Architecture

Layer 1 EVM six-layer stack. Protocol layer integrates Multi-Token Gas and On-Chain AI; off-chain AI Workers and NuaSwap handle execution and swaps.

L1 EVM PoS Multi-Gas On-Chain AI
01 · Stack

Six-Layer Stack

Bottom-up: data persistence → network propagation → consensus → contract execution → protocol extensions → application access.

L6

Application

DApp · Pantanal AI API · SDK · Web · Agent

User and developer access; DApp and AI service invocation layer.

L5

Protocol

Multi-Token Gas · On-Chain AI · Governance

On-chain protocol extensions: core logic for gas payment and AI service economics.

L4

Execution

EVM · WASM · Precompiled Contracts

Smart contract and precompiled contract execution environment.

L3

Consensus

PoS · Block Production · Finality

Validator election, block production, and finality guarantees.

L2

Network

P2P · Block Sync · State Broadcast

Inter-node block and state synchronization.

L1

Data

State Trie · Event Log · IPFS Interface

On-chain state, event logs, and off-chain data interface.

02 · Separation

Chain vs. Execution Separation

Blockchain maintains identity, registration, stake, verification, and settlement; AI Workers execute inference, training, and Agent tasks; NuaSwap provides gas token swap liquidity.

Blockchain Layer
IdentityPaymentRegistryStake VerificationReputationDisputeSettlementMulti-Gas
OFF-CHAIN
Off-Chain Execution
AI GatewaySchedulerAI Worker Model RuntimeNuaSwap Swap
03 · Protocol

Protocol Layer Modules

Protocol layer composed of Multi-Token Gas and On-Chain AI extensions; shared PRC20 payment abstraction.

Multi-Token Gas

Extends EVM gas model to a PRC20 token set.

View Multi-Token Gas Protocol
01

Gas Token Registry

Gas token admission, state management, and risk rating

02

NuaSwap Integration

Liquidity pool quote source and depth constraints

03

Swap Engine

On-chain PRC20 → PANTA swap execution

04

Gas Settlement

Miner fee deduction and gas revenue distribution

On-Chain AI

AI service registration, pricing, verification, and on-chain settlement.

View On-Chain AI Protocol
01

Service Payment

PRC20 pricing, locking, and on-chain settlement

02

Node Registry

Node identity, hardware claims, pricing, and stake

03

Model Registry

Model ID · Hash · Version · Runtime · Capabilities

04

Task Registry

Task ID, input commitment, price cap, and status

05

Payment Channel

max_spend lock and Usage Receipt streaming settlement

06

Verification / Dispute

Task verification, dispute resolution, and stake slashing

04 · Payment

PRC20 Payment Abstraction

Gas Fee and AI Service Fee metered independently; shared PRC20 interface with separate settlement paths.

Gas Fee Path

PRC20 → Swap Engine → PANTA → Miner Fee Settlement

AI Service Fee Path

PRC20 → Payment Channel → Usage Receipt → Settlement
05 · Gas Path

Gas Payment Path

Registry
Validate gas token admission and NuaSwap pool binding
Quote
Read NuaSwap pool quote
Swap
Swap Engine executes PRC20 → PANTA
Execute
Deduct miner fee and broadcast transaction
Settle
Distribute gas revenue to validators
06 · AI Path

AI Service Path

Gateway
Pantanal AI API unified entry point
Scheduler
Route by model, price, latency, and reputation
Worker
AI Worker executes inference / training / Agent
Receipt
Streaming signed Usage Receipt
Verify
Proof of Inference and dispute handling
Settle
PRC20 on-chain final settlement
07 · Boundary

On-Chain / Off-Chain Boundary

Chain maintains commitments, state, and settlement; off-chain carries compute load and data flow.

ON-CHAIN
  • Task ID / Gas Token State
  • User · Worker · Model ID
  • Price · Max Spend · Usage Commitment
  • Result Hash · Proof · Settlement
OFF-CHAIN
  • Prompt · Response · Streaming
  • Model Execution · Agent Runtime
  • NuaSwap Pool Swap
  • Training Data · Intermediate Tensors
08 · Data Flow

System Data Flow

Users → API / SDK / DApp
        ├─ Gas Path: PRC20 → Registry → Swap Engine → PANTA → Miner Fee
        └─ AI Path: Gateway → Scheduler → Worker → Usage Receipt → Verification → PRC20 Settlement
09 · Client

Client Modules (AI-Geth)

Hosts EVM, consensus interface, Multi-Token Gas module, and On-Chain AI module. Does not embed AI model runtime.

ai-geth 
AI-Geth
├── EVM / WASM Runtime
├── Consensus Interface
├── Multi-Token Gas Module
│   ├── Gas Token Registry
│   ├── NuaSwap Integration
│   ├── Swap Engine
│   └── Gas Settlement
├── On-Chain AI Module
│   ├── Service Payment (PRC20)
│   ├── Node / Model / Task Registry
│   ├── Payment Channel
│   ├── Stake / Reputation
│   └── Verification / Settlement
└── Governance Interface

AI inference runs on off-chain AI Workers; gas swaps use NuaSwap liquidity.

10 · Performance

Performance Parameters

≥ 50,000
Peak TPS
≤ 1s
Block Time
Registry-gated
Gas Token Support
EVM + WASM
Virtual Machine
Payment Channel
AI Settlement Mode
99.99%
Availability
11 · Security

Security Model

Consensus

Validator Stake / Slashing · Block Finality

Multi-Gas

Signature · Nonce · Gas Limit · Pool Depth · Registry Risk Rating

On-Chain AI

Proof of Inference · Dispute Resolution · Stake Slashing

Settlement

Atomic Finalization · Payment Channel Price Lock · Double-Entry Verification

Governance

Protocol Upgrade Gate · Emergency Delisting · Treasury Controls

12 · Docs

Protocol Docs

Multi-Token Gas Protocol→ On-Chain AI Protocol→