Jarvis

Jarvis is a personal voice assistant built on open source software. It runs everything on the host machine and is intended for people looking to run a GPT-like assistant on their own infrastructure.
Jarvis uses OpenAI's Whisper for speech-to-text transcription and llama.cpp (with the OpenLLaMa 3b model) as a Large Language Model, focused on user interaction. OpenLLaMa is an openly licensed reproduction of Meta's original LlaMa model.

Dependencies

The project uses the folowing projects as dependencies:

• arecord (voice recording);
• Python >= 3.9
• virtualenv
• whisper (speech-to-text);
• llama.cpp (Text generation);

Installation

1. Clone the repository:

git clone https://github.com/mlemosf/jarvis-assistant

2. Enter the main folder:

cd jarvis-assistant

3. Create a virtual environment and activate it:

virtualenv venv
source venv/bin/activate

4. Clone the whisper repository:

git clone https://github.com/openai/whisper.git

5. Clone the llama.cpp repository and follow the installation procedures. We recommend using the CUDA build:

git clone https://github.com/ggerganov/llama.cpp

6. Run the setup script:

./setup.sh

7. Run the Jarvis script

./run.sh

References

1. OpenLLaMa