Quickstart

Prerequisites

Before meeting Yumi, ensure you have the following installed on your machine:

• Python 3.12+
• A modern web browser (Chrome, Edge, Firefox)
• Microphone access enabled on your system.

---

1. Installation

First, clone the Yumi repository to your local machine and navigate into the directory.

git clone https://github.com/CodeNeuron58/Yumi.git
cd Yumi

Install the dependencies using uv (required — do NOT use pip):

uv sync

Note: Do not use pip install. Yumi pins torch to a CPU-only build. Using pip will try to download the full 2 GB CUDA version.

2. Attunement (First-Time Setup)

Yumi needs an LLM to think and a Text-to-Speech provider to speak.

Run the interactive dashboard.

If you activated your virtual environment:

# Windows
.venv\Scripts\activate
yumi

# macOS / Linux
source .venv/bin/activate
yumi

Or without activating:

uv run yumi

Tip: After activation, you can just type yumi for the rest of your terminal session. If you used the one-line installer, yumi always works globally without activation.

Since this is your first time, Yumi will automatically guide you through her Attunement process. You will be prompted to provide:

1. LLM API Key (We highly recommend Groq for its blazing fast speed).
2. TTS API Key (We recommend ElevenLabs for lifelike expression).

3. Waking Her Up

Once configured, simply select "🌸 Wake Yumi Up" from the main dashboard menu (or run uv run yumi wake-up).

1. The backend server will initialize on port 8000.
2. A browser tab will automatically open to http://localhost:8000/.
3. Allow microphone permissions if prompted.
4. Say hello!

Troubleshooting