nicolamassarenti / meta-assistant

An MVP that uses Google STT, OpenAI LLM, Nvidia Audio2Face
62 stars 12 forks source link

meta-assistant

An MVP of a virtual assistant that chats with you.

In a nutshell

meta-assistant

This application allows a user to talk and chat with a virtual assistant hosted in Nvidia Audio2Face tool. The key features are:

Setup

Local development

To install the application locally, you need to have poetry installed. To install the dependencies, run the following command:

poetry install

Before running the application, remember to open Nvidia Omniverse Audio2Face and to activate the streaming gRPC server. You can do it by clicking on Audio2Face, then click on Open Demo Scene and then on Full Face Core + Streaming Player. See image below for reference.

In addition, you have to have a Google Cloud account and a Google Cloud project with the following APIs enabled:

Then, you have to create a service account and download the JSON key file.

You also have to have a valid OpenAI API key.

Finally, you have to set the following environment variables:

GOOGLE_APPLICATION_CREDENTIALS="/path/to/your/service-account.json"
OPENAI_KEY="your-openai-key"
OPENAI_MODEL="text-davinci-003" # or any other model you want to use

Then, you're ready to run the application:

GRPC_SERVER="localhost:50051"
OPENAI_INSTRUCTION="answer to this sentence like you are chatting with a friend"

poetry run python -m meta_assistant \
    --grpc-server=$GRPC_SERVER \
    --openai-instruction=$OPENAI_INSTRUCTION

Contributing

If you want to contribute to this project, please read the contributing guidelines.

License

This project is licensed under the terms of the MIT license.

Authors