dringai-case

This repo contains a simple real time voice assistant controlled with SPACE/ESC keys and designed for conversational interactions which are ideal for customer service and support.

What could be done better?

• Latency is huge, the data should flow through the streams to reduce latency
• More robust error handling, maybe custom exceptions
• Smarter assistant/llm logic

---

Installation & Usage

I'm not gonna include installation & usage with Docker, because I was unable to test it since I'm not on Linux and do not really had the time to create a VM. The image is building though.

git clone https://github.com/canzden/dringai-case.git
cd dringai-case

# create and activate venv
python -m venv .venv
source .venv/bin/activate

# generate deps from .in file to avoid os specific deps
pip install pip-tools
pip-compile -o requirements.txt requirements.in

# install deps
pip install -r requirements.txt

# copy .env file and fill api keys
cp .env.example .env

# run (assuming accessibility settings are granted)
python src/orchestrator.py

Logs

Each new conversation is saved in to data/logs as jsonl files.

{
    "ts": "2025-08-13T23:55:37.075+00:00",
    "turn_id": 1,
    "user_text": "Merhaba, bana randevu konusunda yardımcı edin misin lütfen?",
    "assistant_text": "Merhaba, ben DringAI müşteri hizmetleri asistanıyım. Randevu konusunda size nasıl yardımcı olabilirim? Hangi tarih ve saat için randevu almak istiyorsunuz?"
}