ThinkOva

Ask a question. Watch the web teach you.

Image disclaimer: The banner above was generated using an AI image generation tool for presentation purposes only.

---

What is it?

ThinkOva is a visual learning tool that takes any topic you type, opens a real browser, navigates the web automatically, and narrates what it's doing — step by step, out loud.

No links. No walls of text. Just watch and listen.

---

How it works

1. You type a topic and hit search
2. An AI agent (Nova LLM on AWS Bedrock) breaks it into a step-by-step plan
3. Nova Act executes each step in a real headless browser
4. Live screenshots stream to your screen at 10fps
5. ElevenLabs reads the narration out loud for each step
6. The next step only starts after the audio finishes — so you never fall behind

---

Stack

What	Why
Next.js	Frontend — search bar, live canvas, WebSocket
FastAPI + Redis	Backend — sessions, orchestration
AWS Bedrock (Nova LLM)	Generates the step plan
Nova Act	Drives the browser, streams screenshots
ElevenLabs TTS	Narrates each step out loud

---

Architecture

---

Running it locally

Backend

cd apps/backend
cp .env.example .env    # add your keys
docker-compose up

Frontend

cd apps/frontend
npm install

Create apps/frontend/.env.local:

NEXT_PUBLIC_ELEVENLABS_API_KEY=sk_yourkey
NEXT_PUBLIC_VOICE_ID=JBFqnCBsd6RMkjVDRZzb
NEXT_PUBLIC_WS_URL=ws://localhost:8080

npm run dev

Open http://localhost:3000

---

Environment variables

Backend (apps/backend/.env)

NOVA_ACT_API_KEY=
AGENT_RUNTIME_ARN=
AWS_REGION=
REDIS_URL=
LOCAL_DEV=1

Frontend (apps/frontend/.env.local)

NEXT_PUBLIC_ELEVENLABS_API_KEY=
NEXT_PUBLIC_VOICE_ID=
NEXT_PUBLIC_WS_URL=

Never commit either of these files. Both are gitignored.

---

Project structure

visualizer/
├── apps/
│   ├── agent/          # Nova LLM agent + prompts
│   ├── backend/        # FastAPI + Redis + Nova runner
│   └── frontend/       # Next.js app + ElevenLabs TTS

---

Built with

Nova Act · AWS Bedrock · ElevenLabs · Next.js · FastAPI