# lel

X11 transcription daemon that records audio, runs Whisper, logs the transcript, and can optionally run AI post-processing before injecting text.

## Requirements

- X11 (not Wayland)
- `ffmpeg`
- `whisper` (OpenAI Whisper CLI)
- `xclip`
- Tray icon deps: `libappindicator3` and `gtk3` (required by `systray`)

## Build

```bash
make build
```

## Config

Create `~/.config/lel/config.json`:

```json
{
  "hotkey": "Cmd+m",
  "ffmpeg_input": "pulse:default",
  "ffmpeg_path": "",
  "whisper_model": "base",
  "whisper_lang": "en",
  "whisper_device": "cpu",
  "whisper_extra_args": "",
  "record_timeout_sec": 120,
  "whisper_timeout_sec": 300,
  "segment_sec": 5,
  "streaming": false,
  "injection_backend": "clipboard",

  "ai_enabled": true,
  "ai_provider": "ollama",
  "ai_model": "llama3.2:3b",
  "ai_temperature": 0.0,
  "ai_system_prompt_file": "",
  "ai_base_url": "http://localhost:11434",
  "ai_api_key": "",
  "ai_timeout_sec": 20
}
```

Env overrides:

- `WHISPER_MODEL`, `WHISPER_LANG`, `WHISPER_DEVICE`, `WHISPER_EXTRA_ARGS`
- `WHISPER_FFMPEG_IN`
- `WHISPER_STREAM`, `WHISPER_SEGMENT_SEC`, `WHISPER_TIMEOUT_SEC`
- `LEL_RECORD_TIMEOUT_SEC`, `LEL_HOTKEY`, `LEL_INJECTION_BACKEND`
- `LEL_FFMPEG_PATH`
- `LEL_AI_ENABLED`, `LEL_AI_PROVIDER`, `LEL_AI_MODEL`, `LEL_AI_TEMPERATURE`, `LEL_AI_SYSTEM_PROMPT_FILE`
- `LEL_AI_BASE_URL`, `LEL_AI_API_KEY`, `LEL_AI_TIMEOUT_SEC`

## Run manually

```bash
./leld --config ~/.config/lel/config.json
```

Disable the tray icon:

```bash
./leld --no-tray
```

## systemd user service

```bash
mkdir -p ~/.local/bin
cp leld lelctl ~/.local/bin/
cp systemd/lel.service ~/.config/systemd/user/lel.service
systemctl --user daemon-reload
systemctl --user enable --now lel
```

## Usage

- Press the hotkey once to start recording.
- Press it again to stop and transcribe.
- The transcript is logged to stderr.

Execution flow (single in-flight state machine):

- `recording` -> `transcribing` -> `processing` (optional) -> `outputting` -> `idle`

Injection backends:

- `clipboard`: copy to clipboard and inject via Ctrl+V (requires `xclip` + `xdotool`)
- `injection`: type the text with simulated keypresses (requires `xdotool`)

AI providers:

- `ollama`: calls the local Ollama HTTP API (`/api/generate`)
- `openai_compat`: calls a chat-completions compatible API (`/v1/chat/completions`)

Dependency checks:

- Recording requires `ffmpeg` (or set `ffmpeg_path`)
- Transcribing uses the `whisper` CLI
- Outputting requires `xclip` (and `xdotool` for injection backends)

Control:

```bash
lelctl status
lelctl reload
lelctl stop
```