batestguy/crop-value-predictor

Offline-first crop value planning pilot for Nigeria with cloud-built data and forecasts. # Fieldmargin Crop Value Planner An offline-first, anonymous crop comparison tool for Nigerian farmers. The current UI is a runnable pilot-seed implementation of the delivery plan in `IMPLEMENTATION_PLAN.md`. ## Run locally ```bash npm install npm run dev ``` For a production build: ```bash npm run build python pipeline/validate.py ``` The PWA caches its static shell and snapshot files through `public/sw.js`. The current snapshot is intentionally labeled as seed data; replace it with the audited pipeline output before presenting automated forecasts as production evidence. ## Project layout - `src/` — responsive React/TypeScript decision interface and pure calculations - `public/data/v1/` — versioned static JSON interfaces - `pipeline/` — no-network snapshot validation and future source adapters - `IMPLEMENTATION_PLAN.md` — staged roadmap, source policy, and acceptance gates The browser never calls upstream agricultural APIs. User-entered costs and yield overrides remain local to the device, and recommendations include source, freshness, price type, uncertainty, and fallback context. ## Delivery progress Implementation follows the review-gated stages in `IMPLEMENTATION_PLAN.md`. The live status dashboard is `PROJECT_PROGRESS.md`, with an evidence record for each baseline and delivery stage under `docs/phases/`. The prototype release architecture is documented in `docs/batch-training-architecture.md`, and the active Stage 1 source register is `docs/source-register.md`. Cloud execution rules are documented in `docs/cloud-compute.md`. GitHub Actions is the authoritative environment for data processing, forecasting, validation, tests, and production builds; the browser performs only the small deterministic scenario calculation required for offline use.