Bruno is a speech-to-text-to-AI tool designed to facilitate collaborative learning in groupwork. The prototype application listens to human speech, provides a written transcript that identifies individual speakers, and passes the transcript and a chosen prompt to a LLM.
First, you can log into the app with Google.
You will find a full dashboard with several features.
Next, you can click "record" to start the transcript, and pause at any time.
The transcript with diarization of speakers will be created through Deepgram and displayed on the frontend.
You can enter context regarding the conversation and read the instructions for use of Bruno.
Finally, you can query Bruno by prompting and receiving a response to help facilitate your group collaboration.
You can customize the display, persona, and prompt at any time through the settings panel.
If at any time your connection has disconnected from the transcription service, you will see a warning.
Alternatively, you can upload or copy and paste a transcript instead of recording it live. Visit "/upload" for the following page.
The frontend is built with html and tailwindcss. The backend is built with the django rest framework, a decision made due to Django's robust feature set, out of the box admin and security capabilities, and compatability with the existing project code written in python. The project also leverages Docker to normalize runtime environments and increase the ease of collaboration.
The backend serves HTML frontend. After the user logs in with Google, they have access to live transcription through audio data passed through the backend to Deepgram (transcription API) and relayed back to the frontend. This transcript is also stored in the database, allowing the user to ask Bruno (GPT with custom prompts and transcript) questions about the conversation.
If you're interested in developing a specialized version of Bruno, we encourage you to fork this repository and implement your changes there. However, if you have ideas or contributions that could broadly enhance Bruno's functionality, please read our contribution guidelines for more information on how you can contribute to this repository. We greatly appreciate your involvement and input!
v1.0.0 - October 26th, 2023. First open-source stable version.
v1.0.1 - January 6th, 2024. Full replacement of Middlesight with OpenAI
v1.1.0 - January 11th, 2024. Ability to customize Bruno persona and prompt, upload and edit custom transcript
GNU General Public License v3.0 or later
Contributors are able to contribute to the original codebase or create a forked open-source version. We request that you include Principles.md in any derivation of the codebase.
See COPYING to view the full text.
Before you can deploy the application, you must first set your environment variables. This includes Google Oauth, deepgram, and OpenAI API keys. You can use the environment variables as a list of required steps for setting up the project.
For deploying the application locally, the following environment variables are required:
------- [local.env] --------
DEPLOYMENT=local
DJANGO_SETTINGS_MODULE=ai4collab.settings
DEEPGRAM_API_KEY=<add here>
OPENAI_API_KEY=<add here>
GOOGLE_CLIENT_ID=<add here>
GOOGLE_SECRET_KEY=<add here>
For deploying the application to a development environment, the following environment variables are required:
------- [development.env] --------
DEPLOYMENT=development
DJANGO_SETTINGS_MODULE=ai4collab.settings
SECRET_KEY=<add here>
PSQL_DATABASE_URL=<add here>
ALLOWED_HOSTS=<add here>
DEEPGRAM_API_KEY=<add here>
OPENAI_API_KEY=<add here>
GOOGLE_CLIENT_ID=<add here>
GOOGLE_SECRET_KEY=<add here>
Finally, deploying the application to a production environment requires the following environment variables:
------- [production.env] --------
DEPLOYMENT=production
DJANGO_SETTINGS_MODULE=ai4collab.settings
SECRET_KEY=<add here>
PSQL_DATABASE_URL=<add here>
ALLOWED_HOSTS=<add here>
DEEPGRAM_API_KEY=<add here>
OPENAI_API_KEY=<add here>
GOOGLE_CLIENT_ID=<add here>
GOOGLE_SECRET_KEY=<add here>
After selecting your deployment environment, the list of environment variables functions as a to-do list for setting up the project. Below are instructions for each variable.
The DEPLOYMENT variable is required for all deployments and is self explanatory; It can be set to local, development, or production.
The DJANGO_SETTINGS_MODULE variable is required for all deployments and does not need to be customized. It should be set to ai4collab.settings.
The SECRET_KEY variable is required only for development and production deployments, as a hard-coded insecure key is used for local. The secret key is a requirement of Django applications and used for security. The key is typically 50 characters long, with a diverse set of letters, digits, and special characters. There are several generators that can be found online, or you can generate one by running the following command:
python -c 'import secrets; print(secrets.token_urlsafe(38))'
The PSQL_DATABASE_URL variable is required only for development and production deployments, as local deployment uses an SQLite database. PSQL_DATABASE_URL should include a PostgreSQL database with the username, password, host, port, and database name in the URL.
The ALLOWED_HOSTS variable is a comma-seperated list of strings representing the allowed hosts for the Django app. This is required only for development and production deployments, as allowed hosts is set to "*" in local. An example for development is the following: ALLOWED_HOSTS=localhost,127.0.0.1,exampledev.railway.app
The DEEPGRAM_API_KEY variable is required for all deployments. It can be obtained by creating a free account on deepgram.com, which will issue your account free credits to start.
The OPENAI_API_KEY variable is required for all deployments. It is the OpenAI API key used for powering Bruno. This is the only third-party API key for Bruno which may need payment information to activate.
The GOOGLE_CLIENT_ID and GOOGLE_SECRET_KEY are required for all deployments. In order to configure Google Oauth, you may consult the following link: https://support.google.com/googleapi/answer/6158849?hl=en. Authorized JavaScript origins for local deployment should include "http://127.0.0.1:8000". Authorized redirect URIs for local deployment should include "http://127.0.0.1:8000/allauth/google/login/callback/". For development and production deployments, "http://127.0.0.1:8000/" should be replaced with deployment URL.
Once you have set all the variables in a file named {deployment}.env (e.g. local.env, development.env, production.env), you can move to the next step. Using docker is recommended, but instructions for deployment without docker are included.
If you are deploying to production or development outside of localhost, you will need to modify settings.py to include the new URL under CSRF_TRUSTED_ORIGINS, and modify the url_to_use variable in audio/views.py.
docker build -f Dockerfile.norail -t ai4collab:latest .
docker run --env-file {DEPLOYMENT}.env -p 8000:8000 ai4collab:latest
DEPLOYMENT = local, development, or production. Run with -d for detached.
The application is now up and running! You can check the status of the deployment for any errors in the logs.
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
touch {DEPLOYMENT}.env
DEPLOYMENT = local, development, or production
export $(cat {DEPLOYMENT}.env | xargs)
python manage.py test
python manage.py check --deploy
python manage.py makemigrations
python manage.py migrate
python manage.py collectstatic
python manage.py createsuperuser
uvicorn ai4collab.asgi:application
gunicorn --bind 0.0.0.0:8000 --workers 4 -k uvicorn.workers.UvicornWorker ai4collab.asgi:application
The application is now up and running!
You can check the status of the deployment for any errors in the logs, or visit /deployment while logged in as a superadmin through the admin panel at /admin.
You can make requests for new features under issues, or contribute code.
Current list:
- Ability to customize prompt
- Add increased test coverage
- Add account-specific rate limiting