Skip to content

Ahmed-fall/Silo

BranchesTags

Repository files navigation

🌾 Silo

النظام الذكي للحفاظ على الحبوب الوطنية

Intelligent National Grain Preservation System

Next.js FastAPI TensorFlow XGBoost PostgreSQL Docker Ollama

Egypt loses 30–40% of its grain every year to spoilage, disease, and poor storage conditions. Silo is the early warning system that changes that.

Features · Architecture · Quick Start · Docker · API Reference · Team

What is Silo?

Silo is an enterprise-grade, AI-powered grain silo management platform built for national agricultural operators. It unifies live sensor telemetry, AI grain disease diagnostics, soil health analysis, predictive risk scoring, geospatial monitoring, and a conversational AI agronomist into a single command center for Egyptian grain storage facilities.

Mission: Transform raw silo data into actionable intelligence — before spoilage, pest infestations, or environmental hazards cause irreversible loss.

Features

🖥️ Command Center Dashboard

Real-time overview of all monitored silos with animated risk-reactive cards. Each card shows live temperature, humidity, risk level, and crop type. A stats row at the top displays total silos, at-risk count, and nominal count at a glance. Falls back to realistic mock data if the backend is offline — presentations never crash.

🗺️ Geospatial Command Map

Interactive SVG map of Egypt plotting every silo's live risk status. 30+ Egyptian governorate coordinates are calibrated manually. Each silo appears as a color-coded pulsing node — Turquoise (Nominal), Bronze (Medium), Red (Critical) — with pulse speed increasing with risk severity. Click any node to open a slide-over detail drawer without leaving the map.

🏭 Silo Diagnostics Page

Per-silo full diagnostic view including: animated risk badge, live sensor widgets (temperature, humidity, fill percentage), 24-hour sensor history chart with 12-hour AI forecast overlay, thermal digital twin cylinder, alert history, AI vision scanner, and a PDF report export button.

📈 Sensor History & AI Forecast Chart

Dual-line Recharts chart showing temperature and humidity history with SVG gradient strokes. A dashed 12-hour AI forecast overlay extends from the last reading using XGBoost trend-adjusted extrapolation. Custom tooltip distinguishes historical readings from AI predictions with a 🔮 badge.

🤖 AI Vision — Grain Disease Detection

Drag-and-drop grain image upload with laser scanning animation. The TensorFlow/Keras CNN classifies the image across 14 disease categories and returns a detected label with confidence score. Results are stored per silo with full history.

14 disease classes: Aphid, Black Rust, Blast, Brown Rust, Fusarium Head Blight, Healthy Wheat, Leaf Blight, Mildew, Mite, Septoria, Smut, Stem Fly, Tan Spot, Yellow Rust

🌱 AI Soil Analysis

Dedicated soil health analysis service. Upload a soil image from the Soil Analysis page and receive an AI-powered classification of soil health status. Accessible directly from the sidebar navigation.

💊 Disease Treatment Protocols

Full treatment guide for every detectable disease. The Diseases page presents tiered treatment protocols — what the disease is, severity level, and exactly what action to take. Treatments are searchable and expandable by tier. Directly connected to the AI Vision detection flow.

💬 AI Agronomist Chatbot

Floating chat assistant powered by LLaMA 3.1 (via Ollama). Before every message, the backend queries the live silos database and injects current silo states into the system prompt — giving the model real context about your actual facility. Supports multi-turn conversation with full history. Quick suggestion chips on first open.

⚡ Real-Time Alert System

WebSocket-powered live alert broadcasting. When a high-risk sensor reading is ingested, the backend automatically creates an alert and pushes it to every connected operator simultaneously. A morphing Dynamic Island pill at the top of every page shows live alerts. Auto-reconnects with 5-second retry if connection drops.

📄 PDF Report Export

Every silo diagnostics page exports as a clean, print-ready A4 report via window.print() with full print CSS overrides. Useful for government inspectors and agricultural supervisors who need physical documentation.

⚙️ Settings & Compact Mode

Slide-over settings panel with compact grid mode toggle and alert mute toggle. All settings applied instantly via React Context with no page reload.

