGrid Adaptor Nodes: Receive $DGAI for gateway routing, settlement management, and validation duties (e.g., monitoring node stability, penalizing malicious actors).

Rewards are distributed weekly via the Bill Contract, with details archived to DStorage for auditability.

Staking Collateral

To ensure accountability, $DGAI staking is required for node participation, with tiered requirements based on node type:

• Standard DGrid Nodes
• High-Performance DGrid Nodes
• DGrid Adaptor Nodes

Staked tokens act as collateral against misconduct (e.g., submitting fake inference results, downtime >72 hours). Malicious nodes face slashing (5–20% ofstaked $DGAI) or temporary jailing, enforced automatically via smart contracts.

Network Governance

$DGAI holders govern the evolution of DGrid.AI through on-chain voting, with proposals including:

• Adjusting fee structures (e.g., base CU pricing, latency penalties).
• Whitelisting new LLM models for network support.
• Upgrading core protocols (e.g., GridRPC, GridVM) or settlement rules.
• Allocating treasury funds for ecosystem growth (e.g., developer grants, security audits).

Voting power is proportional to staked $DGAI, ensuring decisions align with the interests of active network participants.