The Oryn Browser SDK

You build the app.
Oryn runs the model.

Add local AI to your website with one SDK. Oryn handles supported engines, downloads and device resources. Keep your existing web experience for other browsers.

Developer alpha · Apple Silicon Macs.
Public browser and SDK downloads are not available yet.

Your machine. Your decision.
f.FIELDNOTESYOUR RESEARCH SPACE
A LITTLE ROOM TO THINK

Good ideas
start with a note.

ORYN LOCAL AIBrowser-controlled permission02 / 04

Let this site use local AI?

fieldnotes.example wants to summarize your notes using a model on your machine.

Illustration only. No permission is granted.

Interactive product illustration · no model runs on this page.

Chromium at its coreInference on your devicePermission in your handsModels, with clear boundaries

01 / How it works

Your app asks.
Oryn runs the AI.

Use the Oryn Browser SDK to request AI work from your website. Oryn handles permission, supported models and local execution. Your app receives the results.

User’s computer

Website app

Uses the Oryn Browser SDK

Oryn Browser

Permission & orchestration

Enginesmodels

User’s hardware

CPU · GPU · Memory

Results back to your app. Oryn runs the model on the user’s hardware and returns its output through the SDK.

From request to result

A local chat request.

Describe the AI task your app needs. Oryn handles model choice and any needed download, with the user’s permission.

Send the user’s message, receive the result and display it in your interface. Release the session when the work is done.

Open the complete SDK example

Before this excerpt, create the SDK client and check availability. Run it from a user action; your app supplies the message, cancellation signal and result display, and handles refusal or errors.

TypeScript · request and result
const session = await localAI.models.request({
  role: "chat",
  requirements: {
    task: "text-generation",
    outputModalities: ["text"],
    capabilities: { chat: true },
  },
}, { signal });

try {
  const result = await invokeAndWait(localAI, {
    sessionId: session.sessionId,
    input: {
      messages: [{ role: "user", content: message }],
    },
  }, { signal });
  renderResult(result);
} finally {
  await localAI.models.releaseSession(session.sessionId);
}

Import createLocalAIClient and invokeAndWait from @ocno/oryn-sdk. The SDK package is not published yet.

02 / Working examples

See what
you can build.

Try websites built with the Oryn Browser SDK, then explore how to add local AI to your own app.

03 / Who handles what

You own the experience.
Oryn handles local execution.

Oryn Browser

Permission, supported models and engines, downloads, device resource checks and visible job controls.

Your website

Interface, prompts, output quality, error handling and an alternative when local AI is unavailable.

Check the model and device.

Support depends on the exact model files, task and device. Test the quality, speed and memory use your app needs.

What should I test before shipping?

Use representative inputs on your target Macs. Test denied permission, cancellation, errors and the alternative your app offers outside Oryn.

What can my website access?

Its own authorized results, job events and artifacts. No direct engine access, local filesystem paths or global model inventory. Trust cannot override unsupported or unsafe execution.

Read model and permission details

Developer alpha · Apple Silicon

Bring one useful feature.
Start your evaluation.

Review the SDK example and choose a task to test. The browser and SDK are in developer alpha; the package is not on the registry.