浏览器与电脑操作

voice-browser

@serejaris1JavaScriptMIT更新于 2026-09-17可当场跑

基于Jev、Chrome MV3和本地AI Gateway桥接,实现语音和文本浏览器控制。

英文原文

Voice and text browser control with Jev, Chrome MV3 and a local AI Gateway bridge

serejaris/voice-browser

它在哪儿调用了 Jev

const response=await apiFetch(direct?'https://api.typesafe.ai/v1/systemone':base+'/api/decide',{method:'POST',headers:{'Content-Type':'application/json',...(dir

app.js:28

链接指向我们抓取当天的那个 commit,行号是准的。

它问 Jev 的问题

下面是从这个项目源码里原样取出来的 question 组合。

取自 tests/gateway.test.mjs:5

  1. action选项

    Choose action

    • e1click Docs
    • noneDo nothing

用你自己的内容跑一遍

代码
import { experimental_evaluate as evaluate } from 'ai';

const { answers } = await evaluate({
  model: 'typesafe-ai/jev',
  state,
  questions: {
    action: { type: 'choice', instructions: 'Choose action',
      criteria: { e1: 'click Docs', none: 'Do nothing' } },
  },
});

同类的其他项目