ChatGPT Streamlit

This project for making a demo based on ChatGPT and speech interfaces via streamlit.

• Docs of ChatGPT and Whisper API : OpenAI Documentation
• Workflow of demo
  1. Recording audio
  2. STT (Speech-to-text) with Whisper API
  3. Get an answer from ChatGPT API
  4. (optional) Making an audio from ChatGPT's answer

Usage

• Environment

  • python 3.10
  • GPU : cuda 11.6, RTX Titan
  • OS : ubuntu20.04 - wsl2
  • IDE : VSCode

• OpenAI API

  • You must get a secrey key of OpenAI platform from This Page.

• Install and run

  # with virtualenv
  $ virtualenv .venv
  $ . .venv/bin/activate
  
  # openai - chatgpt, whisper only
  $ pip install .
  
  # openai with simple TTS (with cpu version of pytorch)
  $ pip install .[tts] --extra-index-url https://download.pytorch.org/whl/cpu
  
  # run streamlit - with environment variable of API KEY
  $ OPENAI_API_KEY=[YOUR-API-KEY] streamlit run chatgpt_streamlit/app.py 

Others

• Demo Page (all tasks of workflow are done)