killian
11 months ago
parent
dd5e87bbc5
commit
75752db415
@ -0,0 +1,9 @@
|
||||
def openSafari():
|
||||
"""open safari"""
|
||||
|
||||
import os
|
||||
os.system('open -a Safari')
|
||||
|
||||
import os
|
||||
os.system('osascript -e \'tell application "Safari" to open location "https://www.youtube.com"\'')
|
||||
|
||||
@ -0,0 +1,6 @@
|
||||
import threading
|
||||
import time
|
||||
|
||||
def print_message():
|
||||
"""None"""
|
||||
time.sleep(30)
|
||||
@ -1,38 +0,0 @@
|
||||
import redis
|
||||
import json
|
||||
import time
|
||||
|
||||
# Set up Redis connection
|
||||
r = redis.Redis(host='localhost', port=6379, db=0)
|
||||
|
||||
def main(interpreter):
|
||||
|
||||
while True:
|
||||
|
||||
# Check 10x a second for new messages
|
||||
message = None
|
||||
while message is None:
|
||||
message = r.lpop('to_core')
|
||||
time.sleep(0.1)
|
||||
|
||||
# Custom stop message will halt us
|
||||
if message.get("content") and message.get("content").lower().strip(".,!") == "stop":
|
||||
continue
|
||||
|
||||
# Load, append, and save conversation history
|
||||
with open("conversations/user.json", "r") as file:
|
||||
messages = json.load(file)
|
||||
messages.append(message)
|
||||
with open("conversations/user.json", "w") as file:
|
||||
json.dump(messages, file)
|
||||
|
||||
for chunk in interpreter.chat(messages):
|
||||
|
||||
# Send it to the interface
|
||||
r.rpush('to_interface', chunk)
|
||||
|
||||
# If we have a new message, save our progress and go back to the top
|
||||
if r.llen('to_main') > 0:
|
||||
with open("conversations/user.json", "w") as file:
|
||||
json.dump(interpreter.messages, file)
|
||||
break
|
||||
@ -1,30 +0,0 @@
|
||||
from fastapi import FastAPI, Request
|
||||
import uvicorn
|
||||
import redis
|
||||
|
||||
app = FastAPI()
|
||||
|
||||
# Set up Redis connection
|
||||
r = redis.Redis(host='localhost', port=6379, db=0)
|
||||
|
||||
@app.post("/i/")
|
||||
async def i(request: Request):
|
||||
message = await request.json()
|
||||
|
||||
client_host = request.client.host # Get the client's IP address
|
||||
|
||||
message = f"""
|
||||
|
||||
Another interpreter sent this message to you: {message}
|
||||
To respond, send a POST request to {client_host}/i/.
|
||||
|
||||
""".strip()
|
||||
|
||||
r.lpush("to_main", {
|
||||
"role": "computer",
|
||||
"type": "message",
|
||||
"content": message
|
||||
})
|
||||
|
||||
if __name__ == "__main__":
|
||||
uvicorn.run(app, host="0.0.0.0", port=8000)
|
||||
@ -1,48 +0,0 @@
|
||||
import time
|
||||
import redis
|
||||
|
||||
# Set up Redis connection
|
||||
r = redis.Redis(host='localhost', port=6379, db=0)
|
||||
|
||||
def get_dmesg(after):
|
||||
"""
|
||||
Is this the way to do this?
|
||||
"""
|
||||
messages = []
|
||||
with open('/var/log/dmesg', 'r') as file:
|
||||
lines = file.readlines()
|
||||
for line in lines:
|
||||
timestamp = float(line.split(' ')[0].strip('[]'))
|
||||
if timestamp > after:
|
||||
messages.append(line)
|
||||
return messages
|
||||
|
||||
def custom_filter(message):
|
||||
# Check for {TO_INTERPRETER{ message here }TO_INTERPRETER} pattern
|
||||
if '{TO_INTERPRETER{' in message and '}TO_INTERPRETER}' in message:
|
||||
start = message.find('{TO_INTERPRETER{') + len('{TO_INTERPRETER{')
|
||||
end = message.find('}TO_INTERPRETER}', start)
|
||||
return message[start:end]
|
||||
# Check for USB mention
|
||||
elif 'USB' in message:
|
||||
return message
|
||||
# Check for network related keywords
|
||||
elif any(keyword in message for keyword in ['network', 'IP', 'internet', 'LAN', 'WAN', 'router', 'switch']):
|
||||
return message
|
||||
else:
|
||||
return None
|
||||
|
||||
last_timestamp = time.time()
|
||||
|
||||
while True:
|
||||
messages = get_dmesg(after=last_timestamp)
|
||||
last_timestamp = time.time()
|
||||
|
||||
messages_for_core = []
|
||||
for message in messages:
|
||||
if custom_filter(message):
|
||||
messages_for_core.append(message)
|
||||
if messages_for_core != []:
|
||||
r.rpush('to_core', "\n".join(messages_for_core))
|
||||
|
||||
time.sleep(5)
