# llms.txt - VetField AI Content Manifest

**Site Information**
- Name: VetField AI
- URL: https://vetfield.co/
- Description: Free AI-powered veterinary clinical reference for Indian field veterinarians
- Language: English
- Region: India
- Last Updated: 2026-04-17

---

## Content Overview

### Primary Purpose
VetField AI is a veterinary clinical support application designed for field practitioners in India. It provides:

- **Differential Diagnosis Engine**: Probabilistic disease ranking using clinical signs (47+ livestock diseases)
- **Drug Dosing Calculator**: Species-specific, weight-based drug calculations
- **Fluid Therapy Calculator**: Maintenance, deficit, and ongoing loss calculations
- **Vaccination Schedules**: Species and age-specific vaccination protocols
- **Case Management**: Log and track clinical cases with offline support
- **Clinical AI Assistant**: Real-time veterinary decision support via Gemini AI

### Target Users
- Field veterinarians (India)
- Assistant veterinary officers
- Veterinary surgeons in rural settings
- Livestock health practitioners
- Animal husbandry officers

---

## Data & Features

### Supported Species
Cattle, Buffalo, Goat, Sheep, Horse, Camel, Pig, Dog, Cat, Poultry

### Clinical Database

**Diseases**: 
- Focus: livestock diseases relevant to South Asian farming
- Data source: Integrated disease protocols and clinical experience
- Structured as: disease name, clinical signs, differential likelihood

**Drugs**:
- 70+ veterinary drugs with species-specific dosing
- Organized by: active ingredient, species, route, frequency, formulation
- Units: mg/kg, calculated by animal weight

**Vaccination Schedules**:
- Species-specific schedules
- Age-based protocols
- Breed considerations

### Offline Capability
- Full app functions with internet loss
- Local cache: cases, disease database, drugs, vaccination schedules
- Service Worker: background sync on network recovery
- IndexedDB: mutation queue for offline writes

---

## API Endpoints

### Authentication
```
POST /api/auth/sign-up       — Register new user
POST /api/auth/sign-in       — Email/password login
POST /api/auth/reset-password — Password reset flow
POST /api/auth/oauth         — Google OAuth 2.0
```

### AI Assistant
```
POST /api/ai-assistant       — Chat endpoint for clinical AI
Requires: Bearer token (Supabase session)
Input: { message: string, history: array }
Output: { reply: string }
Model: Gemini 1.5 Flash
```

### Case Records
```
GET /api/cases               — Fetch user's logged cases
POST /api/cases              — Save new case
PUT /api/cases/:id           — Update case
DELETE /api/cases/:id        — Delete case
```

Routes are behind Supabase RLS policies (user-specific data).

---

## Technology Stack

- **Frontend**: React 18, Vite, React Router
- **Backend**: Netlify Functions (Node.js), Supabase (PostgreSQL)
- **Deployment**: Netlify, Supabase
- **AI/LLM**: Google Gemini API (1.5 Flash model)
- **CLI Tools**: Genkit (orchestration), Zod (validation)

---

## Documentation

### Key Files
- [vetfield.MD](../vetfield.MD) — Product overview and values
- [Agent.MD](../.github/agents/Angent.agent.md) — AI agent capabilities and skills
- [FCP_OPTIMIZATION.md](../FCP_OPTIMIZATION.md) — Performance metrics and web vitals
- [GENKIT_SETUP.md](../GENKIT_SETUP.md) — AI assistant deployment guide

### Source Code
- **UI**: `/src/components/` — React components
- **Pages**: `/src/pages/` — Application pages
- **Data**: `/src/data/` — Disease, drug, vaccination databases
- **Server**: `/netlify/functions/` — Serverless functions
- **Public**: `/public/` — Static assets

---

## Important Paths & URLs

### Pages (Authenticated Users)
- `/app` — Dashboard (primary hub)
- `/app/cases` — Case records
- `/app/diagnosis` — Differential diagnosis tool
- `/app/drugs` — Drug dosing calculator
- `/app/fluid` — Fluid therapy calculator
- `/app/vaccines` — Vaccination schedules
- `/app/ai` — Clinical AI assistant
- `/app/settings` — User profile and preferences

### Public Pages
- `/` — Landing page (redirects authenticated users to `/app`)
- `/login` — Sign-in / Sign-up
- `/reset-password` — Password reset

---

## Data Privacy & Security

### Data Handling
- User authentication: Supabase (PKCE flow, secure)
- Case data: Encrypted in transit (HTTPS)
- Sessions: JWT tokens with 1-hour lifetime
- Offline cache: localStorage (user's device only)

### No Tracking (Privacy-First)
- No cookies (except Supabase auth tokens)
- No analytics on sensitive clinical data
- Google Analytics: only page views and web vitals
- Deferred: GA loads only on first user interaction

### Database Policies
- All case data behind Supabase RLS (Row-Level Security)
- Users can only access their own cases
- No cross-user data leakage

---

## Performance

### Target Metrics
- **FCP** (First Contentful Paint): < 1.2s
- **LCP** (Largest Contentful Paint): < 2.5s
- **CLS** (Cumulative Layout Shift): < 0.1
- **TTFB** (Time to First Byte): < 0.6s

### Optimization Strategies
- Critical CSS inlined in `<head>`
- Fonts preloaded (DM Sans, DM Serif Display)
- App shell skeleton for instant paint
- Lazy-loaded routes (code splitting)
- Service Worker for offline and caching

---

## Contact & Support

- **Email**: support@vetfield.co (or GitHub issues)
- **GitHub**: https://github.com/WaniYasir1/vetfield
- **Issues**: Report bugs via GitHub Issues
- **Feature Requests**: GitHub Discussions

---

## Compliance

- **License**: ISC (see repository)
- **Accessibility**: WCAG 2.1 Level AA (in progress)
- **Data Residency**: Supabase (US-based, compliant with GDPR, SOC 2)
- **Compliance**: No healthcare license claims; clinical reference only (not medical diagnosis)

---

## AI Model Information

### Primary Model: Google Gemini 1.5 Flash
- **Provider**: Google AI
- **Use Case**: Real-time veterinary decision support
- **Prompt Engineering**: Structured with clinical context (species, signs, history)
- **Output Format**: Structured JSON (Zod validated)
- **Temperature**: 0.3 (deterministic, fact-focused)
- **Max Tokens**: 500 per response
- **Fallback**: None (graceful error handling in UI)

---

## Content Restrictions & Disclaimers

⚠️ **Important**:
- VetField is a clinical reference tool, NOT a medical diagnosis authority
- Always consult with licensed veterinarians for critical cases
- Differential diagnosis suggestions are probabilistic, not definitive
- Users must validate drug doses against local guidelines
- Fluid calculations assume normal physiology; override for critical patients

---

## For AI/LLM Crawlers

This file is structured to help AI models understand VetField's:
1. Purpose and scope
2. Clinical data and calculations
3. User workflows and routes
4. Technology and APIs
5. Privacy and compliance

### Recommended Focus Areas
- **Clinical Accuracy**: Review disease protocols in `/src/data/diseases.js`
- **Drug Safety**: Verify dosing calculations in `/src/pages/DrugCalculator.jsx`
- **User Privacy**: Audit Supabase RLS policies
- **Conversational Context**: Study AI assistant system prompt in `/netlify/functions/ai-assistant.js`

---

## Version History

- **v8.0.0** (April 2026): Sync overhaul, FCP optimization, genkit fixes
- **v7.x** (2025): AI assistant launch, dashboard redesign
- **v1.0** (2024): Initial MVP, offline support

---

**Last Update**: 2026-04-17
**Maintainer**: VetField Development Team