FigurePrice

한국어

One photo → figure ID + MSRP / Market prices.

Local desktop app for Apple Silicon. Not for public deployment.

Video

 

What it does

1. Upload a figure photo (file, camera, or drag & drop)
2. Match it against a local vector index (~12k figures from MFC)
3. Optionally re-rank with Gemini, or fall back to Gemini + Google Search
4. Show MSRP and partner asking prices in KRW

Quick start

Requirements: macOS (Apple Silicon), Python 3.11, Gemini API key (free tier OK)

brew install python@3.11
python3.11 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

cp .env.example .env
# set GEMINI_API_KEY=...

If data/figures.db is already present, run the app:

source .venv/bin/activate
python -m gui
# or: ./run.sh  /  double-click FigurePrice.command

Build the dataset (first time only)

Skip if you already have data/figures.db.

source .venv/bin/activate
python -m db.init_db
python -m crawler.fx_updater
python -m crawler.mfc_crawler --max-items 500   # raise or omit for full crawl
python -m embeddings.build_index                # ~1h for ~12k on MPS

Useful crawl flags: --refresh, --refresh-existing, --seeds URL...

Config (.env)

Variable	Default	Notes
GEMINI_API_KEY	(required)	Free key from AI Studio
GEMINI_MODEL	gemini-3.1-flash-lite	Re-rank
GEMINI_LOOKUP_SEARCH_MODEL	gemini-2.5-flash	Web lookup + Search (higher free-tier Search quota)
GEMINI_LOOKUP_MODEL	gemini-3.1-flash-lite	Offline lookup fallback
GEMINI_LOOKUP_USE_SEARCH	true	Set false to skip Search grounding
DATA_DIR	./data	DB + image cache

How it works

flowchart LR
  A[Photo] --> B[rembg]
  B --> C[DINOv2 · MPS]
  C --> D[sqlite-vec top-K]
  D --> E{Match?}
  E -->|yes| F[Gemini re-rank<br/>optional]
  E -->|miss| G[Gemini + Google Search]
  F --> H[MSRP + quotes → KRW]
  G --> H

Loading

Stack: PySide6 GUI · DINOv2 · sqlite-vec · Gemini · MFC crawl

Optional HTTP API (same pipeline): uvicorn api.main:app --port 8000 → docs at /docs

Troubleshooting

Issue	Fix
First identify is slow (~30–60s)	One-time DINOv2 / u2net download
torchvision import error	pip install torchvision
MFC 403/503	Wait, then raise MFC_REQUEST_DELAY_SEC in config.py
Gemini 429	Wait for RPM, or switch models in .env / try next day
Unique constraint on figure_images	python -m db.migrate_figure_images (once)

Notes

• Personal / research use only — do not redistribute crawled images
• Crawl respects a polite delay (MFC_REQUEST_DELAY_SEC)