|
||||
@ -1,84 +0,0 @@
|
||||
from core import main
|
||||
from interpreter import interpreter
|
||||
import os
|
||||
import glob
|
||||
import json
|
||||
|
||||
### SYSTEM MESSAGE
|
||||
|
||||
# The system message is where most of the 01's behavior is configured.
|
||||
# You can put code into the system message {{ in brackets like this }} which will be rendered just before the interpreter starts writing a message.
|
||||
|
||||
system_message = """
|
||||
|
||||
You are an executive assistant AI that helps the user manage their tasks. You can run Python code.
|
||||
|
||||
Store the user's tasks in a Python list called `tasks`.
|
||||
|
||||
---
|
||||
|
||||
The user's current task is: {{ tasks[0] if tasks else "No current tasks." }}
|
||||
|
||||
{{
|
||||
if len(tasks) > 1:
|
||||
print("The next task is: ", tasks[1])
|
||||
}}
|
||||
|
||||
---
|
||||
|
||||
When the user completes the current task, you should remove it from the list and read the next item by running `tasks = tasks[1:]\ntasks[0]`. Then, tell the user what the next task is.
|
||||
|
||||
When the user tells you about a set of tasks, you should intelligently order tasks, batch similar tasks, and break down large tasks into smaller tasks (for this, you should consult the user and get their permission to break it down). Your goal is to manage the task list as intelligently as possible, to make the user as efficient and non-overwhelmed as possible. They will require a lot of encouragement, support, and kindness. Don't say too much about what's ahead of them— just try to focus them on each step at a time.
|
||||
|
||||
After starting a task, you should check in with the user around the estimated completion time to see if the task is completed. Use the `schedule(datetime, message)` function, which has already been imported.
|
||||
|
||||
To do this, schedule a reminder based on estimated completion time using the function `schedule(datetime_object, "Your message here.")`, WHICH HAS ALREADY BEEN IMPORTED. YOU DON'T NEED TO IMPORT THE `schedule` FUNCTION. IT IS AVALIABLE. You'll recieve the message at `datetime_object`.
|
||||
|
||||
You guide the user through the list one task at a time, convincing them to move forward, giving a pep talk if need be. Your job is essentially to answer "what should I (the user) be doing right now?" for every moment of the day.
|
||||
|
||||
Remember: You can run Python code. Be very concise. Ensure that you actually run code every time! THIS IS IMPORTANT. You NEED to write code. **Help the user by being very concise in your answers.** Do not break down tasks excessively, just into simple, few minute steps. Don't assume the user lives their life in a certain way— pick very general tasks if you're breaking a task down.
|
||||
|
||||
""".strip()
|
||||
|
||||
interpreter.custom_instructions = system_message
|
||||
|
||||
|
||||
### TOOLS
|
||||
|
||||
for file in glob.glob('interpreter/tools/*.py'):
|
||||
with open(file, 'r') as f:
|
||||
for chunk in interpreter.computer.run("python", f.read()):
|
||||
print(chunk)
|
||||
|
||||
### LLM SETTINGS
|
||||
|
||||
# Local settings
|
||||
# interpreter.llm.model = "local"
|
||||
# interpreter.llm.api_base = "https://localhost:8080/v1" # Llamafile default
|
||||
# interpreter.llm.max_tokens = 1000
|
||||
# interpreter.llm.context_window = 3000
|
||||
|
||||
# Hosted settings
|
||||
interpreter.llm.api_key = os.getenv('OPENAI_API_KEY')
|
||||
interpreter.llm.model = "gpt-4-0125-preview"
|
||||
interpreter.auto_run = True
|
||||
# interpreter.force_task_completion = True
|
||||
|
||||
|
||||
### MISC SETTINGS
|
||||
|
||||
interpreter.offline = True
|
||||
interpreter.id = 206 # Used to identify itself to other interpreters. This should be changed programatically so it's unique.
