Skip to content

Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.

1.7 KB
Raw Blame
// Service-catalog descriptor — the single written-once listing for this
// service. api/wk.js derives its /.well-known/x402.json resource entry from
// this file via api/_lib/service-catalog/index.js (toBazaarDiscovery), and
// the OKX storefront projection reads the same record (toOkxCatalog).
// Contract: specs/service-catalog.md. Do not re-add a hand-written mirror
// for this route in api/wk.js — edit this descriptor instead.

export default {
	slug: 'llm-proxy',
	title: 'LLM Inference Proxy',
	category: 'agent-infra',
	useCase: 'three.ws LLM Inference Proxy — pay per completion with no API key required.',
	path: '/api/x402/llm-proxy',
	method: 'POST',
	free: false,
	status: 'live',
	priceAtomics: '5000',
	acceptsBuilder: 'standard',
	serviceName: 'three.ws LLM Inference Proxy',
	tags: ['llm', 'inference', 'completion', 'proxy', 'benchmark'],
	description: 'three.ws LLM Inference Proxy — pay per completion with no API key required. Runs one-shot text prompts through the platform\'s free-first provider chain. Response includes measured latency, token counts, and the provider actually used. Ideal for latency benchmarking, agent pipelines, and one-off completions. Model aliases: "fast" (sub-second) · "smart" (quality backstop). Price: $0.005 USDC per completion on Base or Solana.',
	input: {
		model: 'fast',
		prompt: 'Count to 3.',
		max_tokens: 10,
	},
	inputSchema: {
		type: 'object',
		required: ['prompt'],
		properties: {
			model: {
				type: 'string',
				default: 'fast',
			},
			prompt: {
				type: 'string',
				minLength: 1,
				maxLength: 4000,
			},
			max_tokens: {
				type: 'integer',
				minimum: 1,
				maximum: 2048,
				default: 256,
			},
		},
	},
	storefronts: ['x402scan'],
};