>ParalleliX AI (Consumer Demand)

ParalleliX AI is a consumer chat app (ChatGPT / Claude style) that runs open-source models on the registered operator nodes. Using it costs $PRLX, metered as credits (Part VIII.7). Each prompt becomes one inference request that the coordinator dispatches whole to a single capable node; the user never thinks about routing, only about sending a message and getting a response. The shape of the flow:

1. Sign up. Create a ParalleliX AI account.
2. Fund credits. Deposit $PRLX once on-chain to a credit balance. The deposit is credited to an off-chain ledger; there is no per-message gas and no per-message signature.
3. Send a prompt. The request is metered off-chain (message_cost = model_rate × tokens_processed) and debited from the credit balance.
4. Dispatch. The coordinator routes the whole inference request to one online node (or, if none is available, to project-operated fallback hardware).
5. Receive. The model response streams back to the user. The $PRLX spent flows to the credit ledger, and an 85 % share settles to the operators whose nodes served the work.

This closes the loop: ParalleliX AI usage is the demand, each request is work for an operator node, and serving requests keeps a node's uptime high and its rewards flowing (Part VIII.4).

ParalleliX AI is custodial v1 and runs open-source models: honestly lower quality than ChatGPT or Claude, stated plainly, not trustless and not on-chain-settled.