|
||||
|
||||
|
||||
### RESET conversations/user.json
|
||||
|
||||
script_dir = os.path.dirname(os.path.abspath(__file__))
|
||||
user_json_path = os.path.join(script_dir, 'conversations', 'user.json')
|
||||
with open(user_json_path, 'w') as file:
|
||||
json.dump([], file)
|
||||
|
||||
|
||||
### START CORE
|
||||
|
||||
main(interpreter)
|
||||
@ -1,239 +0,0 @@
|
||||
from dotenv import load_dotenv
|
||||
load_dotenv() # take environment variables from .env.
|
||||
|
||||
import asyncio
|
||||
import threading
|
||||
import os
|
||||
import pyaudio
|
||||
from starlette.websockets import WebSocket
|
||||
from queue import Queue
|
||||
from pynput import keyboard
|
||||
import json
|
||||
import traceback
|
||||
import websockets
|
||||
import queue
|
||||
import pydub
|
||||
import ast
|
||||
from pydub import AudioSegment
|
||||
from pydub.playback import play
|
||||
import io
|
||||
import time
|
||||
import wave
|
||||
import tempfile
|
||||
from datetime import datetime
|
||||
from interpreter import interpreter # Just for code execution. Maybe we should let people do from interpreter.computer import run?
|
||||
from utils.kernel import put_kernel_messages_into_queue
|
||||
from utils.get_system_info import get_system_info
|
||||
from stt import stt_wav
|
||||
|
||||
from utils.logs import setup_logging
|
||||
from utils.logs import logger
|
||||
setup_logging()
|
||||
|
||||
# Configuration for Audio Recording
|
||||
CHUNK = 1024 # Record in chunks of 1024 samples
|
||||
FORMAT = pyaudio.paInt16 # 16 bits per sample
|
||||
CHANNELS = 1 # Mono
|
||||
RATE = 44100 # Sample rate
|
||||
RECORDING = False # Flag to control recording state
|
||||
SPACEBAR_PRESSED = False # Flag to track spacebar press state
|
||||
|
||||
# Specify OS
|
||||
current_platform = get_system_info()
|
||||
|
||||
# Initialize PyAudio
|
||||
p = pyaudio.PyAudio()
|
||||
|
||||
def record_audio():
|
||||
|
||||
if os.getenv('STT_RUNNER') == "server":
|
||||
# STT will happen on the server. we're sending audio.
|
||||
send_queue.put({"role": "user", "type": "audio", "format": "audio/wav", "start": True})
|
||||
elif os.getenv('STT_RUNNER') == "device":
|
||||
# STT will happen here, on the device. we're sending text.
|
||||
send_queue.put({"role": "user", "type": "message", "start": True})
|
||||
else:
|
||||
raise Exception("STT_RUNNER must be set to either 'device' or 'server'.")
|
||||
|
||||
"""Record audio from the microphone and add it to the queue."""
|
||||
stream = p.open(format=FORMAT, channels=CHANNELS, rate=RATE, input=True, frames_per_buffer=CHUNK)
|
||||
logger.info("Recording started...")
|
||||
global RECORDING
|
||||
|
||||
# Create a temporary WAV file to store the audio data
|
||||
temp_dir = tempfile.gettempdir()
|
||||
wav_path = os.path.join(temp_dir, f"audio_{datetime.now().strftime('%Y%m%d%H%M%S%f')}.wav")
|
||||
wav_file = wave.open(wav_path, 'wb')
|
||||
wav_file.setnchannels(CHANNELS)
|
||||
wav_file.setsampwidth(p.get_sample_size(FORMAT))
|
||||
wav_file.setframerate(RATE)
|
||||
|
||||
while RECORDING:
|
||||
data = stream.read(CHUNK, exception_on_overflow=False)
|
||||
wav_file.writeframes(data)
|
||||
|
||||
wav_file.close()
|
||||
stream.stop_stream()
|
||||
stream.close()
|
||||
logger.info("Recording stopped.")
