Codex Micro + Typeless

中文 | English

Use the Codex Micro microphone key to launch Typeless while keeping the other native Codex controls unchanged.

让 Codex Micro 的麦克风键启动 Typeless，同时保留其他 Codex 官方按键功能。

Community project. Not affiliated with or endorsed by OpenAI, Work Louder, or Typeless.

中文说明

解决什么问题

Codex Micro 的默认麦克风键会调用 ChatGPT/Codex 自带的语音功能。Codex 应用可以配置官方按键，但 Work Louder Input 不能直接复制 ChatGPT/Codex 创建的官方图层。因此，单独在任意空白图层中设置 Typeless，会导致其他 Codex 专用按键失效。

本仓库提供一个可导入的图层：

• 保留官方 Codex Agent Keys、Command Keys、旋钮和摇杆映射；
• 只把麦克风键替换为 Shift + Command + D；
• 该快捷键默认用于启动 Typeless；
• 其他 Codex 按键继续由官方功能处理。

已验证环境

• macOS
• Codex Micro
• Work Louder Input 0.18.0
• Codex Micro firmware 0.6.1
• Typeless 全局快捷键：Shift + Command + D

Windows 和其他 Input/firmware 版本尚未验证。升级软件或固件后，建议重新核对图层和快捷键。

快速配置

1. 安装并启动 Typeless。
2. 在 Typeless 设置中，将全局听写快捷键设置为 Shift + Command + D。
3. 连接 Codex Micro，打开 Work Louder Input。
4. 进入 Codex Micro > Keymap > Layers。
5. 导入 examples/codex-typeless-layer.json。
6. 确保图层顺序为：
   • 第 1 层：官方 Layer 1
   • 第 2 层：Codex + Typeless
7. 如果两者之间存在一个旧的空白图层，请先导出备份，再删除该空白图层，使 Codex + Typeless 成为第 2 层。
8. 等待 Work Louder Input 显示 layout updated。
9. 退出并重新打开 Work Louder Input，等待键盘数据读取完成。
10. 打开 Codex 测试：普通按键应继续执行 Codex 官方功能，麦克风键应启动 Typeless。

导入后，Work Louder Input 可能显示 To edit this layer please use the ChatGPT app。这是正常现象，因为该图层保留了 Codex 的专用 vendor keycodes。

回滚

1. 切回官方第 1 层；或
2. 删除 Codex + Typeless 图层；或
3. 导入你在修改前导出的图层备份。

不要删除官方 Layer 1。公开问题或日志前，请移除设备序列号、本机路径以及其他个人信息。

让 Codex 帮你配置

可以。让 Codex 读取本仓库后，它能够核对环境、备份配置、打开 Work Louder Input 并导入示例文件。建议复制 CODEX_SETUP_PROMPT.md 中的提示词给 Codex。

这不是无条件的一键安装：Codex 需要获得必要的本机操作权限，Codex Micro 必须已连接，Typeless 快捷键必须一致，而且软件版本差异可能导致界面或图层结构变化。最后仍需要你亲自按一次麦克风键确认。

自定义其他听写工具

示例 JSON 中的 actions[0].keyInputs 定义了快捷键：

[
  { "keycode": "KC_LGUI", "delay": 0, "actionType": 1 },
  { "keycode": "KC_LSFT", "delay": 0, "actionType": 1 },
  { "keycode": "KC_D", "delay": 0, "actionType": 1 }
]

你可以把它改成 Wispr Flow、系统听写或其他工具的全局快捷键，但不同系统和软件版本需要自行验证。

English

What this fixes

By default, the Codex Micro microphone key launches ChatGPT/Codex voice. The Codex app can configure native controls, but Work Louder Input cannot directly duplicate the official ChatGPT/Codex layer. If Typeless is assigned on an otherwise empty layer, the other Codex-specific controls stop working.

This repository provides an importable layer that:

• preserves the native Codex Agent Keys, Command Keys, encoder, and joystick mappings;
• replaces only the microphone key with Shift + Command + D;
• uses that shortcut to launch Typeless;
• keeps the rest of the Codex controls working normally.

Tested environment

• macOS
• Codex Micro
• Work Louder Input 0.18.0
• Codex Micro firmware 0.6.1
• Typeless global shortcut: Shift + Command + D

Windows and other Input/firmware versions have not been tested. Re-check the layer order and shortcut after software or firmware updates.

Quick setup

1. Install and launch Typeless.
2. Set the Typeless global dictation shortcut to Shift + Command + D.
3. Connect the Codex Micro and open Work Louder Input.
4. Go to Codex Micro > Keymap > Layers.
5. Import examples/codex-typeless-layer.json.
6. Make sure the layer order is:
   • Layer 1: the official Layer 1
   • Layer 2: Codex + Typeless
7. If an old empty layer sits between them, export it as a backup and delete it so Codex + Typeless becomes Layer 2.
8. Wait until Work Louder Input shows layout updated.
9. Quit and reopen Work Louder Input, then wait for it to finish retrieving keyboard data.
10. Test in Codex: normal keys should retain their native Codex actions, while the microphone key should launch Typeless.

After import, Work Louder Input may show To edit this layer please use the ChatGPT app. This is expected because the layer retains Codex vendor keycodes.

Rollback

1. Switch back to the official first layer; or
2. delete the Codex + Typeless layer; or
3. import the layer backup you exported before making changes.

Do not delete the official Layer 1. Remove device serial numbers, local paths, and other personal information before sharing diagnostics.

Let Codex configure it

Yes. After reading this repository, Codex can check the environment, back up the current configuration, open Work Louder Input, and import the example file. Copy the prompt in CODEX_SETUP_PROMPT.md into Codex.

This is not an unconditional one-click installer. Codex needs the required local-control permissions, the Codex Micro must be connected, the Typeless shortcut must match, and software-version differences may change the UI or layer structure. You still need to press the physical microphone key once for final confirmation.

Use another dictation tool

The shortcut is stored in actions[0].keyInputs inside the example JSON:

[
  { "keycode": "KC_LGUI", "delay": 0, "actionType": 1 },
  { "keycode": "KC_LSFT", "delay": 0, "actionType": 1 },
  { "keycode": "KC_D", "delay": 0, "actionType": 1 }
]

You can replace it with the global shortcut for Wispr Flow, macOS Dictation, or another tool, but test the result on your own OS and software versions.

License

MIT. See LICENSE.