Skip to content

Nebulyn-Labs/MediFlow

BranchesTags

Repository files navigation

MediFlow

AI-powered medical logistics platform focused on smart resource allocation

Flutter Firebase Gemini AI ORS

Table of Contents

Project Overview

MediFlow is an enterprise-grade medical logistics platform engineered to solve the "Last Mile" medical supply crisis. By combining Generative AI for demand forecasting with Heuristic Heuristics, we optimize redistribution of medical supplies across a network of urban and rural healthcare facilities.

The Problem | The Solution

The Crisis: Rural clinics often face 30% higher stockout rates for essential antibiotics, while urban hospitals simultaneously dispose of expired stock due to over-purchasing. This inequality claims lives.

The MediFlow Solution: We don't just track inventory; we predict shortages before they happen and automate the movement of medicine from surplus hospitals to deficit clinics using road-optimized routing.

Screenshots & UI Preview

Here is a preview of the MediFlow home page:

MediFlow home page

MediFlow provides two types of roles:

  • Role -> Admin

Here is a preview of Admin Dashboard Admin dashboard

  • Role -> Facility

Here is a preview of Facility manager Dashboard Facility dashboard

It also provides following features:

  • AI analysis of supply facility supply ai analytics

  • Route optimization for delivery via AI (Admin only) route optimization op

  • Demand forecast forecast of usage trend

Core Feature Set

Hospital / Facility Module

Feature Detailed Description
Smart Logging Engine Atomically track daily usage while the system computes burn rates in real-time, ensuring zero data loss even in low-connectivity areas.
AI Forecasting (30-Day) Powered by Gemini-1.5-Flash, predicting seasonal spikes based on historical usage trends (e.g., ORS demand for summer) with a transparency-first "AI Reasoning".
Automated Request Drafting Intelligent auto-population of restock indents and redistribution offers based on AI predictions, reducing administrative overhead for clinic managers.
AI Chat Assistant A 24/7 logistics expert that facility managers can query for stock status, expiry alerts, or burn-rate insights using natural language.

Central Administration Module

Feature Detailed Description
Global Command Center Real-time regional oversight with deep-dive analytics into every facility's stock health, parity, and regional logistics KPIs.
Approval Pipeline A secure hub for regional admins to review, edit, and prioritize redistribution plans proposed by the optimization engine.
Interactive Logistics Map High-visibility markers distinguishing surplus sites from deficit clinics with integrated OSRM/ORS paths that calculate real-world travel time and distance.
Global Optimization A "Global Redistribution Plan" that matches thousands of shortage items to local surpluses in seconds using our proprietary matching logic.

Technical Architecture

MediFlow utilizes a decoupled, serverless architecture that bridges a responsive frontend client with intelligent background processing.

Architecture Overview

graph TD
    A[Flutter Web / Mobile Client] -->|Auth & Live Sync| B[Firebase Auth & Firestore]
    A -->|Direct Call| C[Cloud Functions Node.js]
    C -->|Secret Key| D[Gemini 1.5 Flash API]
    C -->|Data Archiving| E[Google Cloud BigQuery]
    A -->|Geospatial Queries| F[OpenRouteService / OSRM API]
    B -->|Trigger Functions| C
Loading

Key Technical Pillars

  1. AI Engine (Gemini 1.5 Flash): We leverage Gemini's large context window to process months of anonymized usage logs. The model acts as a Predictive Reasoning Layer, identifying non-obvious redistribution opportunities.

  2. Optimization Heuristic (OTS): Our proprietary Optimal Transfer Score ensures that redistribution is both efficient and equitable:

    $$OTS = (w_{dist} \cdot Proximity) + (w_{prior} \cdot RuralPriority) + (w_{qty} \cdot QtyMatch)$$

    • Proximity: Minimizes logistics cost and time.
    • Rural Priority: A weight multiplier ensuring that remote facilities are never starved by the algorithm.

  3. Geospatial Routing System: Integrated with flutter_map and OSRM/OpenRouteService, our routing engine decodes complex polylines to provide precise, road-accurate delivery paths.

Project Structure

lib/
├── constants/
│   └── colors.dart             # Project-wide design tokens & premium palette
│
├── models/                     # Immutable Data Domain
│   ├── daily_usage_log.dart    # Atomic snapshots of medicine consumption
│   ├── facility.dart           # Metadata & Geospatial profiles for nodes
│   ├── inventory_item.dart     # Stock tracking & expiry metadata
│   ├── request.dart            # Ledger for redistribution & restock flows
│   └── usage_log.dart          # Helper models for analytics visualization
│
├── services/                   # Business Logic & Intelligence Layer
│   ├── ai_service.dart         # Gemini-1.5-Flash forecasting & reasoning
│   ├── chat_service.dart       # NLP pipeline for the AI Assistant
│   ├── firebase_service.dart   # Firestore infrastructure & transactions
│   ├── optimization_service.dart # OTS heuristic & matching algorithm
│   ├── routing_service.dart    # Geospatial OSRM/ORS pathfinding logic
│   ├── simulation_service.dart # Real-time demo data generation engine
│   └── tool_dispatcher.dart    # AI tool-calling & data registry
│
├── views/                      # Presentation Layer (UI)
│   ├── admin/                  # Central Command Module
│   │   ├── admin_indent_approval_page.dart
│   │   ├── admin_indent_status_page.dart
│   │   ├── admin_overview.dart
│   │   └── route_optimization_map.dart
│   │
│   ├── auth/                   # Security & Role Gatekeeping
│   │   ├── login_screen.dart
│   │   └── role_selection_screen.dart
│   │
│   ├── facility/               # Local Management Module
│   │   ├── active_indents_page.dart
│   │   ├── ai_forecast_page.dart
│   │   ├── alerts_page.dart
│   │   ├── daily_logging_page.dart
│   │   ├── facility_overview.dart
│   │   └── indent_creation_page.dart
│   │
│   └── shared/                 # Common & Reusable Components
│       ├── ai_chat_page.dart
│       ├── help_page.dart
│       └── sidebar_layout.dart
│
├── firebase_options.dart       # Cross-platform Firebase configuration
└── main.dart                   # Application entry & Router configuration