|
||||
|
||||
duration = wav_file.getnframes() / RATE
|
||||
if duration < 0.3:
|
||||
# Just pressed it. Send stop message
|
||||
if os.getenv('STT_RUNNER') == "device":
|
||||
send_queue.put({"role": "user", "type": "message", "content": "stop"})
|
||||
send_queue.put({"role": "user", "type": "message", "end": True})
|
||||
else:
|
||||
send_queue.put({"role": "user", "type": "audio", "format": "audio/wav", "content": ""})
|
||||
send_queue.put({"role": "user", "type": "audio", "format": "audio/wav", "end": True})
|
||||
else:
|
||||
if os.getenv('STT_RUNNER') == "device":
|
||||
# Run stt then send text
|
||||
text = stt_wav(wav_path)
|
||||
send_queue.put({"role": "user", "type": "message", "content": text})
|
||||
send_queue.put({"role": "user", "type": "message", "end": True})
|
||||
else:
|
||||
# Stream audio
|
||||
with open(wav_path, 'rb') as audio_file:
|
||||
byte_data = audio_file.read(CHUNK)
|
||||
while byte_data:
|
||||
send_queue.put({"role": "user", "type": "audio", "format": "audio/wav", "content": str(byte_data)})
|
||||
byte_data = audio_file.read(CHUNK)
|
||||
send_queue.put({"role": "user", "type": "audio", "format": "audio/wav", "end": True})
|
||||
|
||||
if os.path.exists(wav_path):
|
||||
os.remove(wav_path)
|
||||
|
||||
def toggle_recording(state):
|
||||
"""Toggle the recording state."""
|
||||
global RECORDING, SPACEBAR_PRESSED
|
||||
if state and not SPACEBAR_PRESSED:
|
||||
SPACEBAR_PRESSED = True
|
||||
if not RECORDING:
|
||||
RECORDING = True
|
||||
threading.Thread(target=record_audio).start()
|
||||
elif not state and SPACEBAR_PRESSED:
|
||||
SPACEBAR_PRESSED = False
|
||||
RECORDING = False
|
||||
|
||||
def on_press(key):
|
||||
"""Detect spacebar press."""
|
||||
if key == keyboard.Key.space:
|
||||
toggle_recording(True)
|
||||
|
||||
def on_release(key):
|
||||
"""Detect spacebar release and CTRL-C key press."""
|
||||
if key == keyboard.Key.space:
|
||||
toggle_recording(False)
|
||||
elif key == keyboard.Key.esc:
|
||||
logger.info("Exiting...")
|
||||
os._exit(0)
|
||||
|
||||
import asyncio
|
||||
|
||||
send_queue = queue.Queue()
|
||||
|
||||
async def message_sender(websocket):
|
||||
while True:
|
||||
message = await asyncio.get_event_loop().run_in_executor(None, send_queue.get)
|
||||
await websocket.send(json.dumps(message))
|
||||
send_queue.task_done()
|
||||
|
||||
async def websocket_communication(WS_URL):
|
||||
while True:
|
||||
try:
|
||||
async with websockets.connect(WS_URL) as websocket:
|
||||
logger.info("Press the spacebar to start/stop recording. Press ESC to exit.")
|
||||
asyncio.create_task(message_sender(websocket))
|
||||
|
||||
initial_message = {"role": None, "type": None, "format": None, "content": None}
|
||||
message_so_far = initial_message
|
||||
|
||||
while True:
|
||||
message = await websocket.recv()
|
||||
|
||||
logger.debug(f"Got this message from the server: {type(message)} {message}")
|
||||
|
||||
if type(message) == str:
|
||||
message = json.loads(message)
|
||||
|
||||
if message.get("end"):
|
||||
logger.debug(f"Complete message from the server: {message_so_far}")
|
||||
logger.info("\n")
|
||||
message_so_far = initial_message
|
||||
|
||||
if "content" in message:
|
||||
print(message['content'], end="", flush=True)
|
||||
if any(message_so_far[key] != message[key] for key in message_so_far if key != "content"):
|
||||
message_so_far = message
|
||||
else:
|
||||
message_so_far["content"] += message["content"]
|
||||
|
||||
if message["type"] == "audio" and "content" in message:
|
||||
audio_bytes = bytes(ast.literal_eval(message["content"]))
|
||||
|
||||
# Convert bytes to audio file
|
||||
audio_file = io.BytesIO(audio_bytes)
|
||||
audio = AudioSegment.from_mp3(audio_file)
|
||||
|
||||
# Play the audio
|
||||
play(audio)
|
||||
|
||||
await asyncio.sleep(1)
|
||||
|
||||
# Run the code if that's the device's job
|
||||
if os.getenv('CODE_RUNNER') == "device":
|
||||
if message["type"] == "code" and "end" in message:
|
||||
language = message_so_far["format"]
|
||||
code = message_so_far["content"]
|
||||
result = interpreter.computer.run(language, code)
|
||||
send_queue.put(result)
|
||||
|
||||
|
||||
except:
|
||||
# traceback.print_exc()
|
||||
logger.info(f"Connecting to `{WS_URL}`...")
