`interface.py`

OS/01/interface/interface.py

@@ -1,27 +1,104 @@
-while True:
+import redis
-
+import RPi.GPIO as GPIO
-    if button.is_pressed():
+import asyncio
-        send_to_main(user_start_message)
+import websockets
-        send_to_websocket(user_start_message)
+import sounddevice as sd
-
+import numpy as np
-        audio_chunks = []
+import time
-        for audio_chunk in listen():
+import re
-            audio_chunks.append(chunk)
+
-            if not button.is_pressed():
+def transcribe(audio_chunks):
-                break
+    pass # (todo)
-
+
-        text = stt(audio_chunks)
+def say(text):
-        send_to_main(text)
+    # This should immediatly stop if button is pressed (if GPIO.input(18))
-        send_to_websocket(user_end_message)
+    pass # (todo)
-    
+
-    chunk = get_from_queue('to_io')
+# Connect to button
-    if chunk:
+GPIO.setmode(GPIO.BCM)
-        send_to_websocket(chunk)
+GPIO.setup(18, GPIO.IN, pull_up_down=GPIO.PUD_UP)
-        sentence += chunk["content"]
+
-        if is_full_sentence(sentence)
+# Set the duration and sample rate for the mic
-            tts(sentence)
+chunk_duration = 0.5  # seconds
-            sentence = []
+sample_rate = 44100  # Hz
-
+
-    message = check_websocket()
+# Set up Redis connection
-    if message:
+r = redis.Redis(host='localhost', port=6379, db=0)
-        send_to_main(message)
+
+# Define some standard, useful messages
+user_start_message = {"role": "user", "type": "message", "start": True}
+user_start_message = {"role": "user", "type": "message", "start": True}
+
+# Set up websocket connection
+websocket = websockets.connect('ws://localhost:8765')
+
+# This is so we only say() full sentences
+accumulated_text = ""
+def is_full_sentence(text):
+    return text.endswith(('.', '!', '?'))
+def split_into_sentences(text):
+    return re.split(r'(?<=[.!?])\s+', text)
+
+async def send_to_websocket(message):
+    async with websocket as ws:
+        await ws.send(message)
+
+async def check_websocket():
+    async with websocket as ws:
+        message = await ws.recv()
+        return message
+
+def main():
+    while True:
+
+        # If the button is pushed down
+        if not GPIO.input(18):
+
+            # Send start message to core and websocket
+            r.rpush('to_core', user_start_message)
+            send_to_websocket(user_start_message)
+
+            # Record audio from the microphone in chunks
+            audio_chunks = []
+
+            # Continue recording until the button is released
+            while not GPIO.input(18):
+                chunk = sd.rec(int(chunk_duration * sample_rate), samplerate=sample_rate, channels=2)
+                sd.wait()  # Wait until recording is finished
+                audio_chunks.append(chunk)
+
+            # Transcribe
+            text = transcribe(audio_chunks)
+
+            message = {"role": "user", "type": "message", "content": text, "time": time.time()}
+
+            # Send message to core and websocket
+            r.rpush('to_core', message)
+            send_to_websocket(message)
+        
+        # Send out anything in the to_interface queue
+        chunk = r.lpop('to_interface')
+        if chunk:
+            send_to_websocket(chunk)
+            accumulated_text += chunk["content"]
+            
+            # Speak full sentences out loud
+            sentences = split_into_sentences(accumulated_text)
+            if is_full_sentence(sentences[-1]):
+                for sentence in sentences:
+                    say(sentence)
+                accumulated_text = ""
+            else:
+                for sentence in sentences[:-1]:
+                    say(sentence)
+                accumulated_text = sentences[-1]
+        else:
+            say(accumulated_text)
+            accumulated_text = ""
+
+        message = check_websocket()
+        if message:
+            r.rpush('to_core', message)
+
+if __name__ == "__main__":
+    main()