✨ AI Mate

A modern, ultra-low latency web application for conversational AI via voice. AI Mate uses the GPT-5.4 family to allow real-time conversational STT/TTS (Speech-to-Text & Text-to-Speech), backed by a beautiful Glassmorphic frontend and a powerful Streamlit Admin Dashboard for persona management and history viewing.

Core Features

1. Ultra-Low Latency Streaming: Text is streamed directly from the gpt-5.4-mini model, chunked, and synthesized into MP3 bytes on the fly. The browser natively plays this HTTP stream without waiting for the full generation payload.
2. Glassmorphic Web Interface: A premium dark-mode graphic UI featuring intuitive Push-To-Talk micro-animations.
3. Animated AI Avatar: Features a beautiful 2D digital avatar that automatically synchronizes its mouth movements and body bobbing to the generated speech audio stream in real-time.
4. Streamlit Admin Dashboard: Launch a dedicated internal dashboard (/run_streamlit) to visualize chat logs stored in Google BigQuery, manage your AI's persona parameters (persona.json), and clear histories dynamically.
5. Cloud Run Ready: Built on Flask and ready for highly scalable container deployments.

Requirements

• Python 3.8+
• OpenAI API key
• Google Cloud SDK (for BigQuery logs and Cloud Run deployment)

Installation

1. Clone this repository:

   git clone https://github.com/yourusername/ai_mate.git
   cd ai_mate

2. Install dependencies:

   pip install -r requirements.txt

Usage

1. Set your OpenAI token as an environment variable (note: the code expects OPENAI_TOKEN):

   $env:OPENAI_TOKEN="your-api-key"   # PowerShell
   # or
   export OPENAI_TOKEN="your-api-key" # Bash/Zsh

2. Run the Flask application locally:

   python main.py

3. Open your browser and go to http://localhost:8080.

4. (Optional) Open the Streamlit Admin dashboard by clicking the Admin gear icon in the footer, or navigating directly to http://localhost:8501.

Deployment

To deploy on Google Cloud Run:

1. Make sure you are authenticated with Google Cloud and have set your project (this is also required for the BigQuery logger):

   gcloud auth login
   gcloud config set project YOUR_PROJECT_ID

2. Deploy using the following command (replace parameters as needed):

   gcloud run deploy ai-mate --allow-unauthenticated --region=asia-northeast1 --project=YOUR_PROJECT_ID --source .

Contributing

Contributions are welcome! Please open an issue or submit a pull request.

License

This project is licensed under the MIT License. See the LICENSE file for details.