Methodology & sources

Why a rule engine, not AI?

Every result on PetHealth is produced by a deterministic rule engine: the symptom–disease–urgency mappings are written explicitly in data tables, and the engine applies weighted matching plus fixed escalation rules. That means:

• • Auditable: every recommendation traces back to specific data entries and rules.
• • Stable: identical input always yields identical output — no “AI hallucination”.
• • Zero marginal cost: build once, query forever — traffic never adds compute cost.

Where does the data come from?

Symptoms, disease mappings, breed profiles and toxin data are compiled from these public veterinary resources:

• • Merck 兽医手册（MSD Veterinary Manual）
• • VCA Animal Hospitals 知识库
• • AKC 品种健康资料
• • PetMD 公开内容

We only organize and structure; we don't make original medical claims. Every entry keeps its source context for veterinarian advisors to review item by item.

How is urgency determined?

The urgency level stacks three rule layers: ① red-flag symptoms (trouble breathing, seizures, inability to urinate…) trigger “emergency” directly; ② combination matching — your selected symptoms are weighted-matched against the disease library to set the base level; ③ context escalation — young age, symptom duration and the owner's severity rating raise the level by fixed rules. The “Why this urgency level” list on the result page shows every rule that fired.

Privacy promise

All computation happens locally in your browser. No sign-up, no login, no symptom data uploaded, no tracking cookies (until ads are integrated).