Personal Assistant Chat Application Using OpenAI's GPT-3

Are you looking for a personal assistant that can understand and respond to your natural language commands and queries? Look no further! This project uses OpenAI's GPT-3 model to create a personal assistant that can interact with you via speech or text.

The project includes both Speech-to-Text and Text-to-Speech capabilities, allowing you to interact with the assistant in a natural, conversational way. The user interface is similar to a chat application, with typing animations and the ability to replay messages.

Getting Started

Before you can start using the personal assistant, there are a few requirements and prerequisites that you'll need to take care of:

1. Obtain an API key from OpenAI.
2. Update the openai.key value in the worker.py file
3. Obtain an Entitlment Key from IBM.
4. Ensure Docker is installed in your system.

Once you have your keys and the models deployed, you can use the application by following the steps in the Using the Pesonal Assistant section.

Using the Personal Assistant

To run the application in a Docker container, follow these steps:

1. Open the terminal and navigate to the root directory of the repository.
2. Export your IBM Entitlement Key and login to a specific Docker registry:

IBM_ENTITLEMENT_KEY=... # replace ... with your key
echo $IBM_ENTITLEMENT_KEY | docker login -u cp --password-stdin cp.icr.io

3. Run the following command:

docker compose up --build

Check the logs to make sure that both the TTS and STT containers are running correctly by ensuring the logs display "INFO: Chuck server ready." as shown below:

...
chatapp-with-voice-and-openai-stt-1      | "INFO: Chuck server ready."
...
chatapp-with-voice-and-openai-tts-1      | "INFO: Chuck server ready."
...

To stop the application, press Ctrl + C in the terminal where the docker compose up command was run. To tear down the containers and related resources, run:

docker compose down

For the smoothest experience, run docker compose down and docker compose up --build each time you want to test new changes to ensure the image is rebuilt with the latest changes.

4. Open your browser and navigate to http://localhost:8000 to interact with the Personal Assistant. You can then speak your queries by clicking on the microphone or typing into the message bar. The assistant will then respond with a message and read the message outloud.

Note: Any changes made to the files will require rebuilding and rerunning the service.

Note

• Make sure to keep your API keys secure and not to share them with anyone.
• Make sure the URLs are working and are correctly used in the code
• Make sure that the TTS and STT are ready to use (by checking their logs)

Troubleshooting

If you experince an error lke:

requests.exceptions.ConnectionError: ('Connection aborted.', RemoteDisconnected('Remote end closed connection without response'))

This may indicate that the TTS or STT containers are not ready, please ensure the logs have each logged "INFO: Chuck server ready." before sending queries.

Deploying these Images Globally

If you deploy the TTS and STT images globally, you'll need to update the base_url values in the workey.py file to be correctly configured to the new URL values. Please do not forget this when deploying globally.

Additional Resources

• OpenAI API documentation
• IBM Watson Text-to-Speech documentation
• IBM Watson Speech-to-Text documentation
• IBM Watson Speech GitHub

Demo

Heres a quick video of planning a trip to Portugal with the Assistant. Audio output is not recorded but is there.

demo-video

Here is a screenshot of the application in action in dark mode

Conclusion

With the Personal Assistant, you can now interact with your computer in a natural and conversational way. We hope you find this project useful and enjoy using it as much as we enjoyed building it. Learn to build your own on CognitiveClass.ai