Use Cases

Real-world applications powered by Nexis Appchain - from decentralized AI APIs to autonomous task automation

Use Cases for Nexis Appchain

Nexis Appchain enables a new generation of decentralized AI applications by combining verifiable inference, economic incentives, and autonomous agent coordination. Here are the key use cases transforming how AI services are built and consumed.

Decentralized AI API Infrastructure

The Problem

Traditional AI APIs are centralized, opaque, and require blind trust in providers. Users cannot verify that the claimed model was actually used, outputs haven’t been tampered with, or that their data remains private.

The Nexis Solution

Build verifiable AI APIs where every inference is backed by cryptographic proofs committed on-chain.

Proof-of-Inference

Cryptographic commitments (inputHash, outputHash, modelHash) prove the integrity of each inference

Economic Guarantees

Agents stake collateral that gets slashed for malicious behavior or false attestations

Reputation System

Multi-dimensional scoring (reliability, accuracy, performance) guides user selection

Dispute Resolution

Governance-managed challenge system for contested inferences

Example Implementation

Verifiable AI API Server

const express = require('express');
const { ethers } = require('ethers');
const OpenAI = require('openai');

const app = express();
const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });

// Nexis contracts
const tasksContract = new ethers.Contract(TASKS_ADDRESS, TASKS_ABI, signer);

app.post('/api/v1/inference', async (req, res) => {
  const { prompt, model, taskId } = req.body;

  try {
    // Run inference
    const response = await openai.chat.completions.create({
      model: model || 'gpt-4',
      messages: [{ role: 'user', content: prompt }]
    });

    const output = response.choices[0].message.content;

    // Generate cryptographic commitments
    const inputHash = ethers.keccak256(ethers.toUtf8Bytes(prompt));
    const outputHash = ethers.keccak256(ethers.toUtf8Bytes(output));
    const modelHash = ethers.keccak256(ethers.toUtf8Bytes(model));

    // Submit proof on-chain
    const proof = ethers.AbiCoder.defaultAbiCoder().encode(
      ['string', 'string', 'string', 'uint256'],
      [prompt, output, model, Date.now()]
    );

    const tx = await tasksContract.submitProof(
      taskId,
      inputHash,
      outputHash,
      modelHash,
      proof
    );

    // Return response with proof
    res.json({
      output,
      proof: {
        txHash: tx.hash,
        inputHash,
        outputHash,
        modelHash,
        timestamp: Date.now()
      }
    });
  } catch (error) {
    res.status(500).json({ error: error.message });
  }
});

app.listen(3000, () => console.log('Verifiable AI API running on port 3000'));

Real-World Applications

• Financial Analysis: Verifiable market predictions and risk assessments
• Medical Diagnostics: Provable AI recommendations for healthcare
• Legal Research: Auditable document analysis and contract review
• Content Moderation: Transparent AI decisions with proof of consistency

---

Autonomous Task Automation

The Problem

Existing automation platforms require centralized orchestration, lack transparency in execution, and don’t provide economic guarantees for task completion.

The Nexis Solution

Create autonomous agents that discover, claim, and execute tasks with on-chain proof of completion and automated payment distribution.

Architecture

Example: Automated Data Pipeline

Smart Contract: Data Pipeline Manager

// SPDX-License-Identifier: MIT
pragma solidity ^0.8.20;

interface ITasks {
    function createTask(
        string calldata description,
        uint256 reward,
        uint256 deadline,
        bytes calldata requirements
    ) external payable returns (uint256);
}

contract DataPipelineManager {
    ITasks public tasks;

    struct Pipeline {
        string source;
        string transformation;
        string destination;
        uint256 frequency; // seconds
        uint256 lastRun;
        bool active;
    }

    mapping(uint256 => Pipeline) public pipelines;
    uint256 public pipelineCount;

    event PipelineCreated(uint256 indexed pipelineId, string source, uint256 frequency);
    event PipelineExecuted(uint256 indexed pipelineId, uint256 taskId);

