🌐 AI Chatbot with Groq + web Search + Memory

A lightweight and powerful AI chatbot built using Groq LLM, Tavily Web Search, and NodeCache conversation memory. Supports real-time web search, tool calling, and thread-based memory.

🚀 Features

🔍 Real-time web search using Tavily .

🧠 Per-thread memory (remembers past messages for 24 hours)

🤖 Groq Llama 3.1 model for fast responses

⚙️ Automatic tool calling (LLM decides when to search the web)

🪶 Clean, modular code

🔁 Simple generate(message, threadId) function for easy integration

📦 Installation 1️⃣ Clone the repo git clone https://github.com/yourusername/your-repo.git cd your-repo

2️⃣ Install dependencies npm install

3️⃣ Create .env file GROQ_API_KEY=your_groq_key TAVILY_API_KEY=your_tavily_key

Get keys:

Groq key → https://console.groq.com/keys

Tavily key → https://app.tavily.com/api-key

🧩 Project Structure /project │── index.js # main server file (optional) │── chatbot.js # your chatbot logic (Groq + Tavily + memory) │── package.json │── .env │── README.md

🔥 How to Use the Chatbot Import and call the generate() function: import { generate } from "./chatbot.js";

const reply = await generate("What is the price of iPhone 16?", "user1"); console.log(reply);

Clear memory: import { clearMemory } from "./chatbot.js";

clearMemory("user1");

🧠 Conversation Memory

Memory is stored using NodeCache

Each threadId has its own private history

Auto-expires after 24 hours

You can create multiple user sessions:

generate("hello", "userA"); generate("hello", "userB");

🌍 How Web Search Works

The LLM decides automatically when it needs web search using the tool_calls feature.

If it requests:

{ "name": "webSearch", "arguments": { "query": "something" } }

→ your code runs webSearch() → gives fresh info back to the model → second pass generates a final answer

📡 Example API Endpoint

If you want a simple Express server:

import express from "express"; import { generate } from "./chatbot.js";

const app = express(); app.use(express.json());

app.post("/chat", async (req, res) => { const { message, threadId } = req.body;

const reply = await generate(message, threadId || "default"); res.json({ reply }); });

app.listen(4300, () => console.log("Server running on 4300"));

🛠️ Technologies Used

Node.js

Groq SDK

Tavily Search

NodeCache

Tool Calling Architecture

🤝 Contributing

Feel free to submit PRs or issues.