Firefox AI Platform

This component is an experimental machine learning local inference engine based on Transformers.js and the ONNX runtime. You can use the component to leverage the inference runtime in the context of the browser. To try out some inference tasks, you can refer to the 1000+ models that are available in the Hugging Face Hub that are compatible with this runtime.

To enable the engine, flip the browser.ml.enable preference to true in about:config then visit about:inference (Nightly only) or add the following snippet of code into your (privileged) Javascript code in Firefox or in the browser console:

const { createEngine } = ChromeUtils.importESModule("chrome://global/content/ml/EngineProcess.sys.mjs");
const engine = await createEngine({taskName: "summarization"});
const request = { args:  ["This is the text to summarize"]};
const res = await engine.run(request);
console.log(res[0]["summary_text"]);

Learn more about the platform:

• Architecture
• APIs
• Notifications
• Models management
• How to perftest a model
• WebExtensions AI API
• Trial Inference API Extension Example