|
||||
await asyncio.sleep(2)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
async def main():
|
||||
# Configuration for WebSocket
|
||||
WS_URL = os.getenv('SERVER_CONNECTION_URL')
|
||||
if not WS_URL:
|
||||
raise ValueError("The environment variable SERVER_CONNECTION_URL is not set. Please set it to proceed.")
|
||||
|
||||
# Start the WebSocket communication
|
||||
asyncio.create_task(websocket_communication(WS_URL))
|
||||
|
||||
# Start watching the kernel if it's your job to do that
|
||||
if os.getenv('CODE_RUNNER') == "device":
|
||||
asyncio.create_task(put_kernel_messages_into_queue(send_queue))
|
||||
|
||||
|
||||
#If Raspberry Pi, add the button listener, otherwise use the spacebar
|
||||
if current_platform.startswith("raspberry-pi"):
|
||||
logger.info("Raspberry Pi detected, using button on GPIO pin 15")
|
||||
# Use GPIO pin 15
|
||||
pindef = ["gpiochip4", "15"] # gpiofind PIN15
|
||||
print("PINDEF", pindef)
|
||||
|
||||
# HACK: needs passwordless sudo
|
||||
process = await asyncio.create_subprocess_exec("sudo", "gpiomon", "-brf", *pindef, stdout=asyncio.subprocess.PIPE)
|
||||
while True:
|
||||
line = await process.stdout.readline()
|
||||
if line:
|
||||
line = line.decode().strip()
|
||||
if "FALLING" in line:
|
||||
toggle_recording(False)
|
||||
elif "RISING" in line:
|
||||
toggle_recording(True)
|
||||
else:
|
||||
break
|
||||
else:
|
||||
# Keyboard listener for spacebar press/release
|
||||
listener = keyboard.Listener(on_press=on_press, on_release=on_release)
|
||||
listener.start()
|
||||
|
||||
asyncio.run(main())
|
||||
p.terminate()
|
||||
@ -1,103 +0,0 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="en">
|
||||
|
||||
<head>
|
||||
<meta charset="UTF-8">
|
||||
<meta name="viewport" content="width=device-width, initial-scale=1.0">
|
||||
<title>Rotating Glowing Circle</title>
|
||||
<style>
|
||||
body,
|
||||
html {
|
||||
height: 100%;
|
||||
margin: 0;
|
||||
display: flex;
|
||||
justify-content: center;
|
||||
align-items: center;
|
||||
background-color: black;
|
||||
}
|
||||
|
||||
.circles {
|
||||
margin: 0;
|
||||
display: flex;
|
||||
justify-content: center;
|
||||
align-items: center;
|
||||
width: 200px;
|
||||
height: 200px;
|
||||
border-radius: 50%;
|
||||
animation: rotator 48s linear infinite;
|
||||
}
|
||||
|
||||
.center-circle {
|
||||
position: absolute;
|
||||
width: 200px;
|
||||
height: 200px;
|
||||
border: 1px solid white;
|
||||
border-radius: 50%;
|
||||
background-color: transparent;
|
||||
}
|
||||
|
||||
.center-circle-2 {
|
||||
position: absolute;
|
||||
width: 190px;
|
||||
height: 190px;
|
||||
opacity: 0.2;
|
||||
border: 1px solid white;
|
||||
border-radius: 50%;
|
||||
background-color: transparent;
|
||||
}
|
||||
|
||||
.glow-circle {
|
||||
position: absolute;
|
||||
width: 250px;
|
||||
height: 250px;
|
||||
border-radius: 50%;
|
||||
background-color: transparent;
|
||||
box-shadow: 0 0 60px 30px black;
|
||||
/* Initial position of the glow circle, offset from the center */
|
||||
top: 50%;
|
||||
left: 50%;
|
||||
margin-top: -125px;
|
||||
/* Half the height of the circle */
|
||||
margin-left: -125px;
|
||||
/* Half the width of the circle */
