1. Project Overview

Aria is a generative AI chatbot powered by Anthropic’s Claude model, integrated into a standalone website built with Python (Flask). Unlike rule-based bots that only match predefined keywords, Aria generates intelligent, context-aware responses to any question in real time.

The project covers the full development lifecycle: designing the chatbot personality, building the web interface, integrating the AI model, and deploying to AWS Elastic Beanstalk with auto-scaling.

2. Objectives

• Design an AI-powered chatbot using a generative model (Claude API)
• Enable instant responses to user queries on websites
• Train the chatbot with predefined input patterns for commercial use (system prompt)
• Integrate the chatbot seamlessly with the target website interface
• Optimize and test the chatbot for accuracy and user engagement

3. System Architecture

User Browser
     |
     | HTTP Request
     v
Flask Web Server (AWS EC2 / Local)
     |
     |-- GET  /           --> Landing page (index.html)
     |-- GET  /chat       --> Chat interface (chat.html)
     |-- POST /api/chat   --> Receives message, calls Claude, returns reply
     |-- GET  /api/health --> Health check endpoint
     |
     | HTTPS API Call
     v
Anthropic Claude API (claude-sonnet model)
     |
     | JSON Response
     v
Flask returns reply to browser
     |
     v
JavaScript updates chat UI in real time

How a message flows:

1. User types a message and presses Enter
2. JavaScript collects the full conversation history plus new message
3. A POST request is sent to /api/chat
4. Flask sends the messages to the Claude API with the system prompt
5. Claude generates a response
6. Flask returns the response as JSON
7. JavaScript displays the response in the chat UI with formatting

4. Tech Stack

5. Phase 1 — VS Code Project Setup