SchoolProjects/PizzAI
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

added chatgpt wrap, tts phone number splitting

Browse Source
This commit is contained in:
Looki2000
2023-12-02 17:32:01 +01:00
parent f18e41476f
commit 24f909abff
5 changed files with 129 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

40
frontend/core.py
View File

@@ -1,5 +1,6 @@
from vad_recorder import VADRecorder
from tts_stream import TTSStream
from chatgpt_wrap import ChatGPTWrap
from faster_whisper import WhisperModel
import torch
import time
@@ -8,7 +9,9 @@ import numpy as np
class Core:
def __init__(self, whisper_model_name = "large-v3"):
def __init__(self, whisper_model_name = "large-v3", use_chatgpt_placeholder = False):
self.use_chatgpt_placeholder = use_chatgpt_placeholder
self.device = "cuda" if torch.cuda.is_available() else "cpu"
print("\n=======================================")
print(f"Using {self.device.capitalize()} for:")
@@ -16,19 +19,21 @@ class Core:
print(" - TTS")
print("=======================================\n")
print("Loading Whisper model... ", end="")
print("Loading Faster Whisper model... ", end="")
self.whisper_model = WhisperModel(whisper_model_name, device=self.device, compute_type="float16")
print("Done!")
# VADRecorder and TTSStream have their own console loading messages
# VADRecorder, TTSStream and ChatGPTWrap have their own console loading messages
self.vad_rec = VADRecorder()
self.tts = TTSStream(device=self.device)
self.gpt_wrap = ChatGPTWrap(use_chatgpt_placeholder)
def set_order_settings(self, phone_number, order_items, delivery_address):
def set_order_settings(self, phone_number, order_items, delivery_address, payment_method):
self.phone_number = phone_number
self.order_items = order_items
self.delivery_address = delivery_address
self.payment_method = payment_method
def set_speech_recog_settings(self, speech_recog_timeout, audio_input_device_name, audio_output_device_name, window_size_sec, vad_threshold, min_silence_duration_ms, speech_pad_ms):
@@ -62,7 +67,12 @@ class Core:
)
print("Done!")
self.gpt_wrap.init_order(
self.phone_number,
self.order_items,
self.delivery_address,
self.payment_method
)
@@ -116,9 +126,19 @@ class Core:
print(speech_recog_text)
print("-----------------------------------------\n\n")
time.sleep(1) # fake chatgpt delay
gpt_response = self.gpt_wrap.get_response(speech_recog_text)
print("-----------------------------------------")
if self.use_chatgpt_placeholder:
print("!!!!! CHATGPT PLACEHOLDER RESPONSE !!!!!!")
else:
print("!!!!!!!!!!! CHATGPT RESPONSE !!!!!!!!!!!!")
print("-----------------------------------------")
print(gpt_response)
print("-----------------------------------------\n\n")
gpt_response = "czat dżi pi ti plejsholder 1 2 3"
if not self.assistant_running:
break
@@ -127,7 +147,11 @@ class Core:
# tts
print("Speech synthesis stream started!")
self.tts.tts_speak(gpt_response)
self.tts.tts_speak(gpt_response.replace(" CALLEND", ""))
if "CALLEND" in gpt_response:
self.assistant_stop()
#print(len(audios_for_whisper), time.perf_counter() - last_recog_time, len(speech_recog_text))