prajwalmhetre/medilens-ai

Purpose and entry points

MediLens AI is a full-stack healthcare demo for medicine scanning, medication tracking/reminders, profile-based safety guidance, and simulated AI-pharmacist chat. Flutter is the multi-platform client; FastAPI exposes the REST backend. Primary entry points are lib/main.dart, lib/app/medilens_app.dart, and backend/main.py.

Flutter composition and navigation

The application bootstrap installs app-wide composition, Material styling, and provider-backed state before presenting onboarding and the main tab shell. The shell hosts dashboard, scanning, reminders, chat, and profile presentation modules.

Paths: lib/app/medilens_app.dart, lib/features/shell/presentation/main_shell.dart, lib/features/onboarding/presentation/splash_screen.dart.

Global state and offline boundary

AppStateProvider is the client-side orchestration layer: it owns loaded/optimistic UI state and coordinates profile, medicines, reminders, and conversations. The README specifies offline-first fallback behavior, notably local simulated chat responses when API calls fail; feature screens should remain presentation-focused.

Paths: lib/app/state/app_state_provider.dart, lib/shared/models/medicine.dart, lib/shared/models/reminder.dart.

API integration boundary

Flutter feature/state code reaches the backend only through the HTTP service. Its base URL is runtime-configured through API_BASE_URL, allowing native development, web builds, and container deployments to target different API hosts.

Paths: lib/core/network/api_service.dart, .env.example, pubspec.yaml.

User-facing feature stages

Dashboard renders medication/adherence status; scan simulates medicine identification and allergy-conflict guidance; medicine details expose a selected medicine; reminders support creation and completion state; chat submits and displays medication guidance. These are separate feature presentation areas sharing state and model contracts.

Paths: lib/features/dashboard/presentation/dashboard_tab.dart, lib/features/scan/presentation/scan_tab.dart, lib/features/reminders/presentation/reminders_tab.dart.

FastAPI service and AI workflow

The FastAPI application owns /api endpoints for seeded user data, medicines, reminders, and chat. Chat is a deterministic safety-guidance pipeline: persist user text, classify intent, produce an allergy/safety/scheduling/general response, persist it, then return both messages. It is mock AI today; AI_PROVIDER and OPENAI_API_KEY are future integration seams, not evidence of an active external LLM dependency.

Paths: backend/main.py, backend/schemas.py, backend/AI_IMPLEMENTATION.md.

Persistence and deployment

SQLAlchemy models and database setup isolate persistence from HTTP handlers. SQLite is the default local store, seeded at backend startup; DATABASE_URL permits replacement with a production database. Docker Compose runs the static Flutter web frontend and FastAPI backend together; backend CORS is environment-configured.

Paths: backend/database.py, backend/models.py, docker-compose.yml.

Runtime surface

Flutter/Dart targets Android, iOS, web, Linux, macOS, and Windows through generated platform runners. The material deployment path is Flutter web in the root Dockerfile, with Python/FastAPI/Uvicorn in backend/Dockerfile; no authentication, real OCR, push notification, or production LLM service is currently implemented.

Paths: Dockerfile, backend/Dockerfile, android/app/src/main/kotlin/com/prajwalmhetre/medilens_ai/medilens_ai/MainActivity.kt.