🧠 AltmanGPT

AltmanGPT is an immersive AI "Digital Twin" of Sam Altman, designed to simulate conversation with the OpenAI CEO. It goes beyond simple mimicry by using RAG (Retrieval-Augmented Generation) to ground every answer in Sam's actual blog posts, essays, and public writings.

Featuring a "YC Partner Roast Mode", Voice Input, and a premium dark-mode aesthetic, this project demonstrates how to build high-quality, persona-driven AI applications using Google's Gemini 2.5 Flash model.

(Add your screenshot here)

---

✨ Key Features

• 🧠 Authentic RAG Intelligence: Answers are retrieved from a vector database (Pinecone) containing Sam Altman's real writings, ensuring the advice is grounded in his philosophy.
• 🔥 "Roast My Startup" Mode: A toggleable mode where the AI switches personas from "Calm CEO" to "Ruthless YC Partner," specifically designed to stress-test startup pitches.
• 🎙️ Voice-First Interface: Real-time Speech-to-Text integration allows users to speak directly to the AI, creating a more natural conversational flow.
• 🔐 User Authentication: Secure Google Sign-In integration via Firebase Authentication.
• 🎨 Premium Aesthetic: A bespoke "Crimson/Dark" UI with glassmorphism, smooth animations, and the Inter typography system.

---

🛠️ Tech Stack

Frontend

• Framework: React.js (Vite)
• Styling: Tailwind CSS + Custom CSS Variables
• Auth: Firebase Authentication (Google Provider)
• State: React Hooks (useState, useEffect, useRef)
• Icons: Lucide React

Backend

• Runtime: Node.js & Express
• AI Model: Google Gemini 2.5 Flash (via @google/genai SDK)
• Vector DB: Pinecone (for RAG context retrieval)
• Embedding: Text embedding model for semantic search

---

🚀 Getting Started

Prerequisites

• Node.js (v18+)
• A Google Gemini API Key
• A Pinecone API Key (and an index named samagpt)
• A Firebase Project

1. Clone the Repository

git clone https://github.com/creation22/samagpt.git
cd samagpt

2. Backend Setup

Navigate to the backend folder and install dependencies:

cd backend
npm install

Create a .env file in the backend directory:

API_KEY=your_google_gemini_api_key
PINECONE_API_KEY=your_pinecone_api_key

Start the backend server:

node server.js

Server runs on http://localhost:5000

3. Frontend Setup

Navigate to the frontend folder and install dependencies:

cd ../frontend
npm install

Configure Firebase: Open src/firebase.js and replace the placeholder config with your Firebase credentials:

const firebaseConfig = {
  apiKey: "YOUR_API_KEY",
  authDomain: "YOUR_PROJECT.firebaseapp.com",
  // ... other keys
};

Start the development server:

npm run dev

Frontend runs on http://localhost:5173

---

💡 How It Works (Architecture)

1. Input: User types or speaks a question (e.g., "How do I find product-market fit?").
2. Context: The backend queries Pinecone for the most relevant paragraphs from Sam's blog.
3. Prompting: The system constructs a prompt combining:
   • The User's Question
   • The Retrieved Context
   • The Selected Persona (Standard vs. Roast)
4. Generation: Google Gemini 2.5 Flash generates a response in Sam's style.
5. Delivery: The response is streamed/sent back to the React UI.

---

📄 License

This project is open-source and available under the MIT License.

---

Built with ❤️ for the Hackathon.