Key evidence

A batch pipeline powers a static website.

Retrieval, validation, aggregation, evaluation, forecasting, and export happen before deployment.

Dataset source through27 Jul 202627 Jul 2026

No request-time model or Python API is required.

No account, database, analytics tracker, or browser secret is used.

Partitioned JSON keeps each market detail view focused on the data it needs.

  1. 1AGMARKNET 2.0Fetch public AGMARKNET monthly market price and arrival reports.
  2. 2Validate recordsCheck identity, dates, units, price order, duplicates, ranges, and market references.
  3. 3Aggregate observationsCombine variety reports into documented market-day records without filling missing days.
  4. 4Evaluate forecasting approachesScore simple and complex methods only on later, unseen periods.
  5. 5Publish prepared website dataWrite a small manifest and state/commodity partitions for static delivery.
  6. 6Next.jsBatch pipeline, static website, refresh workflow, and failure boundary.

Technical report

Detailed notes, interpretation, and operational boundaries

This section documents what the published evidence means, how it was produced, and where the interface deliberately avoids making a stronger claim than the data supports.

01

Data preparation layer

A Python pipeline retrieves, validates, normalizes, aggregates, evaluates, forecasts, and exports versioned artifacts.

  • Columnar Parquet files retain full prepared history and evaluation outputs efficiently.
  • The model bundle, report JSON, human-readable reports, data-quality ledger, and public manifest are generated from the same run.
  • Pipeline stages fail when required training or forecast rows are absent instead of publishing an empty success state.
02

Published data contract

The web layer consumes one compact manifest plus state-and-commodity JSON partitions rather than downloading the entire history on every route.

  • Each partition declares market, observation, forecast, and seasonal records with its date range, byte size, and checksum.
  • Stable slugs and market identifiers make routes shareable while visible labels preserve the official market names.
  • Observed records and forecasts remain separate types so an estimate cannot silently replace a source observation.
03

Application layer

Next.js renders crawlable route content while focused client components handle filters, comparison state, copying, and downloads.

  • Market and technical pages read prepared artifacts on the server; browser-side filtering operates on the published catalogue.
  • URL parameters preserve market filters, selected comparisons, dates, quantities, and cost methods for reproducible shared views.
  • Charts include text descriptions and expandable tables so visual summaries are not the only way to inspect values.
04

Operational and trust boundaries

The static release design favors reviewable artifacts, deterministic builds, and low runtime complexity.

  • A new data refresh changes the generated manifest and partitions; the UI does not mutate source history at runtime.
  • Language routes reuse the same numerical contract so translation cannot change prices, dates, or evidence counts.
  • Deployment readiness depends on passing pipeline tests, application tests, type checks, lint, and a production build against the published artifacts.

Related pages