docs: update README to reflect changes in model architecture and JSON…

[Prateekkp]

This pull request makes significant updates to the README.md to reflect a major architectural shift from an API-based model serving approach to a file-based, static JSON artifact serving pattern. The documentation now describes a pipeline where models are trained offline, forecasts are generated as JSON files, and the React/Next.js frontend consumes these artifacts directly, removing the need for a live prediction API. The changes also clarify the directory structure, pipeline stages, and the contract for forecast and diagnostics JSON files.

The most important changes are:

Architecture & Serving Model Updates

• The system now uses offline-trained models with forecast JSON artifacts consumed by a Next.js dashboard, replacing the previous Flask API-based serving approach. The frontend reads static JSON files directly, and a sync script ensures artifacts are available. [1] [2] [3] [4] [5] [6]

Pipeline & Directory Structure

• The pipeline stages and directory structure are updated to include new scripts (train_and_save_xgboost.py, pre_generate.py), support for both XGBoost and LSTM models, and clear separation of model and public data artifacts. The documentation now details where each artifact is stored and how the frontend accesses them. [1] [2] [3] [4]

Frontend & Automation

• The frontend is now a Next.js/React application (gridcast-react/) that loads forecast and diagnostics data from static JSON files. A new sync script (sync-real-data.mjs) is introduced to automate mirroring of artifacts to the frontend. [1] [2] [3] [4]

Forecast & Diagnostics Artifacts

• The documentation introduces a new contract for forecast and diagnostics JSON files, specifying their structure and contents for 24h, 48h, and 72h horizons, as well as metrics and residuals. [1] [2] [3]

Supporting Documentation & Research

• References and links to supporting documentation, research, and design files are reorganized and clarified, reflecting the updated codebase and project structure. [1] [2] [3]

These updates ensure the documentation accurately describes the new artifact-based serving pattern, improves clarity for onboarding, and provides a reliable reference for both pipeline operation and frontend integration.… artifact publishing

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
grid-cast	Ready	Preview, Comment	May 7, 2026 3:52am