shipwithjev

Catalog / Agents & browsers

0061GitHub

jev-use (voice)

Voice control for macOS where Jev is the fallback, not the router.

A Rust core in a Tauri shell that takes spoken commands on macOS. Deterministic matching runs first, and Jev only decides the cases the rules do not cover.

coco-research/jev-useREADME ↗
# jev-use

**Speak a goal. The Mac does it.**

A voice-driven control layer for macOS: it observes the screen through the accessibility
tree and drives it, with everything running locally.

`product source · pre-alpha · no release yet`





</div>

> **Two things live here:** the macOS automation runtime, and the canonical copy of the
> **Repo Standard** every repository in this fleet carries. Both are described below.
>
> **No release yet.** There is no installer, no notarised build and no Homebrew tap, so today
> you clone it and build it. That is the honest state, not a placeholder.

<img src="docs/readme/local-first.png" alt="Repo Standard and the macOS tools" width="100%"/>

## What works today, and what does not

| Piece | State | Evidence |
| --- | --- | --- |
| **Observation** - walk an app's accessibility tree into an indexed element table | **done** | at parity with the reference on four apps: Finder 12/12 elements, Terminal 13/13, PI-Desktop 10/10 (Electron), Notes 8/8, with identical fingerprints. Reproduce: `scripts/parity --live <app>` |
| **Parity harness** - prove the port matches the reference | **done** | `scripts/parity --capture/--check/--live`; the specification is pinned by hash in `reference/PINNED.json` |
| **Settings** - one file, one command, one menu | **done** | `jev-config`, `jev-config doctor` |
| **Voice hook** - decide command vs dictation, then act or type | **done, in daily use** | measured: 0.24-0.33 s to paste a dictation, 6-15 ms to hand a command off |
| **Execution** - press, write, type into the chosen element | **not yet (T6)** | the walk can observe but not act |
| **CLI** - `jev-use-rs --table --app <App>` | **not yet (T7)** | until then the walk is driven through `crates/jev-ax/examples/walk.rs`, which prints the reference's J

Also filed under Agents & browsers

  1. 0295

    Flight search with Browser Use

    Breaking: Browser Use + Jev = Ultrafast ⚡ Findings flights took 7s and cost only $0.0039 🤯 > new action space every step > DOM state space > small LLM fallback to type (this video is at 1x speed btw) Built a tiny open source browser agent. try it below ↓

    @gregpr07 · Agents & browsers · ~$0.004 · ~7 s

  2. 0500

    Jev: AI Decisions as a Typed Function Call | StackToHeap

    # Jev: AI Decisions as a Typed Function Call What I learned using TypeSafe's Jev for code review, and why its more surprising uses are browser control and generative UI.…

    stacktoheap.com · Agents & browsers

  3. 0319

    x-scanner

    Chrome extension that labels every post you scroll past on X with six typed questions per post, and counts what it costs in the corner.

    oso95 · Agents & browsers

  4. 0318

    Vibe Check for X

    Chrome extension that scores draft posts and reply context before posting; optional media descriptions come from a separate vision model.

    RafalWilinski · Agents & browsers