Creating a Chatbot with Deepgram and Gro

Introduction

By the end of this tutorial, you will learn how to create a chatbot that can convert text to speech and respond to user queries using Deepgram and Gro API services. The chatbot will sound like a human, and you can integrate it with your website to provide a conversational interface for users.

Services Used

• Deepgram: A text-to-speech and speech-to-text service that is more accurate than AWS, Microsoft Azure, Whisper, Google, and Assembly AI.
• Gro API: An inferencing service for getting results quickly.

Prerequisites

• Python: Download the latest version from python.org.
• Visual Studio Code Editor: Download and install this code editor.
• API Keys: Obtain API keys from Deepgram and Gro API.

Setting Up the Environment

• Open the terminal in Visual Studio Code and create a new environment using conda create --name envs.
• Activate the environment using conda activate envs.
• Install the required libraries by running pip install deepgram-sdk requests python-dotenv flask Croc.

Configuring Environment Variables

• Create a requirements.txt file and install the required libraries.
• Set up environment variables for Deepgram API key and Gro API key.
• Paste the API keys into the environment variables.

Code Cleanup

• Start a new Python file and import the required libraries.
• Set up the Deepgram and Gro API services using the API keys.
• Create a function to handle user input and respond with a result.

Setting up the Environment and Creating Python Files

• Create three Python files: text_to_speech.py, speech_to_text.py, and Gro_server.py.
• Load up the environment using conda activate and Jarvis new.
• Install all requirements using pip install -r requirements.txt.

Gro Server Code

• Gro_server.py code:
  • Import necessary libraries: from gro import Gro.
  • Set up environment variables using from EnV import load.
  • Load environment variables using load.
  • Create a new instance of the Gro class.
  • Get the Gro API key.
  • Define a new function execute that takes in the model, role, user, and prompt as inputs.
  • Return the responses obtained from the Gro API.
  • Test the function by calling execute with the prompt "tell me a joke".

Text to Speech Code

• text_to_speech.py code:
  • Import necessary libraries: from deepgram import clients.
  • Load environment variables using from EnV import load.
  • Define a function text_to_speech that takes in the text input.
  • Start up the Deepgram API client using the DG API key.
  • Use the voice "Ora Luna" (or other available voices) to convert the text to speech.

Connecting the Text to Speech and Gro Server Code

• Import the execute function from Gro_server.py in the main app.py file.
• Use the execute function to call the Gro server and obtain the response.
• Use the response obtained from the Gro server as input to the text to speech function.
• Call the text_to_speech function with the response as input to convert the text to speech.

Introduction to Text-to-Speech (TTS) and Speech-to-Text (STT) models

• Multiple female voices are available, including Aria, Luna, Stella, and Atlanta.
• Male voices are also available, such as Henry.
• Code snippets show how to use TTS and STT models in Python.

Using TTS models

• Import the text2speech library and select a voice option.
• Define a function to return a wave file with the TTS output.
• Use the name parameter to determine whether to run the code locally or remotely.

Using STT models

• Import the speech2text library and load environment variables.
• Set up API keys and open an audio input file.
• Convert the audio file to text and retrieve the text output.

Creating a Main App with Flask

• Import the Flask library and the request and send_file functions.
• Create an app instance using Flask.
• Define routes for the app, including a root route that returns an index.html template.
• Define a route for the speech-to-text functionality.

Conclusion

By the end of this tutorial, you will have created a chatbot that can convert text to speech and respond to user queries using Deepgram and Gro API services. This chatbot can be integrated with your website to provide a conversational interface for users.