🤖 Physical AI Humanoid Textbook

An AI-powered interactive humanoid textbook platform that combines a notebook-style physical AI book, a React-based RAG chatbot, and an agentic AI backend to simulate a real-world Physical AI learning experience.

This project was built for a hackathon, focusing on modern AI tooling, retrieval-augmented generation, and human-like interaction with educational content.

---

📌 Project Idea (High-Level)

The core idea of this project is to create a Physical AI Book — a notebook-style digital textbook about Physical AI — enhanced with an intelligent chatboard that allows users to ask questions directly from the book content.

• 📘 The book is built using Docusaurus
• 💬 A React-based chatboard is embedded alongside the book
• 🧠 The chatboard uses RAG (Retrieval-Augmented Generation)
• 🤖 The backend is agent-driven using OpenAI Agents SDK
• 🔗 Models are served via OpenRouter
• 📦 Knowledge is stored using Qdrant + Cohere embeddings

---

✨ Key Features

Frontend (Physical AI Book + Chatboard)

• 📘 Docusaurus-based Physical AI Notebook
• 🧩 Structured chapters and learning content
• 💬 Embedded React RAG Chatboard
• 🔄 Real-time question answering from book content
• 🎯 Context-aware responses linked to textbook material

Backend (Agentic AI)

• 🤖 Agent-based backend using OpenAI Agents SDK
• 🧠 RAG pipeline with semantic search
• 🔎 Qdrant vector database
• 📐 Cohere embeddings
• 🌐 OpenRouter-hosted LLMs

---

🧰 Tech Stack

📘 Frontend (Book + UI)

• Docusaurus (Physical AI Notebook)
• React
• JavaScript / TypeScript
• Markdown-based content system

🧠 Backend

• Python
• FastAPI
• OpenAI Agents SDK
• OpenRouter API
• Qdrant Vector Database
• Cohere Embeddings

🧪 Tooling

• Gemini CLI
• Notebook-based data preparation
• Prompt engineering for agents

---

📘 Frontend Explained (Very Important)

🧠 Physical AI Notebook (Docusaurus)

The Physical AI Book is built using Docusaurus, which acts as a structured digital notebook:

• Chapters written in Markdown
• Topics focused on Physical AI concepts
• Acts as the knowledge source for the RAG system
• Easy to extend and maintain

This makes the project feel like a real textbook, not just a chatbot.

---

💬 React RAG Chatboard

Alongside the notebook, a React-based chatboard is integrated:

• Users ask questions while reading the book
• Queries are sent to the backend
• Relevant content is retrieved from the book embeddings
• Answers are generated using LLMs with context

The chatboard transforms the book into an interactive humanoid tutor.

---

🧠 RAG + Agent Flow