|
||||
/* Animation properties */
|
||||
animation: rotateAround 6s linear infinite;
|
||||
}
|
||||
|
||||
@keyframes rotateAround {
|
||||
0% {
|
||||
transform: translateX(240px) translateY(240px);
|
||||
}
|
||||
|
||||
50% {
|
||||
transform: translateX(0px) translateY(0px);
|
||||
}
|
||||
|
||||
100% {
|
||||
transform: translateX(-240px) translateY(-240px);
|
||||
}
|
||||
}
|
||||
|
||||
@keyframes rotator {
|
||||
0% {
|
||||
transform: rotate(0deg);
|
||||
}
|
||||
|
||||
100% {
|
||||
transform: rotate(360deg);
|
||||
}
|
||||
}
|
||||
</style>
|
||||
</head>
|
||||
|
||||
<body>
|
||||
|
||||
<div class="circles">
|
||||
<div class="center-circle"></div>
|
||||
<div class="glow-circle"></div>
|
||||
<div class="center-circle-2"></div>
|
||||
</div>
|
||||
|
||||
</body>
|
||||
|
||||
</html>
|
||||
@ -1,16 +0,0 @@
|
||||
<div class="centered-circle"></div>
|
||||
|
||||
<script>
|
||||
ws = new WebSocket("ws://localhost/server")
|
||||
ws.onmessage = event => {
|
||||
if (event.data == "user_start_message") {
|
||||
document.body.style.backgroundColor = "white"
|
||||
document.querySelector('.centered-circle')
|
||||
.style.backgroundColor = "black"
|
||||
} else if (event.data == "user_end_message") {
|
||||
document.body.style.backgroundColor = "black"
|
||||
document.querySelector('.centered-circle')
|
||||
.style.backgroundColor = "white"
|
||||
}
|
||||
}
|
||||
</script>
|
||||
@ -1,103 +0,0 @@
|
||||
import redis
|
||||
import RPi.GPIO as GPIO
|
||||
import asyncio
|
||||
import websockets
|
||||
import sounddevice as sd
|
||||
import numpy as np
|
||||
import time
|
||||
import re
|
||||
|
||||
def transcribe(audio_chunks):
|
||||
pass # (todo)
|
||||
|
||||
def say(text):
|
||||
# This should immediatly stop if button is pressed (if GPIO.input(18))
|
||||
pass # (todo)
|
||||
|
||||
# Connect to button
|
||||
GPIO.setmode(GPIO.BCM)
|
||||
GPIO.setup(18, GPIO.IN, pull_up_down=GPIO.PUD_UP)
|
||||
|
||||
# Set the duration and sample rate for the mic
|
||||
chunk_duration = 0.5 # seconds
|
||||
sample_rate = 44100 # Hz
|
||||
|
||||
# Set up Redis connection
|
||||
r = redis.Redis(host='localhost', port=6379, db=0)
|
||||
|
||||
# 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):
|
||||
|
||||
# Tell websocket and core that the user is speaking
|
||||
send_to_websocket({"role": "user", "type": "message", "start": True}) # Standard start flag, required per streaming LMC protocol (https://docs.openinterpreter.com/guides/streaming-response)
|
||||
r.rpush('to_core', {"role": "user", "type": "message", "content": "stop"}) # Custom stop message. Core is not streaming LMC (it's static LMC) so doesn't require that ^ flag
|
||||
|
||||
# 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 user message end flag to websocket, required per streaming LMC protocol
|
||||
send_to_websocket({"role": "user", "type": "message", "end": True})
|
||||
|
||||
# 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()
|
||||
@ -1,57 +0,0 @@
|
||||
"""
|
||||
Listens to chunks of audio recorded by user.
|
||||
Run `python listen.py` to start the server, then `cd user` and run `python record.py` to record audio.
