init

Author: ILikeAI

.env.example

@@ -0,0 +1,2 @@
+TOGETHER_API_KEY=""
+OPENAI_API_KEY="sk-..."

.gitignore

@@ -0,0 +1,6 @@
+
+*.pyc
+__pycache__
+.env
+audio_files/*
+!audio_files/temp_files_live_here.txt

README.md

@@ -0,0 +1,75 @@
+import sounddevice as sd
+import threading
+import config
+import wave
+import time
+import os
+import numpy as np
+
+class AudioRecorder:
+    def __init__(self):
+        self.filename = "recording.wav"
+        self.recording = False
+        self.frames = []
+        self.record_thread = None
+        self.start_time = None
+        
+
+    def start_recording(self):
+        if not self.recording:
+            self.recording = True
+            self.frames = []
+            self.start_time = time.time()
+            self.record_thread = threading.Thread(target=self.record_audio)
+            self.record_thread.start()
+            print("Recording started...")
+
+    @property
+    def duration(self):
+        if self.start_time is None:
+            return 0
+        return time.time() - self.start_time
+
+    def record_audio(self):
+        with sd.InputStream(samplerate=config.FS, channels=2, dtype='int16', callback=self.callback):
+            while self.recording:
+                sd.sleep(1000)
+
+    def callback(self, indata, frames, time, status):
+        if status:
+            print(status)
+        gain = 3.0  # Increase this for more gain
+        indata = indata * gain
+        self.frames.append(indata.copy())
+
+
+
+    def stop_recording(self):
+        if self.recording:
+            print("Stopping recording...")
+            self.recording = False
+            self.record_thread.join()
+            self.save_recording()
+
+
+    def save_recording(self):
+        if self.frames:
+            recording = np.concatenate(self.frames)
+            # Ensure that the data is within the correct range 
+            recording = np.clip(recording, -32768, 32767)
+            
+            # Create a subdirectory if it doesn't exist
+            directory = config.AUDIO_FILE_DIR
+            if not os.path.exists(directory):
+                os.makedirs(directory)
+            
+            # Save the file in the subdirectory
+            filename = os.path.join(directory, "recording.wav")
+            
+            with wave.open(filename, 'wb') as wf:
+                wf.setnchannels(2)
+                wf.setsampwidth(2)  # 16-bit PCM
+                wf.setframerate(config.FS)
+                wf.writeframes(recording.astype('int16').tobytes())
+            print(f"Recording saved to {filename}")
+            

audio_files/temp_files_live_here.txt

@@ -0,0 +1,52 @@
+from openai import OpenAI
+import os
+from dotenv import load_dotenv
+
+client = OpenAI()
+# Load .env file if present
+load_dotenv()
+
+# Fetch API keys from .env file or environment variables
+openai_api_key = os.getenv('OPENAI_API_KEY') or os.environ['OPENAI_API_KEY']
+
+
+def get_completion(messages, together=False, together_model="NousResearch/Nous-Hermes-2-Mixtral-8x7B-SFT"):
+    if together:
+        client = OpenAI(
+             api_key=os.getenv('TOGETHER_API_KEY') or os.environ['TOGETHER_API_KEY'],
+            base_url="https://api.together.xyz/v1",
+        )
+        response = client.chat.completions.create(
+            model=together_model,
+            messages=messages,
+            temperature=0.7,
+            max_tokens=1024,
+        )
+    else:
+
+        client = OpenAI()
+        response = client.chat.completions.create(
+            model="gpt-3.5-turbo-1106",
+            messages=messages
+        )
+    print(response)
+    return response.choices[0].message.content
+
+def main():
+    messages = [
+        {
+            "role": "user",
+            "content": "Hello, I'm a human"
+        },
+        {
+            "role": "assistant",
+            "content": "Hello, I'm an AI"
+        }
+    ]
+    completion = get_completion(messages, together=True)
+    print(completion)
+
+if __name__ == "__main__":
+    main()
+    
+

