AI Chatbot with Memory

An interactive AI chatbot built with Vue.js (frontend) and Node.js (backend) that remembers previous conversations using LangChain and Pinecone.
Supports multiple users, selectable personalities, and message-based memory storage.

---

🚀 Features

• 🧠 Persistent Memory – Stores user messages in Pinecone so the bot remembers context across sessions.
• 👥 Multiple Users – Each user gets their own isolated memory space (auto-generated UUID).
• 🎭 Personalities – Switch between helpful, funny, and formal response styles.
• 💬 Conversation History – Keeps short-term memory for local continuity (last 6 messages).
• 🔒 Safety Rules – Restricts AI behavior (e.g., no code output).
• 🧩 Modular Code – Clean separation of backend and frontend, easy to extend.

---

🏗️ Tech Stack

Area	Technology
Frontend	Vue.js (Vite) + Axios
Backend	Node.js + Express
AI Engine	OpenAI GPT-4 (via LangChain)
Memory	Pinecone Vector Database
Vector Dimension	1536 (text-embedding-3-small)
UI/UX	Custom CSS

---

⚙️ Setup Instructions

1️⃣ Clone the Repository

git clone https://github.com/<your-username>/ai-chatbot-with-memory.git
cd ai-chatbot-with-memory

2️⃣ Install Dependencies

Backend

cd backend
npm install

Frontend

cd frontend
npm install

3️⃣ Environment Variables

Create a .env file inside the backend folder:

OPENAI_API_KEY=your_openai_api_key_here
PINECONE_API_KEY=your_pinecone_api_key_here
PINECONE_INDEX=chatbot-memories
PORT=3000

4️⃣ Run the Servers

Backend

cd backend
npm run dev

Server runs on: http://localhost:3000

Frontend

cd frontend
npm run dev

Frontend runs on: http://localhost:5173

---

💬 How It Works

1. Each user gets a unique userId (stored in localStorage).
2. When a message is sent:
   • The last 6 messages (short-term context) are sent to the backend.
   • LangChain uses OpenAI embeddings to encode the new message.
   • Pinecone searches for similar past vectors (long-term memory).
   • The most relevant past facts are injected into the GPT prompt.
   • The AI generates a response according to the chosen personality.
   • The user message (and optionally reply) are stored in Pinecone.

🧠 Short-term memory: Maintained locally (recent 6 messages).
🗄️ Long-term memory: Stored in Pinecone, retrieved via semantic similarity.

---

🔒 Restrictions & Safety

This chatbot includes a global restriction layer defined in server.js:

const restrictionPrompt = `
You are a helpful assistant, but you must follow these restrictions:
- Do NOT provide or generate any code snippets or pseudocode.
- Do NOT include Markdown code blocks or language names.
- If a user asks for code, politely refuse and explain conceptually instead.
`;

🛡️ You can modify or expand this section to restrict other behaviors.