{% extends "base.html" %}
{% block title %}Code — Gordon{% endblock %}
{% block content %}

<section class="prose">
  <h1>Code</h1>
  <p class="lead">Gordon is a small, stock-stack FastAPI + Jinja + Plotly app that renders every
     chart, table, and download from the shipped data artifacts. The data layer is rebuilt
     deterministically from the pre-built source CSVs — no network calls, no random seeds.</p>

  <h2>Reproduce the site</h2>
  <pre><code>pip install -r app/requirements.txt
uvicorn app.main:app --port 8080
# then open http://localhost:8080</code></pre>

  <h2>Rebuild the data layer</h2>
  <p>Every chart artifact under <code>app/data/charts/</code> is produced by one script from the
     official BLS/BEA/Fed inputs:</p>
  <pre><code>python app/build_chart_data.py</code></pre>
  <p>The builder reads the pre-built source CSVs (CPI/PCE, employment, JOLTS, wage–price, productivity,
     and the heterodox block) and the FRED job-openings-rate series, collapses the sparse master
     joins to one row per period, and writes each chart as CSV + XLSX + Parquet plus a
     <code>chart_meta.json</code> describing sources, units, and limits. The same script closes
     three mechanical packaging gaps: it recomputes the CPI component-contribution decomposition
     through the latest month, sources the Beveridge job-openings rate, and derives
     service-providing employment as total-private minus goods-producing.</p>

  <h2>The code-toggle convention</h2>
  <p>Charts are drawn client-side from <code>/api/chart/&lt;key&gt;</code>, a columnar JSON feed,
     using the shared <code>ArkPlotly</code> standard (legend below, accent colorway, the top-right
     "Download data (CSV)" button). Client-side transforms (e.g. reindex-to-100 on index series)
     run in the browser, so the same artifact backs every view.</p>

  <p>
    <a class="ark-btn" href="/api/code/gordon_site_repro.zip" download>Download site reproduction bundle (.zip)</a>
    <br><span class="note">app source + data artifacts; runs offline.</span>
  </p>
  <p class="note">This bundle reproduces the <em>website</em> and its data layer from the pre-built
     source CSVs. The upstream <b>research code</b> that fetches the raw BLS/BEA/Federal-Reserve
     series and constructs the derived series is being packaged as a separate, standalone
     reproduction bundle and will publish with a future release.</p>
</section>

{% endblock %}