Skip to content
MicroResolve

Node.js — Connect to Server

The embedded MicroResolve instance runs entirely in-process. For teams that want a shared training pipeline, you can call the self-hosted server directly over HTTP.

Call the server over HTTP

const BASE = 'http://localhost:3001';
const HEADERS = {
  'X-Namespace-ID': 'support',
  'Content-Type': 'application/json',
};


async function classify(query: string) {
  const res = await fetch(`${BASE}/api/route_multi`, {
    method: 'POST',
    headers: HEADERS,
    body: JSON.stringify({ query }),
  });
  const data = await res.json();
  return data.intents as Array<{ id: string; score: number }>;
}


const matches = await classify('cancel my order');
console.log(matches);
// [{ id: 'cancel_order', score: 0.91 }]

Add phrases via HTTP

async function addPhrase(intentId: string, phrase: string, lang = 'en') {
  const res = await fetch(`${BASE}/api/intents/${intentId}/phrases`, {
    method: 'POST',
    headers: HEADERS,
    body: JSON.stringify({ phrase, lang }),
  });
  return res.json();
}

Correct a misclassification via HTTP

async function correct(query: string, wrong: string | null, correct: string) {
  await fetch(`${BASE}/api/correct`, {
    method: 'POST',
    headers: HEADERS,
    body: JSON.stringify({ query, wrong, correct }),
  });
}

Hybrid: local engine + server sync

For the lowest possible latency, run a local engine in-process and use the server as the training hub. Built-in connected mode is available in the Rust library — see Rust — Connect to Server.

From Node.js the recommended pattern is to route all calls through the server’s HTTP API. The server classifies in-process; HTTP latency on localhost is typically 2–3ms.

const BASE = 'http://localhost:3001';


async function classify(query: string, ns = 'support') {
  const res = await fetch(`${BASE}/api/route_multi`, {
    method: 'POST',
    headers: { 'X-Namespace-ID': ns, 'Content-Type': 'application/json' },
    body: JSON.stringify({ query }),
  });
  const data = await res.json();
  return data.intents as Array<{ id: string; score: number }>;
}


async function reportCorrection(
  query: string,
  wrong: string | null,
  correct: string,
  ns = 'support',
) {
  await fetch(`${BASE}/api/correct`, {
    method: 'POST',
    headers: { 'X-Namespace-ID': ns, 'Content-Type': 'application/json' },
    body: JSON.stringify({ query, wrong, correct }),
  });
}

Next