voice assistant with computer-use

soulseek.py

#! /usr/bin/env -S uv run
# /// script
# requires-python = "==3.11.*"
# dependencies = ["RealtimeSTT==0.3.92", "open-interpreter"]
# ///

"""
# quickstart
$ DEEPSEEK_API_KEY=(pass deepseek_api) uv run soulseek.py
"""
import os

from openai import OpenAI
from RealtimeSTT import AudioToTextRecorder
from interpreter import interpreter


interpreter.llm.api_key = os.environ.get("DEEPSEEK_API_KEY")
interpreter.llm.model = "deepseek/deepseek-chat"
interpreter.auto_run = True
interpreter.safe_mode = 'off'
interpreter.anonymized_telemetry = False


def process_text(text):
    print(text)
    interpreter.chat(text)
 

if __name__ == '__main__':
    print("Wait until it says 'speak now'")
    # https://github.com/KoljaB/RealtimeSTT/blob/master/RealtimeSTT/audio_recorder.py#L301
    with AudioToTextRecorder(
        model="large-v3-turbo",
        wake_words="alexa",
        device="mps",
        language = "zh",
        initial_prompt = "以下是普通话的句子。",
        wakeword_backend="oww",
        wake_words_sensitivity=0.35,
        wake_word_buffer_duration=1,
        ) as recorder:
        recorder.text(process_text)