Description

A chat app for end-to-end voice conversations with LLMs that stores the conversation in text and audio format on the server

Setup

• Clone the model repositories to the models directory:

  1. Speech-to-text: https://huggingface.co/openai/whisper-medium
  2. Chat: https://huggingface.co/Qwen/Qwen3-8B-FP8
  3. Text-to-Speech: https://huggingface.co/kyutai/tts-1.6b-en_fr, https://huggingface.co/kyutai/tts-voices

• For the python client the requirements in client/requirements.txt must be installed. For PyAudio the dev libraries of PortAudio need to be installed (cf. server/stt/Dockerfile)

Run

1. Server: Assuming Docker Compose is installed: Execute ./run.sh in the server directory.
2. Python-Client: python -m client.client (in voice_note directory)