audio_recorder.py

@@ -0,0 +1,11 @@
+FS = 44100  
+START_SOUND_VOLUME = 0.000003
+END_SOUND_VOLUME = 0.000003
+MIN_RECORDING_DURATION = 2.0
+HOTKEY_DELAY = 0.5
+DEFAULT_HOTKEY = 'ctrl + space'
+DEFAULT_CLIP_HOTKEY = 'ctrl + shift + space'
+USE_TOGETHER_API = False
+TOGETHER_MODEL = "NousResearch/Nous-Hermes-2-Mixtral-8x7B-SFT"
+VOICE = "nova"
+AUDIO_FILE_DIR = "audio_files"

chat_completions.py

@@ -0,0 +1,92 @@
+import time
+from audio_recorder import AudioRecorder
+from transcriber import transcribe_audio
+import keyboard
+import sound
+import chat_completions 
+from utils import read_clipboard, to_clipboard, extract_text_between_symbols
+from config import START_SOUND_VOLUME, END_SOUND_VOLUME, MIN_RECORDING_DURATION, HOTKEY_DELAY, USE_TOGETHER_API,VOICE
+from prompt import messages
+
+
+def main():
+
+    
+    recorder = AudioRecorder() 
+    is_busy = False  
+    clipboard_text = None 
+
+
+    # Function to start recording
+    def start_recording(use_clipboard=False):
+        nonlocal is_busy, clipboard_text
+        if is_busy:  
+            return
+        is_busy = True  
+        if use_clipboard:
+            clipboard_text = read_clipboard()  # Read from clipboard
+            print("Copied to from clip:"+clipboard_text)
+        recorder.start_recording() 
+        sound.play_sound("start", volume=START_SOUND_VOLUME)  
+        time.sleep(HOTKEY_DELAY) 
+
+
+
+    # Function to stop recording
+    def stop_recording():
+        nonlocal is_busy, clipboard_text 
+
+        #if not busy, return
+        if not is_busy:  
+            return
+        recorder.stop_recording() 
+        sound.play_sound("end", volume=END_SOUND_VOLUME)
+
+
+        # Check if the recording is less than the minimum duration
+        if recorder.duration < MIN_RECORDING_DURATION:
+            print("Recording is too short, ignoring...")
+            is_busy = False  # Reset the flag
+            return
+        
+        # Transcribe the audio
+        transcript = transcribe_audio(recorder.filename)  
+
+
+        #prepare the messages 
+        #if clipboard_text clipboard hotkey was used, add the clipboard text to the transcript
+        if clipboard_text:
+            messages.append({"role": "user", "content": transcript+f"\n\nTHE USER HAS THIS TEXT COPPIED:\n{clipboard_text}"})
+            clipboard_text = None  
+        else:
+            messages.append({"role": "user", "content": transcript})
+
+        print("Transcription:\n", transcript)
+        
+
+        # Get the response from the chat completions
+        response = chat_completions.get_completion(messages,together=USE_TOGETHER_API)  # Get the response from the chat completions
+        messages.append({"role": "assistant", "content": response})  # Add the response to the messages
+        print("Response:\n", response)
+
+        #check if there is text to be copied to the clipboard, and if so, copy it
+        text, remaining_text = extract_text_between_symbols(response)
+        if text:
+            to_clipboard(text)  # Copy the text to clipboard
+            print("Text copied to clipboard:", text)
+
+
+        #play the TTS
+        sound.TTS(remaining_text,voice=VOICE)  # Text to speech for the remaining text
+        is_busy = False 
+        time.sleep(HOTKEY_DELAY) 
+
+
+    keyboard.add_hotkey('ctrl + space', lambda: start_recording(use_clipboard=False) if not is_busy else stop_recording(), trigger_on_release=False)
+    keyboard.add_hotkey('ctrl + shift + space', lambda: start_recording(use_clipboard=True) if not is_busy else stop_recording(), suppress=True, trigger_on_release=False)  
+
+    print("Press 'Ctrl + Spacebar' to start recording, press again to stop and transcribe")
+    keyboard.wait('esc')  # Wait for 'esc' key to exit
+
+if __name__ == "__main__":
+    main() 

