Scanner intelligence
A transparent, calm engine for product analysis
LAFYA identifies products, evaluates ingredient quality and exposure patterns, and generates educational, non-alarmist insights — with reasoning you can always inspect.
1 · Scan entry methods
Barcode
UPC / EAN lookup
Label OCR
Read ingredient lists
Photo upload
Front-of-pack image
Live camera
Real-time framing
- Auto-focus guidance and edge detection overlay
- Low-light and blur warnings before capture
- Calm retry assistance — never alarmist
2 · Processing pipeline
- 1
Capture product
- 2
Barcode / OCR recognition
- 3
Ingredient extraction
- 4
Ingredient normalization
- 5
Intelligence matching
- 6
Risk categorization
- 7
Safety grade calculation
- 8
Personalized insight
- 9
Cleaner alternatives
3 · Ingredient intelligence schema
4 · Risk classification
Green · Low concern
Whole foods, minerals, minimally processed ingredients
Yellow · Moderate concern
Additives, gums, processed sweeteners, stabilizers
Red · Elevated concern
Synthetic dyes, controversial preservatives, contamination indicators
5 · Safety grade engine
Grades start at 100 and decrease by weighted contribution from each factor.
6 · AI insight structure
- Educational, supportive, never medical
- One primary insight per screen
- Maximum five concern highlights
7 · Accuracy & trust signals
8 · Edge cases & fallback
Blurry capture
Prompt rescan in brighter lighting.
Incomplete label
Show partial result with low-confidence flag.
Unknown product
Offer manual ingredient entry.
Foreign-language label
Auto-translate before extraction.
Barcode mismatch
Verify with OCR before scoring.
9 · UX principles
- Calm scientific tone — never fear-based
- Progressive disclosure for deeper analysis
- Readability over data density
- Always explain reasoning behind any score
- Respect user agency — suggest, never prescribe
Have questions about how a specific score was calculated? Tap the "How it works" badge on any scan result to see the contributing factors.
Maintenance
Re-run AI classification (PubChem / EFSA / FooDB / EWG) on ingredients that were previously labeled by the legacy regex.

