AI Agent CLI Tool - Assignment 02

A conversational command-line agent built with Node.js and the Gemini API. This tool acts as an autonomous web developer that takes natural language prompts, reasons through the necessary steps using a ReAct (Reason-Act) loop, and generates a fully functioning, styled website clone locally.

Features

• Conversational Interface: Chat directly with the AI in your terminal.
• Autonomous File Creation: The agent automatically generates index.html, style.css, and script.js files based on design instructions.
• System Command Execution: The agent can run system commands to automatically open the generated webpage in the browser.
• Strict JSON Reasoning: Uses Gemini's JSON mode to reliably loop through THINK, TOOL, and OBSERVE states before finalizing the OUTPUT.

Prerequisites

• Node.js installed on your machine.
• A valid Gemini API Key from Google AI Studio.

Setup Instructions

1. Clone this repository and navigate to the project directory: ```bash git clone https://github.com/MSSAYED11/Cli-Assignment.git cd Cli-Assignment ```

2. Install the required dependencies: ```bash npm install ```

3. Configure your API Key:

   • Rename the .env.example file to .env.
   • Add your actual Gemini API key to the file: ```env GEMINI_API_KEY=your_actual_api_key_here ```

How to Run

1. Start the Agent: Open your terminal and run the following command: ```bash node index.js ```

2. Test the Agent (Demo Prompt): Once the tool is running and waiting for input, paste the following prompt to watch the agent clone the Scaler Academy landing page step-by-step:

   Clone the Scaler Academy landing page in a folder 'scaler_clone' with index.html, style.css, and script.js. Use a professional white-label design: white background, navy text (#0b122f), and primary blue buttons (#0052ff). Section 1: Flexbox Navbar with links and a blue CTA. Section 2: Hero section with the headline 'Become the Professional Built for the Next Decade in AI' (color 'Next Decade in AI' blue). Section 3: Dual CTA buttons with hover effects. Section 4: Dark footer. Use the 'Inter' font and ensure it's responsive. Open the site in my browser when finished.

Architecture

This project utilizes the ReAct (Reason-Act) framework. The agent is provided with a system prompt that enforces a strict JSON output structure, allowing it to think about the task, call specific tools (createFile, executeCommand), observe the outcome, and repeat the cycle until the overall objective is completed.