Architecture

Browser (Next.js 16.2 + TypeScript)
         │
         ├── REST (Axios, 2s timeout)
         ├── WebSocket (/ws/alerts)
         └── POST /chat
              │
        ┌─────▼──────────────────────────────┐
        │     FastAPI Backend (:8000)         │
        │     asyncpg · Alembic · Pydantic v2 │
        └─────┬──────────────────────────────┘
              │
    ┌─────────┼──────────────────────┐
    │         │                      │
    ▼         ▼                      ▼
PostgreSQL  AI Vision (:8001)   AI Predictive (:8002)
(silo_db)   TensorFlow/Keras    XGBoost Classifier
            14-class CNN        89% accuracy
                                      │
                             AI Soil (:8003)
                             Soil Health Classifier
                                      │
                             Ollama LLM (:11434)
                             LLaMA 3.1 8B

All frontend requests route exclusively through the FastAPI backend. The frontend never communicates directly with the database, AI services, or Ollama.

Services

Service Port Technology Responsibility
Frontend 3000 Next.js 16.2, TypeScript, Tailwind CSS v4, Framer Motion UI, WebSocket client, AI chat
Backend 8000 FastAPI, asyncpg, Alembic, Pydantic v2, httpx API gateway, DB, WebSocket hub, AI orchestration
AI Vision 8001 TensorFlow 2.17, Keras 14-class grain disease image classification
AI Predictive 8002 XGBoost, scikit-learn, joblib Sensor-based risk scoring and forecast
AI Soil 8003 FastAPI, PyTorch Soil health image classification
Database 5432 PostgreSQL 15+ Persistent storage — silos, sensors, alerts, images
Ollama 11434 LLaMA 3.1 8B Conversational AI agronomist with live DB context

AI Models

AI Vision — Grain Disease Classifier

Property Detail
Framework TensorFlow 2.17 / Keras CNN
Input 255×255 RGB image, normalized to [0,1]
Validation Accuracy 78.22%
Classes 14 (Aphid, Black Rust, Blast, Brown Rust, Fusarium Head Blight, Healthy Wheat, Leaf Blight, Mildew, Mite, Septoria, Smut, Stem Fly, Tan Spot, Yellow Rust)
Training Dataset Kaggle Wheat Disease Dataset

AI Predictive — Risk Scorer

Property Detail
Framework XGBoost + scikit-learn Pipeline
Input Features Temperature, Humidity, Soil Moisture, NDVI
Test Accuracy 89%
Cross-Validation 5-fold mean 90.4%, std 0.0047
Output Critical → high (score 75), Stressed → medium (45), Healthy → low (15)
Inference Latency 60ms end-to-end

AI Chatbot

Property Detail
Engine Ollama (local inference)
Default Model llama3.1:8b
Pattern RAG-lite — live silo DB state injected into system prompt per query
Timeout 120 seconds

Quick Start

Prerequisites

  • Python 3.11
  • Node.js v20+
  • PostgreSQL 15+
  • Ollama
  • conda / mamba

1. Clone

git clone https://github.com/Ahmed-fall/Silo.git
cd Silo

2. Python environment

mamba create -n silo python=3.11 -y
mamba activate silo

pip install fastapi==0.111.0 uvicorn[standard]==0.29.0 asyncpg==0.29.0 \
    pydantic==2.7.1 pydantic-settings==2.2.1 python-dotenv==1.0.1 \
    httpx==0.27.0 python-multipart==0.0.9 alembic==1.13.1 \
    psycopg2-binary==2.9.9 email-validator==2.3.0 \
    tensorflow==2.17.0 keras==3.13.2 pillow==10.4.0 \
    xgboost scikit-learn joblib numpy==1.26.4 requests websockets

3. Frontend

cd frontend && npm install && cd ..

