Voice-to-text Chatbot with Streaming

This project builds a chatbot that takes text or voice inputs. It uses OpenAI Whisper for transcribing audio, and OpenAI gpt-4o-mini for providing LLM responses in the chat. It uses Django REST Framework as the backend API, Streamlit for the UI, and the audio_recorder_streamlit plugin for voice recording.

Prerequisites

• Python 3.11 or higher
• Poetry (package manager)
• Git

Project structure

The vt_demo folder is the root folder for the Django project, the chatbot_app is a Django app folder containing the core AI logic and related views, and the ui folder contains the Streamlit frontend.

Getting started

1. Clone the repository

   git clone <repository-url>
   cd <project-directory>

2. Install Poetry (if not already installed)

   curl -sSL https://install.python-poetry.org | python3 -

3. Install dependencies

   poetry install

Configuration

1. Set up environment variables: Create a .env file, add your OpenAI API key, and Django dev secret key.

   OPENAI_API_KEY="sk-..."
   DJANGO_SECRET_KEY="django-..."
   

2. Database Setup

   poetry run python manage.py migrate

Running the Application

1. Split terminal: You can then run the backend in one, and the frontend in another.

2. Start the development backend server in one terminal

   poetry run python manage.py runserver

   The application will be available at http://localhost:8000

3. Start the UI in the other terminal

   poetry run streamlit run ui/streamlit_app.py

   Navigate to http://localhost:8501/admin or whatever the link in the terminal suggests for the UI.

Common Development Tasks

• Create new migrations

  poetry run python manage.py makemigrations

• Run tests

  poetry run python manage.py test

• Install new dependencies

  poetry add package-name

To Do

• Tests need updating mocking OpenAI is broken from latest changes.