TereraiC/RimAI-AI4I-2026

AI-powered agricultural intelligence platform for Zimbabwe — AI4I 2026 Track 3 # RimAI — Zimbabwe Agricultural Intelligence Platform An AI-powered agricultural intelligence platform combining machine learning yield/risk prediction, explainable AI, weather intelligence, and institutional decision support for Zimbabwe's farmers, AGRITEX extension officers, and Ministry of Agriculture. Built for the **2026 AI for Impact Challenge (AI4I) — Track 3: Development**. --- ## Architecture ``` Data Layer (real: FAOSTAT, NASA POWER, NOAA ENSO — synthetic: disclosed agronomic augmentation) │ ▼ Feature Engineering (rainfall anomaly, soil moisture proxy, NDVI proxy, ENSO phase, ...) │ ▼ Machine Learning Models (Ridge regression — yield · XGBoost — risk classification) │ ▼ Explanation Engine → Economic Impact Engine → Recommendation Engine │ ▼ Farmer / AGRITEX / Ministry / Admin Dashboards (Flask + Jinja2, one shared pipeline) ``` See the full technical design and AI-necessity rationale in the written proposal (`docs/` in the submission package, not this repo). ## Project Structure ``` rimai/ ├── app.py # Flask application: routes, auth, RBAC, DB schema ├── core/ # Prediction & reasoning engines │ ├── harvest_model.py Yield (Ridge) + risk (XGBoost) prediction, PROVINCE_META │ ├── crop_advisor.py Combines weather + rotation + pest + prediction into one analysis │ ├── explanation_engine.py Turns model output into cited "why" explanations │ ├── agronomy_engine.py Crop rotation and pest-risk rule logic │ ├── farm_manager.py Farm Health Score, Daily Brief, Smart Calendar, Farm Memory │ └── rimai_assistant_free.py Rule-grounded Virtual Agronomist chat (no external LLM) ├── data_pipeline/ # Data ingestion, synthetic data management, model validation │ ├── fetch_faostat.py Real FAOSTAT yield history (with disclosed fallback) │ ├── build_master_dataset.py Merges yield history with NASA POWER weather │ ├── weather_service.py Live …