4. Environment

cp .env.example .env
# Edit .env with your values

Required variables:

POSTGRES_USER=silo_user
POSTGRES_PASSWORD=silo_pass
POSTGRES_DB=silo_db
POSTGRES_HOST=localhost
POSTGRES_PORT=5432
BACKEND_HOST=0.0.0.0
BACKEND_PORT=8000
AI_VISION_URL=http://localhost:8001
AI_PREDICTIVE_URL=http://localhost:8002
AI_SOIL_URL=http://localhost:8003
OLLAMA_URL=http://localhost:11434
OLLAMA_MODEL=llama3.1:8b
UPLOADS_DIR=/absolute/path/to/Silo/uploads

5. PostgreSQL

sudo -u postgres psql -c "CREATE USER silo_user WITH PASSWORD 'silo_pass';"
sudo -u postgres psql -c "CREATE DATABASE silo_db OWNER silo_user;"
sudo -u postgres psql -c "GRANT ALL PRIVILEGES ON DATABASE silo_db TO silo_user;"
sudo -u postgres psql -d silo_db -c "GRANT ALL ON SCHEMA public TO silo_user;"

cd backend && alembic upgrade head && cd ..

6. Ollama

OLLAMA_HOST=0.0.0.0 ollama serve &
ollama pull llama3.1:8b

7. Model files

Place these manually (not tracked in Git due to size):

ai-vision/app/Final_model.keras
ai-predictive/app/model/weights/crop_health_model_xgb.pkl
ai-predictive/app/model/weights/label_encoder.pkl
ai-soil/app/model/weights/<your-soil-model>.pth

8. Seed & run

python scripts/seed_data.py

Open 6 terminals:

# Terminal 1 — Backend
cd backend && uvicorn app.main:app --host 0.0.0.0 --port 8000 --reload --env-file ../.env

# Terminal 2 — AI Vision
cd ai-vision && uvicorn app.main:app --host 0.0.0.0 --port 8001 --reload

# Terminal 3 — AI Predictive
cd ai-predictive && uvicorn app.main:app --host 0.0.0.0 --port 8002 --reload

# Terminal 4 — AI Soil
cd ai-soil && uvicorn app.main:app --host 0.0.0.0 --port 8003 --reload

# Terminal 5 — Frontend
cd frontend && npm run dev

# Terminal 6 — Sensor Simulator (optional)
python scripts/simulate_sensor.py

Open http://localhost:3000

Docker

The entire stack runs with one command:

docker compose up --build

Note: Ollama runs on the host machine. Start it before running Docker:

OLLAMA_HOST=0.0.0.0 ollama serve &

On Linux, the Docker network gateway is typically 172.18.0.1. The docker-compose.yml is configured to reach Ollama at that address.

Run migrations and seed after first startup:

docker compose exec backend alembic upgrade head
# Seed from host machine:
POSTGRES_HOST=localhost python scripts/seed_data.py

Useful commands:

docker compose up --build        # Build and start everything
docker compose up -d             # Run in background
docker compose logs backend      # View backend logs
docker compose logs ai-predictive
docker compose down              # Stop everything
docker compose down -v           # Stop and delete database volume

Verify All Services

curl http://localhost:8000/health   # {"status":"ok","service":"backend"}
curl http://localhost:8001/health   # {"status":"ok","service":"ai-vision"}
curl http://localhost:8002/health   # {"status":"ok","service":"ai-predictive"}
curl http://localhost:8003/health   # {"status":"ok","service":"ai-soil"}
curl http://localhost:11434/api/tags  # Ollama model list

API Reference

Base URL: http://localhost:8000

Silos

Method Endpoint Description
GET /silos List all silos with latest sensor and risk data
GET /silos/{id} Get single silo detail
POST /silos Create a new silo

Sensors

Method Endpoint Description
POST /sensors/ingest Ingest reading → AI prediction → alert if high risk → WebSocket broadcast
GET /sensors/{silo_id} Get sensor history
GET /sensors/forecast/{silo_id} Get 12-hour AI forecast

Alerts

Method Endpoint Description
GET /alerts/{silo_id} List alerts for a silo
PATCH /alerts/{alert_id}/read Mark alert as read

Images & Vision

Method Endpoint Description
POST /images/upload?silo_id={id} Upload grain image → AI disease classification
GET /images/{silo_id} Get image scan history

Soil Analysis

Method Endpoint Description
POST /soil/analyze Upload soil image → AI soil health classification

Treatments

Method Endpoint Description
GET /treatments Get all disease treatment protocols
GET /treatments/{disease} Get treatment protocol for a specific disease

Chat

Method Endpoint Description
POST /chat Send message to AI agronomist (live DB context injected)

WebSocket

Channel URL Description
Alerts ws://localhost:8000/ws/alerts Real-time alert push to all connected clients

