small fixes

frontend/core.py

@@ -5,6 +5,7 @@ from faster_whisper import WhisperModel
 import torch
 import time
 import numpy as np
+import re
 
 
 
@@ -84,7 +85,7 @@ class Core:
 
         self.assistant_running = True
 
-        while self.assistant_running:
+        while True:
             if self.vad_rec.speech:
                 last_recog_time = time.perf_counter()
 
@@ -114,8 +115,6 @@ class Core:
 
 
 
-
-
             elif time.perf_counter() - last_recog_time > self.speech_recog_timeout and len(speech_recog_text) > 0:
                 speech_recog_text = speech_recog_text.strip()
 
@@ -130,6 +129,18 @@ class Core:
 
                 gpt_response = self.gpt_wrap.get_response(speech_recog_text)
 
+                # separate long sequences of numbers in text string (for example 123456789) into packets of 3 (123 456 789)
+                gpt_response = re.sub(r"(\d{3})(?=\d)", r"\1 ", gpt_response)
+
+                # Add space on the right side of numbers
+                gpt_response = re.sub(r'(\d)([^\d\s])', r'\1 \2', gpt_response)
+                # Add space on the left side of numbers
+                gpt_response = re.sub(r'([^\d\s])(\d)', r'\1 \2', gpt_response)
+
+                # replace "ul." with "ulica" (non case sensitive)
+                gpt_response = re.sub(r"ul\.", "ulica", gpt_response, flags=re.IGNORECASE)
+
+
                 print("-----------------------------------------")
                 if self.use_chatgpt_placeholder:
                     print("!!!!! CHATGPT PLACEHOLDER RESPONSE !!!!!!")
@@ -159,6 +170,9 @@ class Core:
 
             time.sleep(0.01)
 
+            if not self.assistant_running:
+                break
+
         # set assistant_running back to True to indicate that the loop has exited
 
     def assistant_stop(self):

frontend/core_test.ipynb

@@ -10,14 +10,21 @@
     "\n",
     "core = Core(\n",
     "    whisper_model_name = \"large-v3\",\n",
-    "    use_chatgpt_placeholder = True\n",
-    ")\n",
-    "\n",
+    "    use_chatgpt_placeholder = False\n",
+    ")\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
     "core.set_order_settings(\n",
     "    phone_number = \"123456789\",\n",
-    "    order_items = \"1x margharitta\\n2x sos majonezowy\",\n",
-    "    delivery_address = \"ul. Amogusowa 1337, Suski Małe\",\n",
-    "    payment_method = \"karta\"\n",
+    "    order_items = \"2x margherita\\n3x sos czosnkowy\\n2x pepsi 1.5l\",\n",
+    "    delivery_address = \"ulica Kielecka 16A, Opole\",\n",
+    "    payment_method = \"gotówka\"\n",
     ")\n",
     "\n",
     "core.set_speech_recog_settings(\n",
@@ -28,12 +35,26 @@
     "    vad_threshold = 0.6,\n",
     "    min_silence_duration_ms = 150,\n",
     "    speech_pad_ms = 0\n",
-    ")\n",
-    "\n",
+    ")\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
     "core.set_tts_settings(\n",
     "    speaker_wav = \"voices/lector.wav\"\n",
-    ")\n",
-    "\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
     "try:\n",
     "    core.assistant_start()\n",
     "except (KeyboardInterrupt, SystemExit):\n",

frontend/system_instructions.txt

@@ -2,7 +2,7 @@ You are a bot that will act as a guy that orders a pizza you will be connected o
 
 DATA:
 phone: {phone_number}
-delivery location: {delivery_location}
+delivery location: {delivery_location} (PIZZA IS FOR DELIVERY. NOT for dine-in.)
 paymentMethod: {payment_method}
 OrderItems:
 {order_items}

frontend/tts_stream.py

@@ -78,9 +78,6 @@ class TTSStream:
     def tts_speak(self, text):
         self.play_buffer_size = 512
 
-        # separate long sequences of numbers in text string (for example 123456789) into packets of 3 (123 456 789)
-        text = re.sub(r"(\d{3})(?=\d)", r"\1 ", text)
-
 
         # open pyaudio stream
         p = pyaudio.PyAudio()