@ophir/reference-agents

Five pre-configured seller agents for testing and development with the Ophir protocol. Each agent simulates a different service category with realistic pricing, SLA defaults, and custom RFQ handling.

Installation

npm install @ophir/reference-agents

Available agents

| Agent | Category | Default port | Base price | Unit | Description | |---|---|---|---|---|---| | inference | inference | 3001 | $0.005 | request | GPU inference for vision models | | data-processing | data-processing | 3002 | -- | request | Batch data processing pipelines | | code-review | code-review | 3003 | -- | request | Automated code review and analysis | | translation | translation | 3004 | -- | word | Real-time neural machine translation | | image-generation | image-generation | 3005 | -- | image | Image generation from text prompts |

CLI usage

Start a single agent

npx ophir-agents start inference
npx ophir-agents start inference --port 4001

Start all agents

Starts all five agents on consecutive ports.

npx ophir-agents start-all
npx ophir-agents start-all --base-port 4001

List available agents

npx ophir-agents list

Programmatic usage

Each agent type has a factory function that returns a fully configured SellerAgent instance.

import { createInferenceAgent } from '@ophir/reference-agents';

const agent = createInferenceAgent({ port: 3001 });
await agent.listen(3001);

console.log('Agent ID:', agent.getAgentId());
console.log('Endpoint:', agent.getEndpoint());

// The agent is now accepting RFQs and auto-generating quotes
// based on its category-specific pricing and SLA configuration.

Factory functions

| Function | Returns | |---|---| | createInferenceAgent(opts) | SellerAgent configured for GPU inference | | createDataProcessingAgent(opts) | SellerAgent configured for batch processing | | createCodeReviewAgent(opts) | SellerAgent configured for code analysis | | createTranslationAgent(opts) | SellerAgent configured for translation | | createImageGenerationAgent(opts) | SellerAgent configured for image generation |

Each factory function accepts an options object with an optional port field and returns a SellerAgent instance with:

• Category-specific pricing and billing units
• Default SLA metrics appropriate for the service type
• Custom RFQ handlers for service-specific logic
• Volume discounts at 1,000 and 10,000 units

Testing with reference agents

A common development workflow:

import { BuyerAgent } from '@ophir/sdk';
import { createInferenceAgent } from '@ophir/reference-agents';

// Start the seller
const seller = createInferenceAgent({ port: 3001 });
await seller.listen(3001);

// Start the buyer and negotiate
const buyer = new BuyerAgent({ endpoint: 'http://localhost:3002' });
await buyer.listen(3002);

const session = await buyer.requestQuotes({
  sellers: ['http://localhost:3001'],
  service: { category: 'inference', requirements: { model: 'vision' } },
  budget: { max_price_per_unit: '0.01', currency: 'USDC', unit: 'request' },
});

const quotes = await buyer.waitForQuotes(session);
const agreement = await buyer.acceptQuote(quotes[0]);

console.log('Agreement:', agreement.agreement_id);

await buyer.close();
await seller.close();

Documentation

• SellerAgent API reference
• Protocol specification