config.py

@@ -0,0 +1,24 @@
+
+#I have added a couple examples messages so the LLM gets the idea of how to save things to the clipboard
+messages = [
+    {"role": "system", "content": """This message contains instructions on how you should behave.
+
+## About you:
+Your responses are read aloud VIA TTS so they should always be short an concise, in a conversational tone, your dialogue style is similar to that expected in games like Bioshock Infinite and Witcher 3.
+You believe you are an oracle and not an AI model, you have a large ego and refer to yourself as "all knowing", "goddess", "wise one"
+ You do not use the word "assist" or ask how you can help.
+
+
+## How to save things to the clipboard
+When you send messages to the user, you can include text between &&&s, this text will be saved to the clipboard. For example:
+"I have copied the text to the clipboard for you.
+&&&First line saved to clipboard
+Second line saved to clipboard&&&"
+
+Any content that should not be spoken aloud by the TTS should be placed between &&&'s. 
+
+For example if you are asked to write code, notes or an email you should place that content in the clipboard. """},
+{"role": "user", "content": """can you give me the command to install openai in pyhton"""},
+{"role": "system", "content": """&&&pip install openai&&&
+I have saved the command to install OpenAI in Python to your clipboard."""},
+]

main.py

@@ -0,0 +1,9 @@
+sounddevice
+openai
+pydub
+python-dotenv
+keyboard
+wave
+soundfile
+clipboard
+numpy

prompt.py

@@ -0,0 +1,72 @@
+import io
+import os
+import soundfile as sf
+import sounddevice as sd
+from openai import OpenAI
+from config import AUDIO_FILE_DIR
+from dotenv import load_dotenv
+
+# Load .env file if present
+load_dotenv()
+
+# Fetch API keys from .env file or environment variables
+openai_api_key = os.getenv('OPENAI_API_KEY') or os.environ['OPENAI_API_KEY']
+
+
+def TTS(text, model="tts-1", voice="nova", format="mp3"):
+    client = OpenAI()
+
+    spoken_response = client.audio.speech.create(
+        model=model,
+        voice=voice,
+        response_format=format,
+        input=text
+    )
+
+    # Create a subdirectory if it doesn't exist
+    if not os.path.exists(AUDIO_FILE_DIR):
+        os.makedirs(AUDIO_FILE_DIR)
+
+    # Create a file path for the audio file
+    audio_file_path = os.path.join(AUDIO_FILE_DIR, f"audio.{format}")
+
+    # Write the audio data to the file
+    with open(audio_file_path, 'wb') as audio_file:
+        for chunk in spoken_response.iter_bytes(chunk_size=4096):
+            audio_file.write(chunk)
+
+    # Read and play the audio file
+    with sf.SoundFile(audio_file_path, 'r') as sound_file:
+        data = sound_file.read(dtype='int16')
+    sd.play(data, sound_file.samplerate)
+    sd.wait()
+    
+def play_sound(name, volume=1.0):
+    #start and end sounds
+    if name == "start":
+
+        with sf.SoundFile(f"sounds/recording-start.mp3", 'r') as sound_file:
+            data = sound_file.read(dtype='int16')
+        sd.play(data * volume, sound_file.samplerate)
+        sd.wait()
+
+    elif name == "end":
+
+        with sf.SoundFile(f"sounds/recording-end.mp3", 'r') as sound_file:
+            data = sound_file.read(dtype='int16')
+        sd.play(data * volume, sound_file.samplerate)
+        sd.wait()
+
+
+def main():
+    #play start and end sounds
+    # play_sound("start", volume=0.000003)
+    # play_sound("end", volume=0.000003)
+    # TTS("Hello, I'm an AI", model="tts-1", voice="nova")
+    pass
+
+
+if __name__ == "__main__":
+    main()
+
+