Explore Baguette and OpenLess: Command‑Line iOS Simulator Control & Cross‑Platform Voice Input

Baguette: Command‑Line iOS Simulator Control

iOS developers rely on the Simulator daily, but Simulator.app is heavyweight, consumes significant memory, and cannot run on headless servers. Baguette addresses this by providing full simulator control without launching the app.

Key capabilities :

Simulator management – baguette list shows all simulators, baguette boot --udid xxx starts one, baguette shutdown --udid xxx stops it.

Screen streaming – real‑time MJPEG or H.264 streams at 60 fps, configurable bitrate and frame rate, output to stdout or via WebSocket to a browser.

Gesture injection – terminal commands inject tap, swipe, pinch, pan, hardware button events, and even forward Mac keyboard input.

Accessibility tree query – baguette describe-ui returns each UI element’s role, label, value, and coordinates as JSON, useful for automation or AI control.

Web UI console – baguette serve starts a local web service; visiting http://localhost:8421/simulators shows a list of simulators with live video, and the /farm panel arranges multiple devices in a grid for full‑screen control.

Technical implementation uses the private Xcode 26 framework SimulatorKit and IndigoHIDMessageForMouseNSEvent for gesture injection, avoiding dylib injection and per‑app adaptation. The accessibility tree leverages the private AccessibilityPlatformTranslation framework. The project follows a TDD approach with over 290 test cases and mock implementations, allowing swift test to run all tests without launching a simulator.

Installation command:

brew install tddworks/tap/baguette baguette serve open http://localhost:8421/simulators

Requirements: macOS 15+ and Xcode 26, Apple Silicon only.

GitHub: https://github.com/tddworks/baguette

OpenLess: Open‑Source Cross‑Platform Voice Input

Typing can be slow, especially when you need to feed ideas quickly into AI assistants. OpenLess lets you press a global shortcut, speak, and have AI restructure the utterance into ready‑to‑use text.

The standout feature is the “AI structuring mode” that converts free‑form speech into a structured prompt suitable for ChatGPT, Claude, or Cursor.

You say: <code>uh… so… I want ChatGPT to write me a SQL query, from the orders table get last month's orders, group by customer, sort by amount desc, top ten</code> OpenLess returns after one second: <code>Please write a SQL query that: - Pulls orders from last month from the `orders` table. - Groups by customer. - Sorts by total amount, descending. - Returns the top 10 rows only.</code>

Core functions :

Four output modes – Raw, Light Polish, Structured (AI Prompt), Formal.

Two recording modes – Toggle (press once to start, press again to stop) and Push‑to‑Talk (hold to speak, release to stop); ESC cancels at any stage.

Global hotkey – implemented with macOS CGEventTap or Windows low‑level keyboard hook, works in any text field.

Custom dictionary – add proprietary terms, product names, or personal names; they are synced to the ASR hot‑word list for higher recognition accuracy.

Compared with closed‑source competitors such as Typeless, Wispr Flow, and Superwhisper, OpenLess is fully open source and local‑first. It uses Volcano Engine’s streaming ASR together with an Ark/DeepSeek‑compatible chat‑completion API; users register their own accounts, pay per usage, and keep audio data on the device.

GitHub: https://github.com/appergb/openless