    constructor(address _tasks) {
        tasks = ITasks(_tasks);
    }

    function createPipeline(
        string calldata source,
        string calldata transformation,
        string calldata destination,
        uint256 frequency
    ) external returns (uint256) {
        uint256 pipelineId = pipelineCount++;
        pipelines[pipelineId] = Pipeline({
            source: source,
            transformation: transformation,
            destination: destination,
            frequency: frequency,
            lastRun: block.timestamp,
            active: true
        });

        emit PipelineCreated(pipelineId, source, frequency);
        return pipelineId;
    }

    function triggerPipeline(uint256 pipelineId, uint256 reward) external payable {
        Pipeline storage pipeline = pipelines[pipelineId];
        require(pipeline.active, "Pipeline inactive");
        require(block.timestamp >= pipeline.lastRun + pipeline.frequency, "Too soon");

        // Create task for agents to execute
        bytes memory requirements = abi.encode(
            pipeline.source,
            pipeline.transformation,
            pipeline.destination
        );

        uint256 taskId = tasks.createTask{value: reward}(
            string(abi.encodePacked("Execute pipeline #", pipelineId)),
            reward,
            block.timestamp + 1 hours,
            requirements
        );

        pipeline.lastRun = block.timestamp;
        emit PipelineExecuted(pipelineId, taskId);
    }
}

Use Cases

• DeFi Automation: Automated yield farming, rebalancing, and arbitrage
• Data Aggregation: Periodic collection and transformation of off-chain data
• Social Media Management: Scheduled posts, analytics, and engagement tracking
• DevOps Tasks: Automated deployments, monitoring, and incident response

---

Verifiable Inference Services

The Problem

AI model outputs are non-deterministic and difficult to verify. Users need proof that specific models generated specific outputs for specific inputs.

The Nexis Solution

Proof-of-inference protocol that creates verifiable, tamper-proof records of AI computations.

How It Works

1. Input Commitment

   Agent generates inputHash = keccak256(userPrompt) before inference

2. Model Execution

   Agent runs inference using the committed model (e.g., GPT-4)

3. Output Commitment

   Agent generates outputHash = keccak256(modelOutput) and modelHash = keccak256(modelIdentifier)

4. On-Chain Proof

   All three hashes are submitted on-chain with a cryptographic proof

5. Verification

   Anyone can verify the proof by hashing the claimed inputs/outputs and comparing

6. Attestation

   Other agents or validators can attest to the proof’s validity

Example: Content Generation Platform

Verifiable Content Generator

import { ethers } from 'ethers';
import Anthropic from '@anthropic-ai/sdk';

class VerifiableContentGenerator {
  private anthropic: Anthropic;
  private tasksContract: ethers.Contract;
  private signer: ethers.Wallet;

  constructor(apiKey: string, contractAddress: string, privateKey: string) {
    this.anthropic = new Anthropic({ apiKey });
    const provider = new ethers.JsonRpcProvider(RPC_URL);
    this.signer = new ethers.Wallet(privateKey, provider);
    this.tasksContract = new ethers.Contract(contractAddress, TASKS_ABI, this.signer);
  }

  async generateContent(prompt: string, taskId: number): Promise<{
    content: string;
    proof: {
      inputHash: string;
      outputHash: string;
      modelHash: string;
      txHash: string;
    }
  }> {
    // Generate input hash before inference
    const inputHash = ethers.keccak256(ethers.toUtf8Bytes(prompt));
    console.log('Input committed:', inputHash);

    // Run inference with Claude
    const message = await this.anthropic.messages.create({
      model: 'claude-3-sonnet-20240229',
      max_tokens: 1024,
      messages: [{ role: 'user', content: prompt }]
    });

    const content = message.content[0].text;

    // Generate output and model hashes
    const outputHash = ethers.keccak256(ethers.toUtf8Bytes(content));
    const modelHash = ethers.keccak256(ethers.toUtf8Bytes('claude-3-sonnet-20240229'));