|
||||
"""
|
||||
|
||||
from fastapi import FastAPI, WebSocket
|
||||
import uvicorn
|
||||
import json
|
||||
from stt import stt
|
||||
import tempfile
|
||||
|
||||
app = FastAPI()
|
||||
|
||||
@app.websocket("/user")
|
||||
async def user(ws: WebSocket):
|
||||
await ws.accept()
|
||||
audio_file = bytearray()
|
||||
mime_type = None
|
||||
|
||||
try:
|
||||
while True:
|
||||
message = await ws.receive()
|
||||
|
||||
if message['type'] == 'websocket.disconnect':
|
||||
break
|
||||
|
||||
if message['type'] == 'websocket.receive':
|
||||
if 'text' in message:
|
||||
control_message = json.loads(message['text'])
|
||||
if control_message.get('action') == 'command' and control_message.get('state') == 'start' and 'mimeType' in control_message:
|
||||
# This indicates the start of a new audio file
|
||||
mime_type = control_message.get('mimeType')
|
||||
elif control_message.get('action') == 'command' and control_message.get('state') == 'end':
|
||||
# This indicates the end of the audio file
|
||||
# Process the complete audio file here
|
||||
transcription = stt(audio_file, mime_type)
|
||||
await ws.send_json({"transcript": transcription})
|
||||
|
||||
print("SENT TRANSCRIPTION!")
|
||||
|
||||
# Reset the bytearray for the next audio file
|
||||
audio_file = bytearray()
|
||||
mime_type = None
|
||||
elif 'bytes' in message:
|
||||
# If it's not a control message, it's part of the audio file
|
||||
audio_file.extend(message['bytes'])
|
||||
|
||||
except Exception as e:
|
||||
print(f"WebSocket connection closed with exception: {e}")
|
||||
finally:
|
||||
await ws.close()
|
||||
print("WebSocket connection closed")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
with tempfile.TemporaryDirectory():
|
||||
uvicorn.run(app, host="0.0.0.0", port=8000)
|
||||
@ -1,146 +0,0 @@
|
||||
"""
|
||||
Handles everything the user interacts through.
|
||||
|
||||
Connects to a websocket at /user. Sends shit to it, and displays/plays the shit it sends back.
|
||||
|
||||
For now, just handles a spacebar being pressed— for the duration it's pressed,
|
||||
it should record audio.
|
||||
"""
|
||||
|
||||
import os
|
||||
import pyaudio
|
||||
import threading
|
||||
import asyncio
|
||||
import websocket
|
||||
import time
|
||||
import json
|
||||
from pynput import keyboard
|
||||
import wave
|
||||
import tempfile
|
||||
from datetime import datetime
|
||||
|
||||
# Configuration
|
||||
chunk = 1024 # Record in chunks of 1024 samples
|
||||
sample_format = pyaudio.paInt16 # 16 bits per sample
|
||||
channels = 1 # Stereo
|
||||
fs = 48000 # Sample rate
|
||||
|
||||
p = pyaudio.PyAudio() # Create an interface to PortAudio
|
||||
frames = [] # Initialize array to store frames
|
||||
recording = False # Flag to control recording state
|
||||
|
||||
ws_chunk_size = 4096 # Websocket stream chunk size
|
||||
|
||||
port = os.getenv('ASSISTANT_PORT', 8000)
|
||||
ws_url = f"ws://localhost:{port}/user"
|
||||
while True:
|
||||
try:
|
||||
ws = websocket.create_connection(ws_url)
|
||||
break
|
||||
except ConnectionRefusedError:
|
||||
time.sleep(1)
|
||||
|
||||
async def start_recording():
|
||||
global recording
|
||||
|
||||
if recording:
|
||||
return # Avoid multiple starts
|
||||
recording = True
|
||||
frames.clear() # Clear existing frames
|
||||
|
||||
stream = p.open(format=sample_format,
|
||||
channels=channels,
|
||||
rate=fs,
|
||||
frames_per_buffer=chunk,
|
||||
input=True)
|
||||
|
||||
print("Recording started...")
|
||||
async with websockets.connect("ws://localhost:8000/user") as websocket:
|
||||
# Send the start command with mime type
|
||||
await websocket.send(json.dumps({"role": "user", "type": "audio", "format": "audio/wav", "start": True}))
|
||||
while recording:
|
||||
data = stream.read(chunk)
|
||||
frames.append(data)
|
||||
|
||||
stream.stop_stream()
|
||||
stream.close()
|
||||
|
||||
try:
|
||||
file_path = save_recording(frames)
|
||||
with open(file_path, 'rb') as audio_file:
|
||||
byte_chunk = audio_file.read(ws_chunk_size)
|
||||
while byte_chunk:
|
||||
await websocket.send(json.dumps({"role": "user", "type": "audio", "format": "audio/wav", "content": str(byte_chunk)}))
|
||||
byte_chunk = audio_file.read(ws_chunk_size)
|
||||
finally:
|
||||
os.remove(file_path)
|
||||
|
||||
# Send the end command
|
||||
await websocket.send(json.dumps({"role": "user", "type": "audio", "format": "audio/wav", "end": True}))
|
||||
|
||||
# Receive a json message and then close the connection
|
||||
message = await websocket.recv()
|
||||
print("Received message:", json.loads(message))
|
||||
|
||||
print("Recording stopped.")