Project Structure

Silo/
├── .env                              # Environment config (not in Git)
├── .env.example                      # Environment template
├── docker-compose.yml
├── backend/
│   ├── app/
│   │   ├── main.py                   # FastAPI app, CORS, WebSocket, health
│   │   ├── api/
│   │   │   ├── silos.py
│   │   │   ├── sensors.py            # Ingest pipeline + forecast
│   │   │   ├── alerts.py
│   │   │   ├── images.py
│   │   │   ├── treatments.py         # Disease treatment protocols
│   │   │   └── chat.py               # Ollama + live DB context
│   │   ├── core/
│   │   │   ├── config.py
│   │   │   └── database.py
│   │   ├── data/
│   │   │   └── treatment_protocols.py  # 14-disease treatment data
│   │   ├── models/
│   │   │   ├── silo.py
│   │   │   ├── sensor.py
│   │   │   ├── alert.py
│   │   │   ├── image.py
│   │   │   └── treatment.py
│   │   └── ws/alerts.py
│   ├── migrations/
│   └── requirements.txt
├── ai-vision/
│   ├── app/
│   │   ├── main.py
│   │   └── Final_model.keras         # Not in Git — place manually
│   └── requirements.txt
├── ai-predictive/
│   ├── app/
│   │   ├── main.py
│   │   └── model/
│   │       ├── predict.py
│   │       └── weights/
│   │           ├── crop_health_model_xgb.pkl   # Not in Git
│   │           └── label_encoder.pkl           # Not in Git
│   └── requirements.txt
├── ai-soil/
│   ├── app/
│   │   └── main.py
│   └── requirements.txt
├── frontend/
│   ├── app/
│   │   ├── page.tsx                  # Command Center Dashboard
│   │   ├── live-map/page.tsx         # Geospatial Command Map
│   │   ├── silos/[id]/page.tsx       # Silo Diagnostics
│   │   ├── diseases/page.tsx         # Disease Treatment Protocols
│   │   └── soil-analysis/page.tsx    # Soil Analysis
│   ├── components/
│   │   ├── Sidebar.tsx
│   │   ├── SiloCard.tsx
│   │   ├── SensorChart.tsx
│   │   ├── ThermalSiloMap.tsx
│   │   ├── AIVisionScanner.tsx
│   │   ├── AISoilScanner.tsx
│   │   ├── LiveAlertsPanel.tsx
│   │   ├── ChatBot.tsx
│   │   ├── SettingsDrawer.tsx
│   │   ├── SplashScreen.tsx
│   │   └── WheatParallaxField.tsx
│   ├── context/
│   │   ├── AlertContext.tsx
│   │   └── SettingsContext.tsx
│   └── lib/api.ts
└── scripts/
    ├── seed_data.py
    └── simulate_sensor.py

Known Limitations

  • No authentication — all endpoints are publicly accessible
  • No rate limiting on sensor ingestion or image upload
  • Chatbot responds in English only
  • Model files must be placed manually (not tracked in Git due to size)
  • Ollama requires separate host setup — not containerized

Troubleshooting

Issue Cause Fix
permission denied for table alembic_version PostgreSQL 15+ schema permissions Run GRANT ALL ON SCHEMA public TO silo_user;
Backend can't connect to DB in Docker POSTGRES_HOST=localhost in .env Override with POSTGRES_HOST=db in docker-compose.yml environment
Ollama unreachable from Docker on Linux host.docker.internal not supported on Linux Use gateway IP 172.18.0.1 and allow firewall: sudo ufw allow from 172.18.0.0/16 to any port 11434
Frontend exits silently Corrupt node_modules rm -rf node_modules && npm install
Module not found: sklearn Missing dependency pip install scikit-learn xgboost
Port already in use Previous service still running sudo fuser -k 8000/tcp (replace port as needed)

Team

Built by the Silo Engineering Team Egypt-Japan University of Science and Technology (E-JUST) Supported by nWeave

Built for Egypt. By Egyptian engineers.

About

A Smart intelligent system that identifies and recognizes the pattern of spoilage for Egypt plants.

Resources

Readme
Activity

Stars

1 star

Watchers

0 watching

Forks

2 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages