Overview

This project implements a minimal AI-powered chatbot using Cloudflare Workers and Workers AI (Llama 3.3). It supports conversational memory via Durable Objects and a simple web UI with Markdown rendering and voice input.

The webpage can be accessed at https://cf-ai-chat.sriram-venkatesh.workers.dev/

Features

• Chatbot API: /api/chat endpoint for chat interactions, powered by Llama 3.3.
• Session Memory: Persists chat history per session using Durable Objects.
• CORS Support: Handles CORS preflight and headers for API endpoints.
• Markdown Output: Responses are formatted in GitHub-Flavored Markdown, including fenced code blocks.
• Web UI: Minimal HTML page for chat, with Markdown rendering and voice input (Web Speech API).
• Session Reset: /api/reset endpoint to clear session memory.

Getting Started

Prerequisites

• Cloudflare Workers
• Workers AI
• Durable Objects

Installation

1. Clone the repository: git clone https://github.com/Sriramv739/cloudflare-ai-chatbot.git && cd cloudflare-ai-chatbot
2. Install dependencies: npm install
3. Run on localhost: npx wrangler dev

API Endpoints

• POST /api/chat
  • Request: { sessionId: string, message: string }
  • Response: { reply: string } Handles chat messages, persists session history, and returns AI-generated Markdown-formatted replies.
• POST /api/reset
  • Request: { sessionId: string }
  • Response: { ok: true } Clears session memory for the given session.
• GET / Returns the HTML chat UI.

Web UI

• Type your question and press Send or hit Enter.
• Hold the 🎙️ button to use voice input.
• Session data is temporary—closing the tab erases chat history.
• Demo only—do not share sensitive information.

Code Structure

• src/index.ts: Main Worker entry point, API routing, and session management.
• SessionDO: Durable Object class for session memory.
• getIndexHtml(): Generates the HTML page for the chat UI.
• Helper functions for CORS, JSON responses, Markdown formatting, and code block detection.

Customization

• System Prompt: Modify SYSTEM_PROMPT to change assistant behavior or formatting.
• Model: Change the model name in env.AI.run() to use a different LLM.