ChatAadi

ChatAadi is a full-stack AI chatbot application, for general purpose conversation, inspired by ChatGPT. It features Google-based authentication, persistent chat history, and token tracking – built with modern web technologies.

• In localhost

• In live deploy

🚀 Tech Stack

• Framework: Next.js (App Router, TypeScript)
• Authentication: NextAuth.js with Google OAuth
• Database: PostgreSQL via NeonDB
• ORM: Prisma
• UI: Tailwind CSS (optional)
• AI Integration: Ollama llama3.2 model as default for localhost and mistral-tiny model for production (both are free)

---

✨ Features

• 🔐 Google Sign-In via NextAuth
• 💬 Persistent chat history per user
• 🧠 Token usage tracking per message/session
• 🔁 Server-side chat streaming (optional)
• ⚡ Fast, modern architecture with Prisma + Neon

---

📦 Installation

1. Clone the Repository

git clone https://github.com/abchatterjee7/chat-aadi.git
cd chat-aadi

2. Install Dependencies

npm install
# or
yarn install

3. Set Up Environment Variables

Create a .env file in the root:

# PostgreSQL (NeonDB) connection
DATABASE_URL="postgresql://your_user:your_password@your_neon_host/dbname?sslmode=require"

# Google OAuth
GOOGLE_CLIENT_ID=your_google_client_id
GOOGLE_CLIENT_SECRET=your_google_client_secret

# NextAuth
NEXTAUTH_SECRET=your_random_secret
NEXTAUTH_URL=http://localhost:3000

# CORS
NEXT_PUBLIC_ALLOWED_ORIGINS=http://localhost:3000,https://chat-aadi.vercel.app

#AI Providers
#OLLAMA (localhost provider as default)
OLLAMA_URL=http://localhost:11434/api/generate
OLLAMA_MODEL=llama3.2
DISABLE_OLLAMA=false  # Set to true to force using Mistral


#Mistral (fallback provider, will be used in live deploy)
MISTRAL_API_KEY=your_mistral_api_key
MISTRAL_API_URL=https://api.laplateforme.io/mistral/v1
MISTRAL_MODEL=mistral-tiny

🔐 You can generate NEXTAUTH_SECRET using:

openssl rand -base64 32

---

🧪 Prisma Setup

Initialize the DB (for fresh project):

npx prisma migrate dev --name init

Or if you prefer schema push without migrations:

npx prisma db push

Generate Prisma Client

npx prisma generate

---

🏁 Run the App

npm run dev
# or
yarn dev

Visit: http://localhost:3000

---

🗂️ Project Structure

/chat-aadi/src
├── prisma/                 # Prisma schema & migrations
├── public/         
├── package.json            
├── .env                   
├── .tsconfig.json   
├── README.md             
└── src/              

---

🔐 Authentication: NextAuth + Google

You're using Google as the OAuth provider. Once logged in, user data is stored in the database (User, Account, Session models).

Make sure you've:

1. Created a project on Google Cloud Console
2. Enabled OAuth 2.0
3. Added http://localhost:3000 as an authorized redirect URI

---

🧠 Models Overview

User

• Basic info, profile image, credits
• Relations: accounts, sessions, chats, token usage

Chat

• Belongs to a user
• Contains many messages
• Token usage tracking

Message

• Long content using @db.Text
• Roles: user, assistant

TokenUsage

• Tracks token count per chat

---

🌐 Deployment

You can deploy this to:

• Vercel (recommended for Next.js)

NeonDB for PostgreSQL

• Free serverless Postgres
• Optimized for edge functions
• Get started: https://neon.tech

---

📅 Roadmap

• Integrate Mistral API (Mistral 7B free model) & Ollama llama3.2 (for localhost chat completion)
• Stream responses from the server
• Token counting (approximate & actual)
• User settings page
• Export chat history
• Mobile-first responsive UI

---

🙋 FAQ

Q: Why use NeonDB? A: It’s a modern, serverless Postgres DB perfect for Vercel or other edge-first platforms.

Q: What about Ollama and Mistral integration? A: This template lays the foundation. You’ll soon be able to hook it up with Mistral API (In localhost and default, using ollama llama3.2 model) to generate responses.

Q: How are messages stored? A: Each message is stored in the Message model as a TEXT field, allowing for large content.

---

🧑‍💻 Author

ChatAadi by [Aaditya B Chatterjee] Made with 💙 and inspired by ChatGPT.

---

📜 MIT License

This project is licensed under the MIT License.

---

Major Pending Task

• Generating text to image and image to image like Ghibli images (say using stability.ai api)
• subscription monthly implementation with razorpay and token spent
• forgot/reset password & email verification with otp
• ollama not working.
• edit profile with image,name
• Mobile Compatable UI

Vercel Deployment

• Ensure .env file. It must also have CORS as,

NEXT_PUBLIC_ALLOWED_ORIGINS=http://localhost:3000,https://chat-aadi.vercel.app

• Add .nvmrc file with 18 (for NodeJs version)
• Add .vercelignore file
• During build need

npm i
npx prisma generate (to generate prisma client)
npm run build
npm run start

• Under "Build & Development Settings":

Set "Build Command" to: npm run vercel-build
Set "Output Directory" to: .next
Set "Install Command" to: npm install

• In case it got deployed, find it live here:
• https://chat-aadi.vercel.app