README.md

metadata

title: Thai Energy AI Ambassador
emoji: ⚡
colorFrom: purple
colorTo: pink
sdk: gradio
python_version: '3.10'
app_file: app.py
license: mit
short_description: This is an AI ambassador for the Thai energy sector.
pinned: false
sdk_version: 5.47.2

Thai Energy AI Ambassador

This project demonstrates a secure, PDPA-compliant AI avatar that acts as a brand ambassador for the Thai energy sector. It utilizes a custom-trained Thai language model that is exclusively knowledgeable about documents from the MEA, PEA, EGAT, and other Thai energy authorities.

Project Phases

• Phase 1: Data Pipeline: Automated web scraping of official websites to create a private and secure dataset.
• Phase 2: LLM Training: Fine-tuning a small, efficient Thai language model on the collected data.
• Phase 3: AI Avatar Development: Building a desktop application with a visual avatar and text-to-speech (TTS) capabilities.

How it Works

1. Data Collection: A Zapier workflow scrapes public-facing documents and news from official Thai energy websites.
2. Dataset Transformation: The raw data is automatically formatted into a structured dataset for model training.
3. Model Training: The custom dataset is used to fine-tune a small language model on a Hugging Face Space.
4. Secure Deployment: The trained model is served privately from the Hugging Face Space to a local Windows application, ensuring no data leaves the controlled environment.

---

Getting Started

To get started with this project, ensure you have the required files in your repository:

• README.md (this file)
• requirements.txt
• app.py