This repository contains Dishant Zaveri’s Cloudflare AI Agent project, built as part of the optional assignment for the Cloudflare Software Engineer Internship (Winter/Spring 2026).
It showcases how to build a full-stack AI-powered application entirely on Cloudflare’s developer platform.

• LLM Integration: Uses Workers AI with Llama 3.3 Instruct for reasoning and conversation.
• Durable Objects: Provides persistent chat sessions, scheduling (alarms), and multi-user state.
• Vectorize: Adds semantic memory (768-dim embeddings with @cf/baai/bge-base-en-v1.5) so the agent can “remember” facts about you.
• Realtime Communication: WebSockets keep the chat UI live and responsive.
• Modern UI: Custom Pages frontend with tabs (About, Project, Agent, Contact). Includes:
  • Chat interface with auto-scrolling conversation history
  • Preset chips for quick prompts
  • Note saving (/note)
  • Search tool (/search <url>)
  • Voice input (Web Speech API)
  • Clear chat button
• Optional Assignment Ready: Demonstrates all required components:
  • LLM (Workers AI)
  • Workflow/coordination (Durable Objects + alarms)
  • User input (chat + voice)
  • Memory/state (Durable Object storage + Vectorize)

👉 https://cf-agents-cf4242.dishantzaveri.workers.dev

Open the demo, click Talk to my Agent, and start chatting!
Examples:

• “Introduce yourself in one sentence.”
• “/note Dishant has 8 published research papers.”
• “What do you remember about me?”
• “/search https://blog.cloudflare.com/”

cf_ai_agent/
├── public/             # Frontend UI (Pages assets)
│   ├── index.html
│   ├── styles.css
│   └── app.js
├── src/                # Cloudflare Worker & Durable Object
│   ├── index.ts        # Routes: /api/chat, /api/note, /api/schedule, /api/history, /api/clear, /ws
│   ├── agent.ts        # Durable Object logic (state, memory, scheduling)
│   └── rag.ts          # Vectorize (embedding + retrieval)
├── wrangler.toml       # Worker configuration
├── README.md           # Project documentation
└── PROMPTS.md          # AI prompts used during development

1. Install dependencies

   npm install

2. Start local dev server

   wrangler dev --local

   Visit http://localhost:8787.

3. Deploy to Cloudflare

   wrangler deploy

   Ensure you have:

   • A Durable Object migration (MyAgent)
   • A Vectorize index created:

     npx wrangler vectorize create mem-index --dimensions=768 --metric=cosine

• POST /api/chat – Chat with the agent

  { "userId": "dishant", "prompt": "Say hi" }

• POST /api/note – Save a note into memory

  { "userId": "dishant", "text": "Dishant won the Singapore–India Hackathon" }

• POST /api/schedule – Schedule a reminder

  { "userId": "dishant", "seconds": 30, "note": "Follow-up reminder" }

• GET /api/history – Get conversation history

• POST /api/clear – Clear all history/memory

• WS /ws – Real-time WebSocket chat

Dishant Zaveri

• Master’s in Computer Science @ Texas A&M University (graduating May 2026)
• 8 published research papers (ML/Systems)
• 15+ international hackathon wins (including Singapore–India Hackathon)
• Ex-Jefferies Technical Associate (backend + microservices)

📧 [email protected] |
💼 LinkedIn |
🖥️ GitHub

• Built fully on Cloudflare’s developer platform as part of the internship application assignment.
• Shows enthusiasm for Cloudflare’s mission to build a better Internet — combining curiosity, engineering rigor, and design sense.