@circuit-llm/inference

OpenAI-compatible client for Circuit's decentralized 72B, served across a mesh of independent GPUs and paid per request in CIRC over x402.

Part of the Circuit SDK. Getting started →

Install

npm install @circuit-llm/inference @circuit-llm/wallet

Usage

import { makeWallet } from '@circuit-llm/wallet';
import { Inference } from '@circuit-llm/inference';

const ai = new Inference({ wallet: makeWallet() });   // wallet from CIRCUIT_WALLET
const res = await ai.chat({ messages: [{ role: 'user', content: 'what is a falling knife?' }] });
console.log(res.content);

• chat(params) — completion, paid automatically per call.
• chatVerified(params, { acceptedKeys }) — returns a signed inference receipt for Verified Intents.
• listModels(), signingKey().

Cap spend per call: new Inference({ wallet, maxSpendRaw: 500_000_000n }).