⚡ Fully offline · No server needed
Patient profile
Describe symptoms in plain English
0/1400
Try:
Enter symptoms above and click Analyse.
No file linked. Use "Link file" to auto-save history to a local JSON file.
🐾

No entries yet.
Run a screening and save it — it will appear here.

Fully offline NLP + ANN pipeline
1

NLP — Keyword & Phrase Extraction

Your free-text description is scanned by a built-in JavaScript NLP engine. It tokenises the text, applies negation detection ("no coughing", "not limping"), intensifier weighting ("very", "extremely", "a little"), and matches 300+ symptom phrases across 18 symptom dimensions (skin, eyes, behaviour, mobility, respiratory, GI). No external API — runs entirely in your browser.

2

Breed-aware feature engineering

Extracted scores (0–10) are normalised and combined with breed one-hot encoding (12 breeds), age, weight, estimated temperature, appetite and exposure flags into a 35-dimensional feature vector.

3

In-browser ANN (TensorFlow.js)

A 4-layer neural network (48→28→16→10) trains on 1,000 synthetic breed-stratified samples in your browser using TensorFlow.js loaded from CDN. Softmax output gives probabilities across 10 condition categories.

4

Rule-based triage safety layer

Runs independently of the ANN. Flags critical thresholds — breathing difficulty ≥8, temp ≥40.5°C, bloat ≥8, combined severe GI + lethargy, eye pain ≥8 — as Urgent regardless of model output.

5

Local file persistence

History saves to localStorage automatically. Use "Link file" (Chrome/Edge) to also write to a real .json file on your disk — survives browser resets and can be backed up or shared.

⚠️ Educational screening only — not a veterinary diagnosis. For severe breathing difficulty, collapse, seizures, suspected poisoning, or eye injury seek veterinary care immediately.
✅Saved.