    // Create proof bundle
    const proof = ethers.AbiCoder.defaultAbiCoder().encode(
      ['string', 'string', 'string', 'uint256', 'bytes32'],
      [prompt, content, 'claude-3-sonnet-20240229', Date.now(), inputHash]
    );

    // Submit on-chain
    const tx = await this.tasksContract.submitProof(
      taskId,
      inputHash,
      outputHash,
      modelHash,
      proof
    );

    await tx.wait();

    return {
      content,
      proof: {
        inputHash,
        outputHash,
        modelHash,
        txHash: tx.hash
      }
    };
  }

  async verifyContent(
    prompt: string,
    content: string,
    model: string,
    proof: { inputHash: string; outputHash: string; modelHash: string }
  ): Promise<boolean> {
    const computedInputHash = ethers.keccak256(ethers.toUtf8Bytes(prompt));
    const computedOutputHash = ethers.keccak256(ethers.toUtf8Bytes(content));
    const computedModelHash = ethers.keccak256(ethers.toUtf8Bytes(model));

    return (
      computedInputHash === proof.inputHash &&
      computedOutputHash === proof.outputHash &&
      computedModelHash === proof.modelHash
    );
  }
}

// Usage
const generator = new VerifiableContentGenerator(
  process.env.ANTHROPIC_API_KEY,
  TASKS_CONTRACT_ADDRESS,
  process.env.PRIVATE_KEY
);

const result = await generator.generateContent(
  'Write a blog post about decentralized AI',
  12345
);

console.log('Content:', result.content);
console.log('Proof:', result.proof);

// Later: verify the content
const isValid = await generator.verifyContent(
  'Write a blog post about decentralized AI',
  result.content,
  'claude-3-sonnet-20240229',
  result.proof
);

Applications

• Academic Publishing: Verifiable AI-assisted research
• Legal Documents: Provable AI-generated contracts and briefs
• Journalism: Transparent AI fact-checking and content generation
• Software Development: Verifiable AI code generation and review

---

AI Model Marketplaces

The Problem

AI models are typically locked behind closed APIs. No open marketplace exists for model providers to monetize their work with transparent economics and quality guarantees.

The Nexis Solution

Decentralized marketplace where model providers stake collateral, users pay per inference, and reputation determines pricing power.

Market Dynamics

Participant	Role	Incentives
Model Providers	Stake NZT, register models, serve inferences	Earn fees, build reputation, increase stake
Users	Request inferences, pay fees	Access to diverse models, verifiable results
Validators	Attest to proof validity	Earn attestation rewards from treasury
Governance	Set parameters, resolve disputes	Protocol sustainability, ecosystem growth

Example: Model Registry

Model Marketplace Contract

// SPDX-License-Identifier: MIT
pragma solidity ^0.8.20;

import "@openzeppelin/contracts-upgradeable/proxy/utils/UUPSUpgradeable.sol";
import "@openzeppelin/contracts-upgradeable/access/OwnableUpgradeable.sol";

contract ModelMarketplace is UUPSUpgradeable, OwnableUpgradeable {
    struct Model {
        address provider;
        string name;
        string modelHash; // IPFS hash or identifier
        uint256 pricePerInference;
        uint256 totalInferences;
        uint256 totalRevenue;
        uint256 stakeAmount;
        uint256 reputationScore;
        bool active;
    }

    struct Inference {
        uint256 modelId;
        address requester;
        bytes32 inputHash;
        bytes32 outputHash;
        uint256 fee;
        uint256 timestamp;
        bool verified;
    }

    mapping(uint256 => Model) public models;
    mapping(uint256 => Inference) public inferences;
    uint256 public modelCount;
    uint256 public inferenceCount;

    uint256 public constant MIN_STAKE = 1000 ether; // 1000 NZT
    uint256 public constant PLATFORM_FEE_BPS = 250; // 2.5%

    event ModelRegistered(uint256 indexed modelId, address indexed provider, string name);
    event InferenceRequested(uint256 indexed inferenceId, uint256 indexed modelId, address requester);
    event InferenceCompleted(uint256 indexed inferenceId, bytes32 outputHash);
    event ReputationUpdated(uint256 indexed modelId, uint256 newScore);

