Skip to content

pranjalbhatia710/trashmytech

BranchesTags

Repository files navigation

trashmy.tech

Paste a URL. 12 AI personas stress-test your site in 60 seconds. Get a scored report with exactly what to fix.

trashmy.tech

What is this?

We ship fast. But who's actually testing? trashmy.tech lets you paste any URL and instantly see how real users would experience it — accessibility issues, broken flows, SEO gaps, security holes, and content problems.

One free analysis per visitor. $5 one-time payment for unlimited.

How it works

  1. Paste a URL, click "Trash It"
  2. 12 AI personas launch in parallel — each a real Playwright browser instance. A blind user on a screen reader. A skeptic looking for trust signals. Googlebot checking crawlability. A chaos agent breaking forms.
  3. Scoring engine calculates a deterministic 0-100 score across 6 categories
  4. Report generates with annotated screenshots, ranked fixes, and a copy-paste prompt for ChatGPT/Claude to implement the fixes

Results are deterministic — same site always gets the same score. Reports are cached for 7 days.

Features

  • 40 persona types across 7 categories: accessibility, demographic, security, usability, chaos, content/SEO
  • 6-category scoring: Accessibility (25%), SEO (20%), Performance (20%), Content (15%), Security (10%), UX (10%)
  • Hard caps: No HTTPS = max 40/100. Critical a11y failure = max 50 accessibility score
  • Lite mode: Near-zero API cost using only free checks (SSL, DNS, robots.txt, axe-core)
  • Full mode: Google PageSpeed, Mozilla Observatory, Safe Browsing integration
  • Real-time streaming: WebSocket updates as agents crawl, test, and find issues
  • Auth + payments: NextAuth.js (Google OAuth), Stripe ($5 one-time)
  • Caching: Redis + PostgreSQL, 7-day TTL, domain normalization for deterministic results

Tech stack

Layer Tech
Frontend Next.js 16, React 19, Tailwind CSS v4, Framer Motion, Three.js, Recharts
Backend FastAPI, WebSockets, asyncio
Browser agents Playwright (local or Modal for scaling)
AI Gemini 2.0 Flash (personas, report generation)
Accessibility axe-core
Auth NextAuth.js, JWT, Google OAuth
Payments Stripe (one-time $5)
Database PostgreSQL (Supabase/Neon), Redis (Upstash)
Deploy Vercel (frontend), Render/Railway (backend)

Getting started

Prerequisites

1. Clone and install

git clone https://github.com/pranjalbhatia710/trashmytech.git
cd trashmytech

# Frontend
npm install

# Backend
cd backend
pip install -r requirements.txt
playwright install chromium

2. Configure environment

Backend — create backend/.env:

GEMINI_API_KEY=your-gemini-key
ANALYSIS_MODE=lite
HEADLESS=true
AGENT_COUNT=12

# Optional (for persistence):
# DATABASE_URL=postgresql://user:pass@host:5432/trashmytech
# REDIS_URL=redis://default:token@host:6379

# Optional (for auth):
# JWT_SECRET=your-random-secret
# STRIPE_SECRET_KEY=sk_test_...
# STRIPE_WEBHOOK_SECRET=whsec_...
# STRIPE_PRICE_ID=price_...

Frontend — create .env.local:

NEXT_PUBLIC_API_URL=http://localhost:8000

# Optional (for auth):
# NEXTAUTH_URL=http://localhost:3000
# NEXTAUTH_SECRET=your-random-secret
# GOOGLE_CLIENT_ID=your-google-client-id
# GOOGLE_CLIENT_SECRET=your-google-client-secret
# NEXT_PUBLIC_STRIPE_PUBLISHABLE_KEY=pk_test_...

3. Run

# Terminal 1 — Backend
cd backend
uvicorn main:app --reload --port 8000

# Terminal 2 — Frontend
npm run dev

Open http://localhost:3000, paste a URL, hit Trash It.

Minimum setup

Only GEMINI_API_KEY is required. Everything else (database, Redis, auth, Stripe) is optional and degrades gracefully — the app works without them.

