Inference Request (User Job)

Any user, DAO, or application can initiate a verifiable AI task by submitting an inference request. This request is cryptographically tied to:

• A registered model,

• The user’s input (hashed),

• And a HUB token payment.

Once submitted, the request enters the execution pipeline and becomes a provable, revenue-generating event within the system.

---

🔁 Workflow:

1. User selects a model from the on-chain registry.

2. Prepares input data (off-chain JSON, hashed on client side).

3. Sends InferenceRequest transaction with:

   • model_hash

   • input_hash

   • payment_amount (in HUB or stablecoin)

4. Job is queued for executors.

---

📦 On-Chain Struct:

struct InferenceRequest {
  bytes32 input_hash;     // SHA-256 of user’s input
  bytes32 model_hash;     // Tied to registered zk-model
  address user;           // Sender address
  uint256 payment_amount; // In HUB
  uint64 timestamp;       // Unix time
}

---

📥 Input Data Example (Off-Chain):

{
  "APR": 6.42,
  "stake_volume": 125000,
  "risk_profile": "moderate"
}

This input is not stored on-chain, but its hash is used to validate the zk-proof.

---

🌍 RWA Integration:

• Payment Tracking: Requests using RWA-tokenized models will trigger payout logic to token holders.

• Subscription-Based Access: DAOs holding access tokens (NFTs) can submit requests without direct HUB payment (prepaid logic).

• Auditable Revenue: Each request adds to the on-chain usage history of a model, powering dashboards, reputation, and financial reporting.

Every inference request is a cryptographic contract — linking a user’s intent, a model’s logic, and an on-chain monetization outcome.