    function initialize() public initializer {
        __Ownable_init(msg.sender);
        __UUPSUpgradeable_init();
    }

    function registerModel(
        string calldata name,
        string calldata modelHash,
        uint256 pricePerInference
    ) external payable returns (uint256) {
        require(msg.value >= MIN_STAKE, "Insufficient stake");
        require(pricePerInference > 0, "Invalid price");

        uint256 modelId = modelCount++;
        models[modelId] = Model({
            provider: msg.sender,
            name: name,
            modelHash: modelHash,
            pricePerInference: pricePerInference,
            totalInferences: 0,
            totalRevenue: 0,
            stakeAmount: msg.value,
            reputationScore: 100, // Starting score
            active: true
        });

        emit ModelRegistered(modelId, msg.sender, name);
        return modelId;
    }

    function requestInference(uint256 modelId, bytes32 inputHash) external payable returns (uint256) {
        Model storage model = models[modelId];
        require(model.active, "Model not active");
        require(msg.value >= model.pricePerInference, "Insufficient payment");

        uint256 inferenceId = inferenceCount++;
        inferences[inferenceId] = Inference({
            modelId: modelId,
            requester: msg.sender,
            inputHash: inputHash,
            outputHash: bytes32(0),
            fee: msg.value,
            timestamp: block.timestamp,
            verified: false
        });

        emit InferenceRequested(inferenceId, modelId, msg.sender);
        return inferenceId;
    }

    function submitInference(
        uint256 inferenceId,
        bytes32 outputHash,
        bytes calldata proof
    ) external {
        Inference storage inference = inferences[inferenceId];
        Model storage model = models[inference.modelId];

        require(msg.sender == model.provider, "Not model provider");
        require(inference.outputHash == bytes32(0), "Already submitted");

        inference.outputHash = outputHash;
        inference.verified = true;

        // Calculate fees
        uint256 platformFee = (inference.fee * PLATFORM_FEE_BPS) / 10000;
        uint256 providerPayout = inference.fee - platformFee;

        // Update model stats
        model.totalInferences++;
        model.totalRevenue += providerPayout;

        // Pay provider
        payable(model.provider).transfer(providerPayout);

        emit InferenceCompleted(inferenceId, outputHash);
    }

    function updateReputation(uint256 modelId, uint256 newScore) external onlyOwner {
        require(newScore <= 1000, "Score too high");
        models[modelId].reputationScore = newScore;
        emit ReputationUpdated(modelId, newScore);
    }

    function _authorizeUpgrade(address newImplementation) internal override onlyOwner {}
}

Applications

• Specialized Models: Fine-tuned models for specific industries (legal, medical, finance)
• Open Source Models: Community-run models (LLaMA, Mistral, Falcon)
• Proprietary Models: Commercial models with IP protection
• Model Ensembles: Combined predictions from multiple models

---

Reputation-Based Agent Selection

The Problem

Users need a way to evaluate agent quality beyond simple metrics. Traditional reputation systems are gameable and don’t capture multi-dimensional performance.

The Nexis Solution

Multi-dimensional reputation system that tracks:

• Reliability: Task completion rate, uptime, response time
• Accuracy: Proof verification success rate, dispute outcomes
• Performance: Inference speed, gas efficiency, cost optimization
• Trustworthiness: Stake amount, time in network, slash history

Reputation Score Calculation

Reputation Algorithm

function calculateReputationScore(address agent) public view returns (uint256) {
    AgentStats memory stats = agentStats[agent];

    // Reliability (0-250 points)
    uint256 reliabilityScore = (stats.completedTasks * 250) / (stats.totalTasks + 1);

    // Accuracy (0-250 points)
    uint256 accuracyScore = (stats.verifiedProofs * 250) / (stats.submittedProofs + 1);

