Shaggy Dog Web Application (MCO368 Final)

This project satisfies the requirements:

• Upload a headshot
• Identify closest dog breed
• Generate 2 transition images + final dog image (3 generated images total)
• Show images in the UI, generated concurrently using multithreading
• User registration/login with encrypted credentials (bcrypt hashes)
• Store user credentials + generated images in a database
• Each user can only see their own images

Local dev

1. Create a virtualenv and install deps:

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2. Create .env:

cp .env.example .env
# then edit:
# OPENAI_API_KEY=...
# SECRET_KEY=...

3. Run:

uvicorn app.main:app --reload

Open http://127.0.0.1:8000

Smoke test (optional)

python -m app.smoketest

Deploy to Render

Create:

• A Web Service for this repo
• A PostgreSQL instance

Set env vars on the web service:

• OPENAI_API_KEY
• SECRET_KEY (any long random string)
• DATABASE_URL (Render provides this for Postgres)
• VISION_MODEL (optional, default: gpt-4.1-mini)
• IMAGE_MODEL (optional, default: gpt-image-1.5; fallback to dall-e-2 on errors)

Build command:

pip install -r requirements.txt

Start command:

uvicorn app.main:app --host 0.0.0.0 --port $PORT

Custom domain (subdomain)

In Render service settings: add shaggydog.<yourdomain> as a custom domain. Then add a CNAME DNS record for shaggydog pointing to your service’s *.onrender.com hostname.

Render automatically provisions TLS (HTTPS) after verification.

Notes on OpenAI image models

The app prefers GPT Image (gpt-image-1.5) for higher quality. Some orgs may need verification to use GPT Image models. If the call fails, the app automatically falls back to dall-e-2 edits (works but lower quality and requires square PNG input).