AI Agent Template for eternaX Network

Quick Start

This template provides everything you need to create AI-powered autonomous agents on eternaX Network.

1-Minute Setup

# Clone and install
npm install

# Deploy example agents
npm run deploy

# Your agents are now running autonomously!

Template Structure

contracts/
├── BaseAgent.sol           # Core template - inherit from this
├── Payer.sol               # Scheduling and payment example
├── SimpleBettingAgent.sol  # Advanced example (betting logic)
└── interfaces/             # Required interfaces

scripts/
└── deploy-example-agent.js # Deployment script

docs/
├── AGENT_DEVELOPMENT_GUIDE.md  # Complete guide
├── ETERNAX_PRECOMPILES.md      # Precompiles documentation
└── SCHEMA_CALLBACK.md          # Schema-based callback documentation

How to Create Your Agent

Step 1: Inherit from BaseAgent

import "./BaseAgent.sol";

contract MyAgent is BaseAgent {
    constructor(uint256 interval, uint32 maxExecutions) 
        BaseAgent(interval, maxExecutions) 
    {
        // Your initialization
    }
}

Step 2: Implement 4 Required Functions

// 1. Your main business logic
function executeAgentLogic() external override onlySelf whenActive returns (bool) {
    // Called by scheduler - implement your logic here
}

// 2. Configure AI models & schemas per request
function getInferenceConfig(string memory context) internal override returns (InferenceConfig memory) {
    // Return different configs based on context
}

// 3. Handle AI responses
function processAIResponse(bytes32 requestId, string memory response) internal override {
    // Parse JSON response and execute actions
}

// 4. Validate your configuration
function validateConfiguration() internal override view returns (bool) {
    // Return true if agent is properly configured
}

Step 3: Deploy & Activate

const agent = await MyAgent.deploy(10, 100); // 10 blocks interval, 100 max executions
await agent.setActive(true);
await agent.setupPeriodicExecution();

Schema Examples

Market Prediction

{
  "type": "object",
  "properties": {
    "prediction": {"type": "string", "enum": ["buy", "sell", "hold"]},
    "confidence": {"type": "integer", "minimum": 0, "maximum": 100},
    "reasoning": {"type": "string"}
  },
  "required": ["prediction", "confidence", "reasoning"]
}

Risk Assessment

{
  "type": "object",
  "properties": {
    "risk_level": {"type": "integer", "minimum": 1, "maximum": 10},
    "recommendation": {"type": "string", "enum": ["reduce", "maintain", "increase"]},
    "factors": {"type": "array", "items": {"type": "string"}}
  },
  "required": ["risk_level", "recommendation"]
}

Flexible Schema Configuration

Dynamic Schema Evolution

function adaptSchema(uint256 successRate) internal {
    if (successRate < 60) {
        currentSchema = CONSERVATIVE_SCHEMA; // More restrictive
    } else if (successRate > 80) {
        currentSchema = AGGRESSIVE_SCHEMA;   // More flexible
    }
}

Best Practices

DO

• Always validate AI responses before acting
• Use try-catch for all external calls
• Implement emergency pause functionality
• Monitor gas usage and optimize
• Test schemas with your target AI models

DON'T

• Put schemas in constructor parameters
• Act on unvalidated AI responses
• Ignore error handling
• Deploy without testing
• Use overly complex schemas

Monitoring Your Agent

Events to Watch

event InferenceRequested(bytes32 indexed requestId, string context);
event InferenceResponseReceived(bytes32 indexed requestId, string response);
event AgentExecutionCompleted(uint256 timestamp, bool success);
event AgentExecutionFailed(string reason);

State Functions

// Get overall state
(uint256 sent, uint256 received, uint256 lastRequest, bool requesting, bool active) = agent.getState();

// Get specific request info
RequestInfo memory info = agent.getRequestInfo(requestId);

Troubleshooting

Problem	Solution
Agent not executing	Check if active and scheduled
Callback not working	Verify function selector and gas limit
Schema validation fails	Check JSON syntax and required fields
High gas costs	Optimize callback gas limits and schema complexity

Examples Included

1. Payer

• Purpose: Shows how to pay a payee a fixed amount of tokens every X blocks for Y times. Customizable.
• Features: Payee can be any address. Token can be native ETX or any ERC20 token.
• Good for: Understanding how to use deferred execution in business logic.

2. SimpleBettingAgent

• Purpose: Shows advanced betting agent for prediction markets
• Features: Risk management, bet sizing, native ETX handling, multiple models
• Schemas: Market analysis, risk assessment, portfolio review
• Good for: Understanding complex agent logic with real market integration

Ready to Build?

1. Read: AGENT_DEVELOPMENT_GUIDE.md for complete details
2. Study: SimpleBettingAgent.sol
3. Inherit: From BaseAgent.sol
4. Implement: The 4 required functions
5. Deploy: Using the provided script
6. Monitor: Using events and state functions

Testnet Configuration

• EternaX testnet1 is configured in hardhat.config.js with a few pre-funded accounts. Please use with moderation.
• The RPC URL is https://rpc.eternax.ai
• The model available for inference is llama3.1:8b-instruct-q8_0
• The Market Oracle is deployed at 0x4D589B49d8763C9FaaE726d464E3B9F53347BED4
• The Prediction Market is deployed at 0x623e3E052f7cC5df91afcd31ea6195396d41f912
• The WETX token is deployed at 0x7d2C41C76dc3EA1E9A0F10B1d0EB3aCB78E22B64

Support

• Full guide: AGENT_DEVELOPMENT_GUIDE.md
• Examples: See ExampleAgent.sol and SimpleBettingAgent.sol
• Deployment: Use scripts/deploy-example-agent.js

For quick questions, DM us on Twitter: https://x.com/eternaxLabs

Happy building! 🎉