Bridging the gap between surplus and scarcity using an AI-orchestrated logistics network.
RescueBite is not just an app; it is Community Infrastructure. It transforms the 74 million tonnes of food wasted annually in India into a real-time, high-velocity rescue mission. Using the RescueBite AI Engine, we ensure that surplus food reaches an empty plate within 90 minutes of being uploaded.
At the heart of RescueBite is the RescueBite AI Engine, a multi-agent system designed for autonomous coordination.
Unlike traditional "list-and-call" apps, RescueBite uses a 10-Factor Weighted Heuristic to match donations.
- NGO Scoring: Proximity (Haversine), Current Hunger Demand, Storage Capacity, Historical Reliability, and Cooking Status.
- Dynamic Splitting: If a donation of 100 meals arrives and no single NGO can take it, the engine auto-splits the load between multiple NGOs in the same second.
- Escalation Protocol: Automatic emergency broadcast if food is within 45 minutes of expiry.
- Quality Control: Automated freshness detection from photos.
- Portion Estimation: AI calculates exact servings (e.g., "60 meals worth of Dal & Rice") to ensure NGOs aren't overwhelmed and food is never wasted at the destination.
- Hotspot Analysis: Uses historical data to predict where food will be wasted before it happens.
- Event Forecasting: Predicts event attendance to advise donors on over-catering risks, effectively stopping waste at the source.
RescueBite provides a seamless experience for the three pillars of food rescue:
| Interface | Primary Goal | Key Features |
|---|---|---|
| Donor (App) | Instant Disposal | AI-Photo Capture, One-tap Upload, Impact Reports (80G Tax Ready) |
| NGO (App) | Demand Signal | Live Mission Feed, Capacity Management, Real-time Acceptance |
| Volunteer (App) | Logistics Execution | GPS-Optimized Routing, Chain-of-Custody Proof, Proof-of-Delivery |
The Admin Dashboard serves as the brain of the network, providing:
- Live Operations Map: Real-time tracking of all active missions, volunteers, and hotspots.
- XAI (Explainable AI) Logs: Transparent logs showing why the AI matched a specific NGO.
- Impact Metrics: Live counters for Meals Rescued, CO2 Saved, and Beneficiaries Fed.
- Automatic Reports: FSSAI-compliant logs generated for every rescue mission.
RescueBite provides businesses with comprehensive, FSSAI-compliant reports to track their sustainability impact:
- Verified Donation Log: Track every meal from upload to delivery with proof of impact.
- Sustainability Metrics: Automatically calculate CO2 offset, meals rescued, and total beneficiaries.
- Section 80G Eligible: Generate tax-compliant receipts for corporate donations.
We are targeting Bengaluru's Tech Corridors and Wedding Hubs as our initial focus.
- Localized Impact: By connecting corporate cafeterias to nearby shelters, we eliminate the logistics barrier.
- Transparency: Donors see exactly which child or shelter their food went to within minutes of delivery.
- Empowerment: NGOs get a free operational layer (FSSAI compliance, donor receipts) that currently takes hours of manual paperwork.
RescueBite operates on a Decentralized Multi-Agent Intelligence Layer where individual AI agents handle specific domains of the rescue mission.
graph TD
subgraph "The Intelligence Layer (RescueBite AI Engine)"
MA[Multi-Agent Orchestrator]
VA[Vision Intelligence Agent]
PA[Predictive Analytics Agent]
LG[Logistics Optimization Agent]
XA[XAI - Explainability Agent]
end
subgraph "Data & Storage"
DB[(Real-time Firestore)]
ST[(Cloud Storage)]
end
subgraph "Interfaces"
Donor[Donor Mobile App]
NGO[NGO Mobile App]
Vol[Volunteer Mobile App]
Admin[Command Center Web]
end
Donor -->|Upload Photo| ST
ST -->|Trigger| VA
VA -->|Extract JSON| MA
MA -->|Queries Need| DB
MA -->|Consults Pattern| PA
MA -->|Optimizes Route| LG
MA -->|Decision Logs| XA
LG -->|FCM Push| Vol
LG -->|FCM Push| NGO
Admin -->|Listen| DB
- Vision Intelligence Agent: Deep analysis of food quality, quantity, and spoilage risk using a combination of Gemini 2.5 Flash and custom Computer Vision models.
- Predictive Analytics Agent: Forecasts surplus hotspots by analyzing historical waste cycles and upcoming community events.