Data & Schema

MediFlow utilizes a hierarchical Firestore schema designed for high-concurrency performance:

  • /facilities: Metadata, type (urban/rural), and geospatial coordinates.
  • /inventory/{fac_id}/medicines: Sub-collection tracking individual batches and live stock levels.
  • /requests: Global collection for tracking movement, status (Pending/Approved/Fulfilled), and manifest details.

🔑 Demo Credentials

Role Email ID Password Facility Type
CMS Admin admin@mediflow.com password123 Central Management
Facility Head rampur@mediflow.com password123 Rural PHC
Facility Head modinagar@mediflow.com password123 Urban CHC
Facility Head loni@mediflow.com password123 Urban PHC
Facility Head ghaziabad@mediflow.com password123 District Hospital
Facility Head bhojpur@mediflow.com password123 Rural PHC
Facility Head hapur@mediflow.com password123 Urban CHC
Facility Head dasna@mediflow.com password123 Rural PHC
Facility Head pilkhuwa@mediflow.com password123 Rural Sub-Centre

Note

If the dashboard appears empty, use the "Seed DB" button on the Login/Role Selection screen to populate the database with demo records.

Development & Setup Guide

Prerequisites

  • Flutter SDK (>=3.0.0, recommended 3.41.x)
  • Node.js (for Firebase Cloud Functions deployment)
  • Firebase Project configured on Firebase Console
  • Google AI Studio API Key (for Gemini)
  • OpenRouteService API Key (for route calculations)

Local Setup Steps

  1. Clone the Repository:

    git clone https://github.com/Nebulyn-Labs/MediFlow.git
cd MediFlow

  2. Fetch Dependencies:

    flutter pub get

  3. Configure the Environment:

    Create a .env file in the root of the project (and copy from .env.example if needed):

    GEMINI_API_KEY=your_gemini_api_key
ORS_API_KEY=your_openroute_service_key
FIREBASE_PROJECT_ID=mediflow-92e6f

  4. Initialize Firebase CLI (if modifying functions):

    npm install -g firebase-tools
firebase login
firebase use mediflow-92e6f

  5. Run the Application:

    For local web prototype development, run:

    flutter run -d chrome

  6. Configure Firebase App Check (Important):

    This project uses Firebase App Check to protect backend resources.

    • Android: Play Integrity is used in release builds, and debug mode uses a Debug token. You must register your app's SHA-256 certificate in the Firebase Console under App Check.

    • Web: We use reCAPTCHA v3. Provide your site key during build or run time:

      flutter run -d chrome \
  --dart-define=RECAPTCHA_SITE_KEY=your_recaptcha_site_key

      Note: If no key is provided, the app will continue to run, but App Check will remain inactive for web. If App Check enforcement is enabled in the Firebase Console, requests may be rejected.

Troubleshooting

1. asset_does_not_exist Error for .env

If compilation fails with an error indicating .env does not exist:

  • Make sure you created a .env file in the root folder.

  • If you don't have API keys yet, you can create a placeholder .env with dummy values:

    GEMINI_API_KEY=dummy_key
ORS_API_KEY=dummy_key

2. Web Map Fails to Render (CORS or Rendering mode)

If the map does not display in your web browser:

  • If the CanvasKit renderer encounters issues fetching tiles from OpenStreetMap due to CORS, check your OpenRouteService API key and network settings. Ensure all Firebase dependencies are up to date.
  • Verify that your OpenRouteService API key is valid and has not exceeded its request limit.

3. Cloud Functions Authentication Errors

If calling Gemini forecasts yields an authentication error:

  • Ensure you are logged in to the application.

  • Verify that you have deployed the Cloud Functions and configured secrets:

    firebase functions:secrets:set GEMINI_API_KEY="your_actual_key"

Future Developments

  • Offline-First Sync: Native SQLite integration for zero-connectivity environments.
  • Batch Tracking: QR-code integration for granular tracking of individual medicine strips.
  • IoT Cold Chain: Integration with sensors to track temperature-sensitive vaccines during transit.

Contributing

We welcome contributions! Please review our CONTRIBUTING.md and CODE_OF_CONDUCT.md before submitting pull requests.

License

Distributed under the MIT License. See LICENSE for details.

The Team

Built with ❤️ by the MediFlow Team:

  • Aarush Yadav
  • Paavni Bansal
  • Devansh Rana
  • Sharvi Singhal

© 2026 MediFlow Team. Engineering a smarter, healthier supply chain.

About

MediFlow is an AI-powered medical logistics platform focused on smart resource allocation ensuring medicines are efficiently distributed based on real-time needs. By combining predictive forecasting, intelligent alerts, and optimized redistribution plans, it minimizes wastage and prevents stockouts, creating a proactive healthcare supply chain.

mediflow-92e6f.web.app

Resources

Readme
MIT license

Code of conduct

Code of conduct

Contributing

Contributing

Security policy

Security policy
Activity
Custom properties

Stars

3 stars

Watchers

0 watching

Forks

27 forks
Report repository

Releases

Packages

Contributors

Languages