FavourOgboi/HAKI-Agricultural-Intelligence-Platform

HAKI is a role-based agricultural intelligence platform for Zimbabwe farming communities. # HAKI Agricultural Intelligence Platform --- haki-frontend.onrender.com HAKI is a role-based agricultural intelligence platform for Zimbabwe farming communities. It combines: - A fieldworker assessment workflow - A manager dashboard for risk and performance monitoring - A researcher query and data exploration experience - A FastAPI backend for analytics and LLM-assisted workflows - Supabase for authentication and operational data storage ## Project Structure - src/App.tsx: application shell, routing, auth, and role-based navigation - src/pages/fieldworker/NewAssessment.tsx: fieldworker assessment flow - src/pages/manager/ManagerDashboard.tsx: manager dashboard - src/pages/researcher/QueryPage.tsx: natural-language researcher query page - src/pages/researcher/DataExplorerPage.tsx: researcher data explorer - src/pages/OperationalPages.tsx: farmers, history, risks, reports, audit, and export pages - haki_backend/main.py: FastAPI API routes - haki_backend/engine.py: analytics engine and dataset processing - haki_backend/llm.py: LLM layer for chat, extraction, parsing, and summaries - supabase/migrations/20260824012837_create_haki_core_tables.sql: database schema ## Main Features ### Fieldworker - Guided farmer assessment conversation - Structured field extraction from chat - Yield benchmark lookup - Insight generation from similar-farmer and benchmark data - Save farmer profiles and conversation history to Supabase ### Manager - Community dashboard with live aggregate metrics - Risk alerts - Community and yield map context - Reports and export support ### Researcher - Natural-language query interface - Privacy-safe aggregated results - Data Explorer with supported projected fields - Yield map exploration - Anonymised CSV export ## Current and Planned Intelligence Capabilities Currently implemented: - LLM-based field extraction from fieldworker conversations - Natural-language researcher query parsing - Similar-farmer and yield-benchmark insight g …