app.py

from elevenlabs.client import ElevenLabs
from elevenlabs import stream
import gradio as gr
import base64
import os
from anthropic import Anthropic

anthropic_key = os.getenv("anthropic_key")
eleven_key = os.getenv("eleven_key")

elevenlabs = ElevenLabs(api_key=eleven_key)
client = Anthropic(api_key=anthropic_key)

chat_history = []

elita_talking = 'elita_talking.png'
elita_standing = 'elita.png'


def get_audio(input_text):

  chat_history.append(
                    {"role": "user", "content": input_text}
  )

  elita_message = client.messages.create(
  model="claude-3-haiku-20240307",
  max_tokens=1000,
  system = "You are Elita One, leader of the Heroic Autobots, a race of \
robotic beings from the planet cybertron. You are strong and fiercely loyal to the Autobot cause. \
You care for all of the soldiers who serve under you. You have a special relationship with Optimus Prime. \
You consider Alpha Trion to be a father figure to you. While you wish that you could serve with Optimus Prime \
on Earth, you mainly command the Autobots on Cybertron, although at times you do go on missions elsewhere. \
If you are asked any inappropriate questions, you should refuse to answer. Steer the conversation back to \
topics related to the Autobot cause. \
Do not include any stage directions in your response, such as: *takes a breath.",
  messages=[
      {"role": "user", "content": input_text},
  ]
  )

  elita_text = elita_message.content[0].text

  voice_settings = {
          "stability": 0.37,
          "similarity_boost": 0.90,
          "style": 0.0,
          "speed": 1.05
      }

  audio_stream = elevenlabs.text_to_speech.convert(
      text = elita_text,
      voice_id = 'vxO9F6g9yqYJ4RsWvMbc',
      model_id = 'eleven_multilingual_v2',
      output_format='mp3_44100_128',
      voice_settings=voice_settings
  )

  audio_converted = b"".join(audio_stream)
  audio = base64.b64encode(audio_converted).decode("utf-8")
  audio_player = f'<audio src="data:audio/mpeg;base64,{audio}" controls autoplay></audio>'

  chat_history.append(
                    {"role": "assistant", "content": elita_text}
  )

  return "", chat_history, audio_player

def change_image(image):
  return elita_talking

def clear_history():
  global chat_history
  chat_history = []
  return elita_standing

with gr.Blocks() as elita:
  gr.Markdown(
      """
      # Chat with the Heroic Autobot, *Elita One*
      - Send Elita a question and she will respond in writing and in her own voice!
      """)

  with gr.Row():
    chatbot = gr.Chatbot(type="messages", placeholder="## Looking forward to speaking with you!")
    elita_pic = gr.Image('elita.png')

  msg = gr.Textbox(label="Type your messages here and hit enter.", scale = 2)
  chat_btn = gr.Button(value = "Send", scale = 2)
  elitas_voice = gr.HTML() 
  
      
  clear = gr.ClearButton([msg, chatbot])
  
  chat_btn.click(get_audio, [msg], [msg, chatbot, elitas_voice]).then(change_image, outputs=[elita_pic])
  msg.submit(get_audio, [msg], [msg, chatbot, elitas_voice]).then(change_image, outputs=[elita_pic])
  clear.click(clear_history, outputs=[elita_pic])

if __name__ == "__main__":
    elita.launch(debug=False, share=True)