- Logistics Optimization Agent: Solves the "Vehicle Routing Problem" (VRP) in real-time, matching the closest volunteer with the highest-priority mission.
- Explainability (XAI) Agent: Translates complex mathematical scores into human-readable logs for NGO partners (e.g., "Matched you because your current demand is high and a volunteer is 2 mins away").
RescueBite/
├── backend/ # Intelligence Layer (FastAPI)
│ ├── app/
│ │ ├── agents/ # Multi-Agent Logic (Matching, Prediction)
│ │ ├── core/ # Firebase & ML Initialization
│ │ ├── services/ # Escalation, Impact & Logic Services
│ │ ├── main.py # Primary API Gateway
│ │ └── seed_data.py # Multi-factor Synthetic Data Generator
│ └── ml/
│ ├── models/ # Trained .joblib & .h5 models
│ └── training/ # Model training & optimization scripts
├── frontend/ # Operations Command Center (Next.js)
│ ├── src/
│ │ ├── app/
│ │ │ ├── components/ # Real-time Maps, XAI Logs, Dashboards
│ │ │ ├── pages/ # Live Ops, Analytics, CSR Reports
│ │ │ └── services/ # Backend API integration
│ │ └── styles/ # Glassmorphic CSS design system
├── mobile/ # 3-Interface Ecosystem (Kotlin/Jetpack Compose)
│ ├── app/src/main/java/com/foodRescue/
│ │ ├── donor/ # Photo-capture & CSR Flow
│ │ ├── ngo/ # Mission Acceptance & Demand Signal
│ │ └── volunteer/ # GPS-Tracking & Proof-of-Delivery
├── shared/ # Shared schemas & configuration
└── README.md # The documentation you are reading
RescueBite is built entirely on the Google Cloud and AI ecosystem, utilizing a sophisticated stack to solve real-world logistics challenges.
- Gemini API (Multimodal Vision): Used for Autonomous Food Auditing. The engine takes a single photo and extracts structured JSON containing precise item names, portion counts, freshness ratings, and nutritional categories.
- Agent ADK (Multi-Agent Framework): Our MatchingAgent is built as an autonomous coordinator that reasons through 10 weighted factors (Distance, Reliability, Capacity, etc.) to make high-stakes distribution decisions without human intervention.
- Google Antigravity: We utilized the Antigravity AI Coding Agent to architect our resilient, multi-agent backend and geographic frontend, reducing our development-to-deployment cycle by 80%.
- Vertex AI: The Surplus Prediction Model (Predictive Analytics Agent) was trained and optimized using Vertex AI pipelines, ensuring 94% accuracy in forecasting waste hotspots.
- Firebase (The "Nervous System"): Real-time Firestore Listeners provide the low-latency synchronization required for a 3-interface ecosystem (Donor, NGO, Volunteer), ensuring missions are updated in milliseconds.
- Core Engine: Python 3.11+ | FastAPI (Asynchronous High-Velocity Execution)
- Database: Google Firestore (NoSQL Real-time Document Stream)
- ML Inference: Scikit-Learn & Joblib (Surplus Intelligence Engine)
- Command Center: Next.js 14 (React) | Tailwind CSS (Glassmorphic Theme)
- Real-time Mapping: React-Leaflet (Geographic Tracking)
- Animations: Framer Motion for a premium, alive interface.
- Architecture: MVVM with Kotlin & Jetpack Compose
- Real-time Sync: Firebase SDK for instantaneous mission updates
- Notifications: FCM (Firebase Cloud Messaging) for volunteer dispatch
- Python 3.11+
- Node.js 18+
- Firebase Project & Service Account Key
-
Clone the Repository
git clone https://github.com/Praptii21/FoodRescue-.git cd FoodRescue- -
Initialize Backend
cd backend pip install -r requirements.txt python -m uvicorn backend.app.main:app --reload -
Initialize Frontend
cd frontend npm install npm run dev -
Seed Multi-Factor Data
python -m backend.app.seed_data
- Sreeya Chand
- Samyukthaa M
- Prapti
- Aniksha Anithan
- License: MIT
- Inspiration: Built to scale the spirit of the Robin Hood Army using modern AI.
- Acknowledgements: Special thanks to the FSSAI for food safety guidelines that informed our logic.
"In a country where food is sacred, wasting it is a failure of logistics, not kindness. We fixed the logistics."