Case Study

Miniatures Cataloger

A Django app built to help a client resell a 16,000-piece miniatures collection on eBay. It identifies a figure from the number stamped on its base, matches it against a 1,140-figure reference catalog I built, and turns the result into a ready-to-post listing. Built to make a slow, expert-knowledge resale workflow fast and repeatable.

Django Data Scraping Reference Database eBay API OAuth

The Problem

A client has a collection of roughly 16,000 pre-painted D&D and Pathfinder miniatures to sell on eBay. Listing one by hand means photographing it, identifying which set it came from and its figure number, researching a fair price against completed sales, writing a description, and filling out the eBay form: about 10 to 15 minutes each. Across the whole collection that is on the order of 3,200 hours of work, with no system tracking status, condition, or pricing.

16,000

miniatures to list

10-15 min

per item, by hand

~3,200 hrs

total manual effort

The Solution

A full-stack Django app with a mobile-friendly capture flow. Photograph a figure, read the number off its base to find it in the reference catalog, confirm the match, then build and post the eBay listing without leaving the app.

Base-Label Lookup

Every figure has a number stamped on its base, like "23/60". The app reads that number and looks it up in the catalog, using the set total (the "/60") to narrow what would otherwise be dozens of matches down to the right set. It is fast, deterministic, and free, which beat image recognition on every count for this job.

Self-Built Reference Catalog

The lookup only works because of the data behind it: roughly 1,140 known figures, assembled by scraping public set checklists (minisgallery, miniscollector, Wikipedia) and reconciling them against each other. A separate scraper backfills a reference photo for each entry so matches can be confirmed by eye.

Listing Builder

Single figures or multi-item bundles. The app auto-generates an eBay-compliant title (capped at 80 characters) and a description from the catalog data, and offers a one-click "completed sales" search to set a fair price. Everything stays editable before posting.

Direct eBay Posting

Listings post straight to eBay through the Inventory API using OAuth, with a sandbox/live toggle for safe testing, daily API-usage tracking, and seller policy IDs stored in the database so credentials can be updated without redeploying.

Where does AI come in? It is an optional fallback, toggleable in settings, for the occasional figure whose base label is worn or missing. When it is off, those go straight to manual entry. The deliberate decision here was that a catalog I controlled was more accurate, cheaper, and more reliable than leaning on image recognition for the common case.

How It Fits Together

Phone Capture
Photograph figure

Django App

Look up + review + list

SQLite DB
Catalog + collection

Catalog Scrapers

Checklists + images

eBay Inventory API

OAuth + posting

Gemini

Optional fallback

Each physical figure moves through a clear status flow (draft, needs review, manual entry, ready) before it can be added to a listing, so the collection is always in a known state.

Inside the App

DnD Mini Lister ◆ Review

All Needs Review 0 Manual Entry 1 Draft 0

Item	Set	Condition	Status	Action
Beholder Lich	Unhallowed #31	Near Mint	Ready
Young Green Dragon	Dragon Collector's Set #3	Near Mint	Manual Entry	Review
Frost Giant	Tyranny of Dragons #27	Near Mint	Ready
Faceless Stalker	Rise of the Runelords #6	Near Mint	Ready
Roper	Underdark #56	Near Mint	Ready

DnD Mini Lister ◆ Young Green Dragon

← Back to queue

Front Photo

Bottom (base label)

Name

Young Green Dragon

Set name

Dragon Collector's Set

Figure number ?

Rarity

not set

Search catalog ?

Condition

Near Mint

Pricing research View sold comps on eBay →

AI price estimate

Mark as Ready Save changes

Reproduced from the app's own templates and styling, using real items from the database. The Young Green Dragon's base label reads 3 / 5: that number is the catalog key the lookup turns into a full identification.

Impact

Turned identification into a look-up-and-confirm step, removing the need to research collector-level details for every figure by hand
Built a reusable 1,140-figure reference catalog from scratch by scraping and reconciling multiple public sources
Chose a controlled database over image recognition for the common case: cheaper, more accurate, and not dependent on a paid API, with AI kept as a fallback
Generates compliant eBay titles and descriptions automatically, with pricing backed by real completed-sale data
Posts directly to eBay through the Inventory API, with a sandbox mode so the whole flow can be tested without listing anything for real