Architecture

┌─────────────────────────────────────────────────────┐
│  Browser (localhost:3000)                            │
│  Next.js + React + Tailwind + Framer Motion          │
└──────────┬──────────────────────────────┬───────────┘
           │ REST                         │ WebSocket
           ▼                              ▼
┌─────────────────────────────────────────────────────┐
│  FastAPI Backend (localhost:8000)                     │
│                                                      │
│  POST /v1/tests         → create analysis            │
│  WS   /ws/{test_id}     → stream live progress       │
│  GET  /v1/tests/{id}    → get results                │
│  GET  /v1/recent        → recently analyzed sites     │
│                                                      │
│  Pipeline:                                           │
│  ┌──────────┐  ┌──────────┐  ┌────────┐  ┌───────┐  │
│  │ 1. Crawl │→ │ 2. Swarm │→ │3.Score │→ │4.Save │  │
│  │Playwright│  │12 agents │  │  Pure   │  │PG+Redis│ │
│  │ axe-core │  │  Gemini  │  │functions│  │       │  │
│  └──────────┘  └──────────┘  └────────┘  └───────┘  │
│                                                      │
│  External APIs (full mode only):                     │
│  PageSpeed · Observatory · Safe Browsing · SSL · DNS │
└─────────────────────────────────────────────────────┘

Analysis modes

Mode Cost What runs Use case
lite ~$0.01/analysis Playwright + axe-core + Gemini Flash + free checks (SSL, DNS, robots.txt) Development, demos
full ~$0.05/analysis Everything above + Google PageSpeed + Mozilla Observatory + Safe Browsing Production

Set via ANALYSIS_MODE=lite or ANALYSIS_MODE=full in backend/.env.

Scoring

All scores are deterministic — pure functions, no AI calls, no randomness. Same input = same output.

Category Weight Data sources
Accessibility 25% axe-core violations, Lighthouse a11y, persona testing
SEO 20% Meta tags, structured data, robots.txt, sitemap, AI bot access
Performance 20% PageSpeed (full) or agent-measured load times (lite)
Content 15% Readability, trust signals, value proposition clarity
Security 10% SSL, DNS, headers, Observatory (full), HTTPS check
UX 10% Persona task completion, mobile testing, form usability

Hard caps:

  • Site flagged by Safe Browsing → overall max 15
  • No HTTPS → overall max 40
  • Critical accessibility failure → accessibility max 50

Deployment

Frontend → Vercel

Push to GitHub. Vercel auto-deploys from main branch. Set env var:

NEXT_PUBLIC_API_URL=https://your-backend-url.onrender.com

Backend → Render

  1. New Web Service → connect GitHub repo
  2. Root directory: backend
  3. Runtime: Docker
  4. Add env vars: GEMINI_API_KEY, HEADLESS=true, ANALYSIS_MODE=lite, AGENT_COUNT=12

Database → Supabase (optional)

Free tier (500MB). Set DATABASE_URL on Render. Migrations run automatically on startup.

Cache → Upstash (optional)

Free tier (10K commands/day). Set REDIS_URL on Render.

API reference

Endpoint Method Description
/v1/tests POST Create analysis {"url": "...", "mode": "standard"}
/ws/{test_id} WS Stream live progress
/v1/tests/{test_id} GET Get test status + report
/v1/tests/{test_id}/report GET Get report only
/v1/recent?limit=6 GET Recently analyzed sites
/v1/stats GET Aggregate stats
/v1/health GET Health check
/v1/auth/register POST Register (email/password)
/v1/auth/login POST Login, returns JWT
/v1/auth/me GET Current user (requires JWT)

Project structure

trashmytech/
├── app/                    # Next.js pages
│   ├── page.tsx            # Landing page
│   ├── test/[id]/page.tsx  # Live dashboard + report
│   ├── api/auth/           # NextAuth routes
│   └── api/stripe/         # Stripe checkout + webhook
├── components/             # React components
│   ├── url-input.tsx       # URL input with validation
│   ├── auth-modal.tsx      # Auth + payment modal
│   └── ui/                 # Score gauge, toast, counters
├── lib/                    # Config, auth helpers
├── backend/
│   ├── main.py             # FastAPI + WebSocket orchestration
│   ├── pipeline.py         # 4-phase pipeline (crawl→swarm→score→persist)
│   ├── agent.py            # Gemini-powered persona engine
│   ├── scoring.py          # Deterministic scoring (pure functions)
│   ├── report.py           # Report generation
│   ├── personas.py         # 35 base personas
│   ├── personas_content_seo.py  # 5 content/SEO personas
│   ├── crawler.py          # Playwright + axe-core
│   ├── external_apis.py    # PageSpeed, Observatory, etc.
│   ├── analysis_lite.py    # Lite mode config
│   ├── quick_wins.py       # Fix recommendations
│   ├── auth/               # User auth + Stripe webhooks
│   ├── db/                 # PostgreSQL schema + queries
│   ├── cache/              # Redis caching layer
│   └── services/           # Persistence service
└── render.yaml             # Render deployment config

Built at HackIllinois 2026

Solo project. Started 15 hours into the hackathon after the original team dropped out. Won first place.

License

MIT

About

AI Website Stress Testing — 50 AI personas test your site in 60 seconds. Built at HackIllinois 2026.

trashmytech.vercel.app

Resources

Readme
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages