mvanzulli / MeetingAssistant.py

A local deployable version of an AI meeting assitant

🎙️ Meeting Assistant

This program provides functionality to record meetings and then generate summaries of the transcribed audio using OpenAI's GPT-3 language model.

🚀 Prerequisites

To run this program, you will need:

• Python 3.7 or higher
• ffmpeg installed on your system. If not:

  sudo apt install ffmpeg

🔧 Installation

1. Clone this repository
2. Create an enviroment and install dependencies

  python3 -m venv .venv
  source .venv/bin/activate
  pip install -r requirements.txt

🔑 Required Environment Variables

This program requires an OpenAI API key to function. To set up your environment variables:

1. Create a new file named .env in the root of this directory
2. Add the following line to your .env file: OPEN_API_KEY=<your_api_key>
3. Replace <your_api_key> with your actual OpenAI API key.

🛠️ Usage

UI

Launch the Web-UI:

./run.sh

Launch the GUI:

python3 gui/gui.py

Record

To record a meeting, run:

python model/model.py record <output_file_name>.mp3

This will record the audio and save it to a file with the specified name.

Summarize

To generate a summary of an audio file, run:

python3 model/model.py summarize <audio_file_name>.mp3 

This will generate a summary of the transcribed audio using OpenAI's GPT-3 language model in the same audio language.

Translate and Summarize

To generate a summary of an audio file and translate it into <language_key>, run:

python3 model/model.py summarize <audio_file_name>.mp3 <language_key>

You can see the language keys in the language_roles.yaml file.

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

💼 Variables to tune

Feel free to change:

This program has several variables that can be tuned to change its behavior. These variables are declared at the beginning of the program:

• OS: Set this to "linux" or "MAC" depending on your operating system.
• DEVICE: Set this to "cuda:0" if you have an NVIDIA GPU and want to use it to accelerate processing, or "cpu" to use the CPU instead.
• WHISPER_MODEL: The name of the pre-trained Whisper model to use for transcribing the audio.
• ENV_OPENAI_KEY: The name of the environment variable that contains your OpenAI API key.
• TEMPERATURE: The "temperature" parameter to use when generating summaries with GPT-3. Higher values will generate more diverse summaries, while lower values will generate more conservative summaries.
• GPT_MODEL: The name of the GPT-3 language model to use for generating summaries.
• GPT_ENCODER: The name of the GPT-3 tokenizer to use for encoding text.
• SIZE_CHUNK: The size of each "chunk" of text to send to GPT-3 for summarization. Larger chunks will result in fewer requests to the API, but may be slower to process.
• Command prompts and command role in the language_roles.yaml file.