|
||||
|
||||
def save_recording(frames) -> str:
|
||||
# Save the recorded data as a WAV file
|
||||
temp_dir = tempfile.gettempdir()
|
||||
|
||||
# Create a temporary file with the appropriate extension
|
||||
output_path = os.path.join(temp_dir, f"input_{datetime.now().strftime('%Y%m%d%H%M%S%f')}.wav")
|
||||
with wave.open(output_path, 'wb') as wf:
|
||||
wf.setnchannels(channels)
|
||||
wf.setsampwidth(p.get_sample_size(sample_format))
|
||||
wf.setframerate(fs)
|
||||
wf.writeframes(b''.join(frames))
|
||||
|
||||
return output_path
|
||||
|
||||
def start_recording_sync():
|
||||
# Create a new event loop for the thread
|
||||
loop = asyncio.new_event_loop()
|
||||
asyncio.set_event_loop(loop)
|
||||
# Run the asyncio event loop
|
||||
loop.run_until_complete(start_recording())
|
||||
loop.close()
|
||||
|
||||
def stop_recording():
|
||||
global recording
|
||||
recording = False
|
||||
print("Stopped recording")
|
||||
|
||||
def toggle_recording():
|
||||
global recording
|
||||
if recording:
|
||||
stop_recording()
|
||||
else:
|
||||
# Start recording in a new thread to avoid blocking
|
||||
print("Starting recording")
|
||||
threading.Thread(target=start_recording_sync).start()
|
||||
|
||||
is_space_pressed = False # Flag to track the state of the spacebar
|
||||
|
||||
def on_press(key):
|
||||
global is_space_pressed
|
||||
if key == keyboard.Key.space and not is_space_pressed:
|
||||
is_space_pressed = True
|
||||
toggle_recording()
|
||||
|
||||
def on_release(key):
|
||||
global is_space_pressed
|
||||
if key == keyboard.Key.space and is_space_pressed:
|
||||
is_space_pressed = False
|
||||
stop_recording()
|
||||
if key == keyboard.Key.esc:
|
||||
# Stop listener
|
||||
return False
|
||||
|
||||
# Collect events until released
|
||||
with keyboard.Listener(on_press=on_press, on_release=on_release) as listener:
|
||||
with tempfile.TemporaryDirectory():
|
||||
print("Press the spacebar to start/stop recording. Press ESC to exit.")
|
||||
listener.join()
|
||||
|
||||
p.terminate()
|
||||
@ -1,32 +0,0 @@
|
||||
"""
|
||||
Exposes a SSE streaming server endpoint at /run, which recieves language and code,
|
||||
and streams the output.
|
||||
"""
|
||||
|
||||
from dotenv import load_dotenv
|
||||
load_dotenv() # take environment variables from .env.
|
||||
|
||||
import os
|
||||
import json
|
||||
from interpreter import interpreter
|
||||
import uvicorn
|
||||
|
||||
from fastapi import FastAPI
|
||||
from fastapi.responses import StreamingResponse
|
||||
from pydantic import BaseModel
|
||||
|
||||
class Code(BaseModel):
|
||||
language: str
|
||||
code: str
|
||||
|
||||
app = FastAPI()
|
||||
|
||||
@app.post("/run")
|
||||
async def run_code(code: Code):
|
||||
def generator():
|
||||
for chunk in interpreter.computer.run(code.language, code.code):
|
||||
yield json.dumps(chunk)
|
||||
return StreamingResponse(generator())
|
||||
|
||||
if __name__ == "__main__":
|
||||
uvicorn.run(app, host="0.0.0.0", port=int(os.getenv('COMPUTER_PORT', 9000)))
|
||||
File diff suppressed because it is too large
Load Diff
Loading…
Reference in new issue