Full-Stack Cloudflare AI Assistant

Live Demo: https://my-angular-app.sidharthak25.workers.dev

This project is a real-time, stateful AI chat application built entirely on Cloudflare's edge infrastructure. It features a modern, responsive UI built with Angular SSR (Server Side Rendering) that supports both text and voice interactions, utilizing Cloudflare Workers AI for natural language processing and Durable Objects for isolated, auto-cleaning session memory.

Core Components

• 1. LLM (Large Language Model):

  • Uses Cloudflare Workers AI to run Meta's Llama 3 for text responses and OpenAI's Whisper for audio transcription directly at the edge.

• 2. Workflow / Coordination:

  • A Cloudflare Worker serves as the central orchestrator, routing HTTP requests, handling multipart audio files, and sequencing the AI models.

• 3. User Input (Chat & Voice):

  • A custom Angular UI captures text or records voice via the native MediaRecorder API, transcribing audio for user review before prompting the LLM.

• 4. Memory / State:

  • Cloudflare Durable Objects (SQLite backend) provide isolated session histories tied to unique user UUIDs, utilizing the Alarms API for automatic self-deletion of abandoned chats.

Local Development & Deployment

Prerequisites

• Node.js installed.
• Wrangler CLI authenticated (npx wrangler login).

Setup Instructions

1. Install dependencies:

   npm install

2. Build the Angular frontend and SSR engine:

   npm run build

3. Test locally:

   npx wrangler dev

4. Deploy to Cloudflare:

   npx wrangler deploy

   (Note: Ensure your wrangler.json includes the new_sqlite_classes migration array for free-tier Durable Object compatibility).