p-i-

train:
  - input: |-
      🔵 🔴 🟢 ⚫️ ⚫️ ⚫️ ⚫️
      🔴 🟢 ⚫️ ⚫️ ⚫️ ⚫️ ⚫️
      🟢 ⚫️ ⚫️ ⚫️ ⚫️ ⚫️ ⚫️
      ⚫️ ⚫️ ⚫️ ⚫️ ⚫️ ⚫️ ⚫️
      ⚫️ ⚫️ ⚫️ ⚫️ ⚫️ ⚫️ ⚫️
      ⚫️ ⚫️ ⚫️ ⚫️ ⚫️ ⚫️ ⚫️
      ⚫️ ⚫️ ⚫️ ⚫️ ⚫️ ⚫️ ⚫️
    output: |-

p-i-

from pathlib import Path
from os import environ
from time import sleep
from textwrap import dedent

import nestedtext as nt
import httpx
import arrow

API_KEY = environ.get('ANTHROPIC_API_KEY', None)

p-i-

# PS1 etc requires wrapping non-printing characters
_np() { echo -ne "\[$1\]" ; }

# Generating ANSI escape codes
esc() { echo -ne "\e[$1m" ; } ;  _esc() { _np "$(esc $1)" ; }

# Set foreground / background R G B (0-5 each)
rgb() { esc "$1;5;$((16 + 36 * $2 + 6 * $3 + $4))" ; }  # FG/BG r g b

fg() { rgb 38 $@ ; } ;  _fg() { _np "$(rgb 38 $@)" ; }

p-i-

p-i-

'''
# Realtime (Ultralow-latency) Voice Activity Detection (VAD) using WebRTC VAD
v3.2
π 17 Apr 2024

## Usage
```python
    def on_segment(segment):  # np.array of np.int16
        print(f'☏ CALLBACK: Got segment: {len(segment)} frames')

p-i-

'''
Adapted by π 15 Apr 2024
    from https://github.com/davabase/whisper_real_time/blob/master/transcribe_demo.py
    ... which appears to have been itself pilfered from:
        https://github.com/JihyeokKim/whisper_ros/tree/master/whisper_ros/src

To run on macOS:
    - `brew install ffmpeg portaudio`
    - Create requirements.txt:
    ```

p-i-

p-i-

p-i-

p-i-