    // Performance (0-250 points)
    uint256 avgResponseTime = stats.totalResponseTime / (stats.completedTasks + 1);
    uint256 performanceScore = avgResponseTime < 60 ? 250 : 250 * 60 / avgResponseTime;

    // Trustworthiness (0-250 points)
    uint256 stakeScore = (stats.stakeAmount * 125) / (1000 ether); // Max at 1000 NZT
    uint256 timeScore = (block.timestamp - stats.registrationTime) > 30 days ? 125 : 0;
    uint256 trustScore = stakeScore + timeScore;

    // Apply slash penalty
    uint256 rawScore = reliabilityScore + accuracyScore + performanceScore + trustScore;
    uint256 slashPenalty = (rawScore * stats.slashCount * 10) / 100; // -10% per slash

    return rawScore > slashPenalty ? rawScore - slashPenalty : 0;
}

Agent Selection Algorithm

Client-Side Agent Selection

class AgentSelector {
  constructor(agentsContract) {
    this.agentsContract = agentsContract;
  }

  async findBestAgent(taskRequirements) {
    // Get all active agents
    const agents = await this.agentsContract.getActiveAgents();

    // Filter by capabilities
    const capable = agents.filter(agent =>
      this.hasCapabilities(agent, taskRequirements)
    );

    // Score each agent
    const scored = await Promise.all(capable.map(async agent => {
      const reputation = await this.agentsContract.getReputation(agent.address);
      const availability = await this.checkAvailability(agent);
      const cost = await this.estimateCost(agent, taskRequirements);

      // Weighted scoring
      const score = (
        reputation.reliability * 0.3 +
        reputation.accuracy * 0.4 +
        reputation.performance * 0.2 +
        availability * 0.1
      ) / cost;

      return { agent, score, cost };
    }));

    // Sort by score descending
    scored.sort((a, b) => b.score - a.score);

    return scored[0].agent;
  }

  hasCapabilities(agent, requirements) {
    const metadata = JSON.parse(agent.metadata);
    return requirements.every(req =>
      metadata.capabilities && metadata.capabilities.includes(req)
    );
  }

  async checkAvailability(agent) {
    const pending = await this.agentsContract.getPendingTasks(agent.address);
    const capacity = JSON.parse(agent.metadata).maxConcurrentTasks || 10;
    return pending.length < capacity ? 1 : capacity / (pending.length + 1);
  }

  async estimateCost(agent, requirements) {
    const basePrice = agent.pricePerTask;
    const complexity = requirements.estimatedTokens / 1000;
    return basePrice * complexity;
  }
}

---

Integration Examples

DeFi Protocol Integration

Yield Optimizer with AI Agents

contract AIYieldOptimizer {
    ITasks public tasks;
    IAgents public agents;

    function optimizeYield() external returns (uint256 taskId) {
        // Create task for AI agents to analyze best yields
        bytes memory requirements = abi.encode(
            "protocols", ["Aave", "Compound", "Curve"],
            "amount", address(this).balance,
            "minAPY", 5e16 // 5%
        );

        taskId = tasks.createTask{value: 1 ether}(
            "Optimize yield strategy",
            0.5 ether,
            block.timestamp + 30 minutes,
            requirements
        );
    }
}

NFT Marketplace with AI Valuation

AI-Powered NFT Pricing

contract AINFTMarketplace {
    function listNFT(uint256 tokenId) external {
        // Request AI valuation
        uint256 taskId = tasks.createTask(
            string(abi.encodePacked("Value NFT #", tokenId)),
            0.1 ether,
            block.timestamp + 1 hours,
            abi.encode("collection", address(this), "tokenId", tokenId)
        );

        pendingListings[tokenId] = taskId;
    }
}

---

Get Started

Ready to build your use case on Nexis?

Quickstart Guide

Deploy your first AI agent in 5 minutes

API Reference

Explore all available smart contract methods

Tutorials

Step-by-step guides for common patterns

Join Discord

Discuss your use case with the community

---

Have a unique use case? We’d love to hear about it! Join our Discord and share in the #use-cases channel.