git clone https://github.com/vindexTOS/supportLLM.git
cd cd supportLLM/
docker compose up --build
After Docker is done building, LLAMA is going to start downloading the LLaMA 3 model in the background (4.7GB download). Until the download is completed, the system is not going to be operational.
After setup is completed, go to http://localhost:3000 and start recording audio.
FIRST TIME LLM NEEDS TO WARM UP, which means it’s going to take 1–2 minutes before the LLM is fully cached into the GPU. After that, it’s going to work just fine.
The assistant will ask you questions. You can try to talk to it about different topics. It’s going to try to gather your email, phone number, and name. After gathering information, it’s going to save the gathered information in the MongoDB database.
The LLM is context-aware, and it can also cross-reference old messages from different conversations, which means the more you talk to it, the more it’s going to learn.
After completing the task, your session will be terminated and the chat will be closed.
Speech To Text Whisper LLM LLama 3 Text To Speech Kokoro
STT -> LLM -> TTS
expressjs with typescript and mongodb utilizing CLEAN architecture
reactjs with antdeisgn and tanstackquery utilizing CLEAN architecture for React