Shelf: Household DVD Collection Live
An invite-only web app for tracking a household's DVD, Blu-ray and 4K box sets season by season, with loans, a wishlist and an AI assistant that identifies a box set from a photo of its cover.
- Client
- My household (invitation only)
- Role
- Designer and developer
- Period
- Nov 2026 – now
- Focus areas
- Azure, D365 & DevOpsIntegration
- Tech
- Azure App ServiceAzure SQLAI assistantTMDB APIJavaScript
Features
The seasons grid
Every show has one row per season and one column each for DVD, Blu-ray and 4K. Each cell shows whether that season is owned and which set it comes from, and a gap can be clicked to add the set that fills it. Specials are listed separately and never count towards a show being complete.
Sets and discs
- Each physical release is recorded once, with its format, edition, region, shelf location and purchase details.
- A set can cover several seasons, so a "Seasons 1–3" box and three single seasons both work.
- Individual discs can be tracked as present, missing or damaged.
Loans and the wishlist
- Loans record who has what and when it's due back. Overdue loans are flagged on the dashboard.
- The wishlist holds the seasons and formats we're after. Adding a set ticks its seasons off automatically.
The AI assistant
Take a photo of a cover in a shop and the assistant identifies the show, season and format, then says whether we already own it and what's still missing. It answers typed questions about the collection too, such as "Which Frasier seasons don't we have on Blu-ray?"
The assistant is built on an Azure OpenAI instance and the solution liaises with the TMDB API as the knowledge source of DVDs. It proposes changes as cards, and nothing changes until someone presses Save.
Our own details, not just TMDB's
Show details, seasons, episodes and posters come from TMDB. We can rename a title or season, choose a different poster, or hide a season we'll never collect. Refreshing from TMDB picks up new seasons without overwriting those changes.
Catalogued from photos
The first collection was put together from photos of some shelves of DVDs: about 60 titles identified from 10 photos of the spines of the DVDs, so the whole collection was catalogued in minutes.