@octoryn/react

React hooks for governed Octoryn chat.

import { Octoryn } from "@octoryn/sdk";
import { useOctorynChat } from "@octoryn/react";

const client = new Octoryn({ apiKey: token });

function Chat() {
  const { messages, status, sendMessage, resume, regenerate, editMessage, stop } = useOctorynChat({
    client,
    model: "openai/gpt-4.1-mini",
  });

  // Render message.parts rather than assuming every message is plain text.
  // Parts can include governance, tools, sources, files and errors.
}

The hook deliberately does not execute an autonomous agent loop or claim an agent API on the Router.

For single-prompt streaming, use useCompletion:

const { completion, status, complete, stop } = useCompletion({
  client,
  model: "openai/gpt-4.1-mini",
});

useObject streams display-safe partial values and strictly parses the final object at the hook boundary. Pass a Zod schema directly, or a { request, parse } adapter for raw JSON Schema. Custom UITransport and ordered stream transforms are accepted by every hook.