@hiway2llm/client

Official TypeScript / JavaScript SDK for HiWay2LLM — BYOK smart LLM routing with 40-60 % savings and zero markup on inference.

Install

npm install @hiway2llm/client

Runs unmodified in Node 18+, Deno, Bun, Vercel Edge runtime and Cloudflare Workers. Uses native fetch, zero extra dependencies.

Quickstart

import { Hiway } from "@hiway2llm/client";

const h = new Hiway({ apiKey: process.env.HIWAY_API_KEY });

const res = await h.chat.completions.create({
  model: "auto",                                    // smart routing
  messages: [{ role: "user", content: "Say hi" }],
});

console.log(res.choices[0].message.content);
console.log(res._hiway?.routed_model);              // which model was picked

Already using the OpenAI SDK? You can keep it — just point baseURL at HiWay:

import OpenAI from "openai";
const client = new OpenAI({
  apiKey: process.env.HIWAY_API_KEY,                // hw_live_...
  baseURL: "https://app.hiway2llm.com/v1",
});

Both approaches hit the same OpenAI-compatible endpoint.

API surface

| Method | Description | |---|---| | h.chat.completions.create({ model, messages, ... }) | OpenAI-compatible chat completion. | | h.models.list() | Models your workspace can route to (BYOK-filtered). | | h.me() | Current user + plan. |

Errors

All 4xx / 5xx responses throw a HiwayError with .status + .body:

import { Hiway, HiwayError } from "@hiway2llm/client";
try {
  await h.chat.completions.create({ /* … */ });
} catch (e) {
  if (e instanceof HiwayError && e.status === 402) {
    // quota exhausted — prompt to upgrade
  }
}

Links

• Homepage
• Docs
• CLI
• GitHub

License

MIT — © Mytm-Group SAS